Use cases for Generative AI: View as single page

Printable page generated Sunday, 23 November 2025, 1:37 AM
Use 'Print preview' to check the number of pages and printer settings.
Print functionality varies between browsers.

Unless otherwise stated, copyright © 2025 The Open University, all rights reserved.
Printable page generated Sunday, 23 November 2025, 1:37 AM

Use cases for Generative AI

1 Introduction

ChatGPT 4 generated image prompt: ‘Draw an image that can be used in course materials that explores GenAI use cases’

This course explores use cases for GenAI – depending on your time commitments, you may choose to study the course in separate parts, as shown below.

Course sessions

The sessions are:

Session 1: Introduces general use cases for GenAI – 90 minutes (Sections 1, 2, 3 and 4).
Session 2: Focuses on how AI solutions are being applied within the advice and legal sectors – 90 minutes (Sections 5, 6, 7 and 8).

You can work through both sections in order or, depending on your role and interests, you might prefer to focus on just one.

This is a self-paced course of around 150–180 minutes including a short quiz.

A digital badge and a course certificate are awarded upon completion. To receive these, you must complete all sections of the course and pass the course quiz.

An introduction to the course

Timing: Allow 2 minutes

Listen to this audio clip in which Francine Ryan provides an introduction to the course.

Download this audio clip.Audio player: course_4_intro.mp3

Show transcript|Hide transcript

Transcript

FRANCINE RYAN:

Welcome to course four, Use cases for Generative AI. This course explores use cases for Generative AI. The first part considers general use cases and the second part highlights some examples of where AI solutions are being applied in the advice and legal sector. There is growing interest in how organisations can harness GenAI to gain competitive advantage. Large Language Models (LLMs), in particular, present significant opportunities to enhance productivity and streamline repetitive tasks. By automating routine processes, LLMs have the potential to free up capacity and allow teams to focus on higher-value work. LLMs are not infallible – they can produce errors, and their outputs still require human oversight. Moreover, there is a risk that over-reliance on these tools could lead to de-skilling within organisations. As such, while GenAI presents exciting possibilities, its integration must be approached thoughtfully and strategically. We really hope you enjoy this course.

End transcript

Download

Interactive feature not available in single page view (see it in standard view).

You can make notes in the box below.

To use this interactive functionality a free OU account is required. Sign in or register.

Interactive feature not available in single page view (see it in standard view).

Discussion

LLMs present opportunities to enhance productivity and streamline repetitive tasks, as well as freeing up capacity and allowing teams to focus on higher-value work. But they can also produce errors, their outputs require human oversight, and these tools could lead to de-skilling within organisations. As such, AI's integration must be approached thoughtfully and strategically.

Learning outcomes

After completing this course, you will be able to:

Identify common use cases for GenAI and reflect on how these applications might be relevant to your own work or organisation.
Evaluate how AI technologies are being implemented within the advice and legal sectors.
Understand and apply a strategic framework for the adoption of AI tools.
Understand the ADKAR model of change management and how to apply it in your organisation.

Glossary

A glossary that covers all eight courses in this series is always available to you in the course menu on the left-hand side of this webpage.

Session 1: Introduces general use cases for GenAI – 90 minutes

2 AI personas

Image created in ChatGPT 4 using the promp: ‘Please can you create an image that is entitled AI personas and has these personas, The Maximalist, The Underground, The Rebel, The Superfan and The Observer. For each of the personas can you create an image and then write the persona next to it make it visually appealing and engaging for the reader.’

Slack’s Work Force Lab has conducted research exploring employees' attitudes toward AI, referring to these differing perspectives as ‘AI personas’. In the second course, Skills and strategies for using Generative AI, we used the term ‘AI personas’ to describe the roles or characters we ask AI to adopt when responding to prompts. In this context the term is used differently it reflects how individuals perceive, relate to, and engage with AI, based on their personal attitudes and experiences.

Slack suggests there are five AI personas:

The Maximalist
The Underground
The Rebel
The Superfan
The Observer

The research from Slack suggests that people have mixed feelings about AI. Organisations considering integrating AI tools should be aware of and address different attitudes towards AI.

What AI persona are you?

Timing: Allow 15 minutes

Take the quiz 5 Types of People Defining the AI-Powered Future of Work – Salesforce

Discussion

You may wish to reflect on your own AI persona and consider how it shapes your views and assumptions about AI. Exploring the other personas can also offer valuable insights – what perspectives might your colleagues identify with, and how do these differ from your own? Personas are a useful tool for understanding the diverse ways people engage with AI. Do they help surface underlying values, concerns and expectations?

3 General use cases for GenAI

In the first course, Understanding Generative AI, you learnt about GenAI technology and how general-purpose public large language models (LLMs) such as Claude, ChatGPT and Microsoft Co-Pilot can generate human-like text, engage in conversations, create images, write computer code and answer questions.

LLMs can be used to streamline workflows and assist with tasks that might help us become more productive and efficient. Keeping track of the number of GenAI products being released is challenging – at the time of writing, ChatGPT was still the most popular, but there are other models emerging.

As organisations begin to explore GenAI tools it is important to distinguish between ‘everyday’ AI tools (ChatGPT, Claude, etc.) and more specialist AI tools. Starting with the ‘everyday’ tools offers a practical first step before considering the adoption of more advanced or bespoke AI solutions.

This course does not recommend any specific GenAI tools. Instead, it aims to raise awareness of the wide range of tools available and to emphasise the importance of selecting the most appropriate one for your particular task and context.

In this section of the course we explore general use cases for LLMs.

AI is transforming the world of work, are we ready for it?

Timing: Allow 20 minutes

Watch this video ‘AI is transforming the world of work, are we ready for it?’ from the Financial Times (2024).

Interactive feature not available in single page view (see it in standard view).

Reflect on how organisations experiment and invest when things are changing at such a rapid pace.

Discussion

The film highlights that AI might be the biggest change to the workplace in our lifetime but identifies there is a lot of hype and concern. Organisations should think carefully about their needs. They should recognise the imperfections and limitations of the technology and critically assess where it might be appropriate in their business.

Before exploring the types of tasks that free and paid-for general-purpose public LLMs – such as Microsoft Copilot, DeepSeek or Gemini – can support, it is important to emphasise the need for responsible and ethical use of GenAI tools. It is also important to explain there are differences between publicly available GenAI tools (which can be both free or subscription based) and corporate versions, particularly regarding data usage policies.

Click on each label below to read more.

Active content not displayed. This content requires JavaScript to be enabled.

Show description|Hide description

Publicly available GenAI tools − Typically, user inputs are used to improve and train the model, and the tool stores conversations and inputs. The privacy policies typically grant broad rights to use the data, this means your data is being used to train the model. They come at a lower cost or are free and have less privacy protection. You should not put any sensitive personal or organisational information into standard public versions of LLMs. Corporate versions of GenAI tools − Typically used by organisations and businesses. The organisation pays a fee depending on the number of users. Corporate tools often come with a dedicated workspace to manage users at scale. Importantly, users’ data is not used to train or improve the model – the organisation retains control of the data, and the data is encrypted. Corporate tools often state that the organisation retains ownership of both inputs and outputs. However, due to several ongoing lawsuits, the issue of ownership remains complex and not entirely settled. Both ChatGPT Team and Microsoft Copilot offer corporate versions of their tools.

Interactive feature not available in single page view (see it in standard view).

As discussed in the other courses in this series, when using any LLM it is advisable to:

Read the terms and conditions.
Be aware of the legal implications of AI generated content.
Understand the rules on data protection, intellectual property and copyright.
Avoid inputting any sensitive personal or organisational information into standard public versions of LLMs.

Organisations should establish clear guidelines and policies outlining what information can be shared with LLMs and which tasks are appropriate for their use. Human oversight and critical thinking remain essential.

Over-reliance on LLMs can lead to de-skilling so it is important to maintain a balanced approach. LLMs should be viewed as collaborative tools that assist, rather than replace, human skills and expertise.

General use cases for LLMs

If LLMs are used ethically and responsibly, several general applications might be beneficial:

Language processing – real time translation, summarisation, information extraction.
Content creation – creative writing, text creation.
Conversational AI – virtual assistants, customer service chatbots, learning assistants.
Professional tools – code generation, legal document analysis, academic research.
Creative applications – music lyrics, poetry, scriptwriting.
Media application – video creation, audio creation.
Code generation – writing code, error detecting, de-bugging code.

Prompt

To sum up the potential use cases for LLMs we asked the free version of Claude 3.7 Sonnet to:

‘Write a concise 100-word paragraph highlighting the key use cases for Large Language Models in modern business environments. Focus on practical examples that deliver measurable value. Maintain a professional tone for executives.’

This is what it said:

Response

‘Large Language Models are revolutionizing how organisations operate across domains. In customer service, AI chatbots handle inquiries 24/7, reducing wait times and increasing satisfaction. Content teams use LLMs to draft blogs, reports, and marketing copy in seconds rather than hours. HR departments streamline recruitment by automatically screening resumes and generating personalized communications. Software developers accelerate coding with AI-powered assistants that suggest solutions and explain complex algorithms. Legal teams extract key provisions from contracts and identify potential risks automatically. Research departments synthesize findings from thousands of documents instantly. When implemented thoughtfully, LLMs boost productivity, enhance decision-making, and create competitive advantage through automation of knowledge work.’

How would you rate the answer? Claude produced this answer within seconds, much quicker than most of us could write a similar answer. It is slightly over the word count at 104 words, but it does provide a summary of use cases for LLMs in a professional tone.

In the second course, Skills and strategies for using Generative AI, you learnt about prompting frameworks and prompt libraries, using that knowledge you could consider using an LLM for:

Image created in ChatGPT Team using the prompt: ‘Please can you create a mind map, in the middle it has “You can use an LLM for” and then there is a line off to: “Content creation, Research assistance, Visual and audio support, Creative assistance and Productivity” can you make it visually appealing and engaging to the viewer’.

Content creation

Document drafting, first drafts of reports, papers and case studies.
Generating meeting agendas and detailed minutes and actions.
Drafting internal communications, such as newsletters, marketing materials, website content, blogs and policy documents.
Composing emails and summarising long email threads to extract key points and decisions.
Simplifying content or generate content in a specific tone or style.
Proofreading and suggesting improvements to the text.
Generating response templates for common questions or enquiries.
Creating presentations, slide decks, and content based on specific topics and audiences.
Generating speaker notes with talking points and transitions.
Suggesting data visualisation approaches for complex information.

Research assistance

Summarising lengthy research papers and reports into actionable insights.
Extracting relevant data points from multiple sources.

Examples of tools

Perplexity is a search engine that uses LLMs to answer queries that include sources from the web and cite links with the text response.
Elicit is an AI research assistant that researchers can use to support with literature reviews it searches academic papers from the Semantic Scholar corpus.
Paid versions of LLMs like ChatGPT now offer a ‘deep research’ function where the LLM browses the web, reads multiple sources and extracts relevant information, it then analyses that information to provide a report with citations and explanations. For more information on deep research, you can read this blog The Differences between Deep Research, Deep Research, and Deep Research.
NotebookLM is a personalised AI assistant, it is designed to help users better understand and interact with their own documents. NotebookLM allows you to upload sources (like Google Docs, PDFs or copied text), and then ask questions, generate summaries, or brainstorm ideas based on that content. Its key features include source-grounded responses – it only answers based on the documents you upload. It will produce auto-generated summaries, glossaries and FAQs, and it can help you compare and synthesise across different documents. It also allows you to generate an AI-powered audio conversation between two virtual hosts. They discuss the content you have uploaded, and the conversation is automatically generated based on your notes, documents or research materials.

Visual and audio support

Image generation: AI systems such as DALL-E or Midjourney can create images based on textual descriptions.
Audio content: podcast script creation, creating description narration for visual content to improve accessibility, interview question preparation and music composition.
Video content: script creation, storyboarding assistance, YouTube titles and descriptions, content ideation and video editing support.
Video post-production: automatic caption generation, content summarisation, thumbnail text generation, chapter marking and content tagging.

Creative assistance

Idea generation: product ideas, new service offerings.
Brainstorming: marketing strategies, event planning.
Overcoming creative blocks: fresh perspectives or re-framing problems.
Role-playing: using different personas to offer different viewpoints.
Feedback and re-fining: reviewing your work, giving constructive feedback, edits and suggestions.
Questioner: ask the LLM to suggest questions for you to consider on a topic.

Productivity

Meeting summaries or transcripts.
Training programmes, researching, and creating training.
Creating internal FAQs and checklists.
Translation of documents.
Redaction of documents.
Integrating into existing tools such as Outlook or add GenAI tools to browser extensions, for example, Chrome.

Different ways you can use GenAI tools

Timing: Allow 6 minutes

Listen to this audio clip in which Francine Ryan discusses the different ways you can use GenAI tools.

Download this audio clip.Audio player: course_4_podcast.mp3

Show transcript|Hide transcript

Transcript

Francine Ryan:

Welcome to the deep dive. We're here to figure out how generative AI tools can actually help boost productivity and creativity, whether you're working alone or, in a bigger organisation. So think of this as a quick way to get up to speed on the practical side of Generative AI, ideas and ways to really augment how you work.

So, let’s start with content creation. GenAI can help you brainstorm and help you with first drafts for things like reports or case studies, it can help tackle that fear of the blank page. Gen AI can give you a really solid starting point. You can also use it to draft agendas for meetings and then afterwards, it can generate detailed minutes with clear action items. You can use it for general communication, like internal newsletters or marketing materials, drafting website content, blogs, or even policy documents. It can compose emails or summarise long email chains.

GenAI tools can tailor content, they can take complex info and simplify it, or change the tone so you could write in a specific style for, say, different audiences. It can provide practical help, like proofreading, and suggesting improvements. Gen AI can help generate the main content for slides, even speaker notes with talking points. It can suggest ways to visualize data effectively make it easier for the audience to understand.

Gen AI isn't just text, moving beyond text, you've got image generation tools like DALLE or Mid-journey, where you type a description and get an image back. You have also got audio content, like podcasts, it can help write a script or tools like Otter.ai, Descript and Whisper will transcribe audio files for you. It can help you with YouTube titles and descriptions, and for post-production like automatic captioning, summarising the video's content, generating text for thumbnails. It can add chapter markers for navigation, tagging content.

If you're stuck brainstorming like new product ideas or marketing angles, it can offer different perspectives help you break through those creative blocks. You can ask it to role play, using it to test ideas from different viewpoints, getting diverse perspectives on your work. Plus, you can use it for feedback, constructive edits and suggestions, almost like having a built-in reviewer and you can even use it as a questioner, to make you think deeper about a topic, challenge your assumptions.

There are a lot of different ways you can use GenAI tools, but there's one really critical takeaway, it's crucial, no matter what you're using Gen AI for you have to know how to prompt it effectively, how to guide it, and then, importantly, how to critically look at what it gives you back. It's not magic. Evaluating the output is key. That's how you make it truly work for you.

Think about one task, just one thing you do regularly, where could Gen AI potentially lend a hand? Then maybe try experimenting. How would you prompt it to help with that specific task and then see what happens? And then think about how you could tweak your prompt to get an even better result and then reflect on whether it is helpful and whether it does save time.

End transcript

Download

Interactive feature not available in single page view (see it in standard view).

Think about one way you might use a GenAI tool in your work or in your organisation and how it might help you.

To use this interactive functionality a free OU account is required. Sign in or register.

Interactive feature not available in single page view (see it in standard view).

Discussion

There may be lots of ways in which you can use an LLM in your work or in your organisation. However you use an LLM you need to understand the legal and privacy implications as well as how to prompt it, guide it and think critically about its outputs.

We asked ChatGPT 4 (April 2025) for some suggestions of which LLMs should be used for different tasks. You can see below what is suggested.

ChatGPT prompt: ‘Can you create me a table for different LLMs for different tasks − the tasks are linked to content creation, research assistance, video and audio support and creative assistant’

Show description|Hide description

1. Content creation. Recommended LLMs/Tools. ChatGPT (Open AI), Gemini (Google), Claude (Anthropic), Notion AI. Strengths. Excellent at writing articles, blogs, emails, and editing drafts. 2. Research assistance. Recommended LLms/Tools. Perplexity AI, ChatGPT-4, Scite Assistant, Elicit.org. Strengths. Fast, accurate answers with sources. Great for summarising complex information. 3. Video and audio support. Recommended LLMs/Tools. Runway ML, Descript, Adobe Podcast, Eleven Labs. Strengths. Generates voiceovers, edits audio/video, and creates content with ease. 4. Creative Assistant. Recommended LLMs/Tools. ChatGPT, Gemini, Mid Journey (visuals), Poe (multi-LLM access. Strengths. Helps with brainstorming, storytelling, image generation, and creative writing.

ChatGPT prompt: ‘Can you create me a table for different LLMs for different tasks − the tasks are linked to content creation, research ...

You can check for updates and improvements to LLMs through their release notes – read the latest updates to ChatGPT and Google Gemini. The technology is moving fast, and you should always do your research into the tools that you think are appropriate for you or your organisation. This is explored further in the eighth course, Preparing for tomorrow: horizon scanning and AI literacy.

Using LLMs for different tasks

In the second course, Skills and strategies for using Generative AI, you learnt about CLEAR and CRISP for prompting – with the overarching prompting principles being:

Clarity: Use straightforward, unambiguous language.
Specificity: Narrow the topic or task to focus the response.
Context: Include relevant background (e.g. jurisdiction, timeframe, audience).
Purpose: State what the output is for (e.g. training, publication, client communication).
Format: Specify the desired structure (e.g. bullet points, summary, list).

LLMs for different tasks

Timing: Allow 15 minutes

Choose one or two of the following options. Copy the prompt into an LLM tool and generate a response. You may want to change your prompt using one of the prompt frameworks you learnt about in the second course, run the query again and compare the different outputs.

Option 1: Draft a professional email

Prompt to use: I am a [Job title] working for [organisation]. Write a professional email to a client, thanking them for attending your recent webinar on [topic] and offering them a follow-up meeting. Keep it professional and concise.

Option 2: Brainstorm a new product or service idea

Prompt to use: I am a [Job title] working for [organisation]. Brainstorm three creative new product or service ideas for a small [insert industry] business and create a table that I could include as part of a presentation.

You could customise the prompt, for example, brainstorm three creative new service ideas for an advice charity that is focused on supporting clients with family law advice.

Option 3: Generate social media post ideas

Prompt to use: I work for [organisation] as a [Job title]. Generate three catchy social media post ideas to promote an upcoming charity run event to raise funds for [Project]. Make the posts engaging and include a call to action.

Option 4: Create a LinkedIn post

Prompt to Use: I am a [Job title] and I gave a presentation on [Topic] to [Organisation]. Write a LinkedIn post about my recent presentation. Thank the audience for attending, highlight [key takeaway], and invite people to connect or ask questions. Keep the tone professional but approachable.

Reflection questions

Timing: Allow 15 minutes

Consider the quality and usefulness of the General use cases for GenAI that you've read about in this section.

1	How relevant were the ideas or suggestions? Rate 1 (poor) to 5 (very good)	To use this interactive functionality a free OU account is required. Sign in or register.
2	How creative or original did you find them? Rate 1 (poor) to 5 (very good)	To use this interactive functionality a free OU account is required. Sign in or register.
3	Would these ideas be practical or feasible in a real-world setting? Rate 1 (not really) to 5 (very usable)	To use this interactive functionality a free OU account is required. Sign in or register.
4	How much refinement would they need before you could use them? None / Minimal / Moderate / Significant	To use this interactive functionality a free OU account is required. Sign in or register.
5	How helpful was this for saving time or sparking new thinking? Not helpful / Slightly helpful / Very helpful	To use this interactive functionality a free OU account is required. Sign in or register.
6	What did the LLM do well in terms of brainstorming?	To use this interactive functionality a free OU account is required. Sign in or register.
7	Were there any ideas you found surprising or insightful?	To use this interactive functionality a free OU account is required. Sign in or register.
8	Would you use an LLM for brainstorming in your work? Why or why not?	To use this interactive functionality a free OU account is required. Sign in or register.
9	If you modified the prompt, how did that change the results?	To use this interactive functionality a free OU account is required. Sign in or register.

Interactive feature not available in single page view (see it in standard view).

Discussion

It is useful to rate the quality and usefulness of these exercises and any task you ask an LLM to complete, to ensure that, it is helping you to be more productive and efficient.

4 Tips and tricks

Image created by ChatGPT 4 with this prompt: ‘Can you create an image for my course that is tips and tricks on prompting an LLM so it is a reminder that quality varies by prompt, checking and editing is essential, experiment but be aware of its limitations and that it still needs to be checked by a human’

Show description|Hide description

Prompting an LLM. Tips and tricks. Quality varies by prompt, checking and editing is essential. Experiment but be aware of its limitations. it still needs to be checked by a human.

Image created by ChatGPT 4 with this prompt: ‘Can you create an image for my course that is tips and tricks on prompting an LLM so it is ...

Quality varies by prompt

The quality of the response depends on the clarity and detail of your prompt. As you learnt in the second course, Skills and strategies for using Generative AI, the more specific you are – about tone, audience, or purpose – potentially the better the output.

When crafting prompts it is essential to ensure they are straightforward and unambiguous. Avoid using vague language and strive to be as precise as possible. Clear prompts help the AI understand exactly what is being asked, leading to more accurate responses. Narrowing down the topic is crucial to avoid broad or general answers. Specify the exact information or type of response you are seeking. This focus helps the AI provide more relevant and detailed answers, tailored to your needs.

Providing relevant background information is key to guiding the AI. Context helps the AI understand the nuances of the query and generate more accurate responses. By offering sufficient context, you enable the AI to better grasp the situation and deliver high-quality answers. Clear and specific prompts lead to more accurate and relevant responses. When the AI understands the query precisely, it can provide answers that are more aligned with your expectations.

Well-crafted prompts save time by reducing the need for follow-up questions or clarifications. By being clear and specific from the outset, you minimise the chances of miscommunication and streamline the interaction. Providing context helps the AI understand the nuances of the query, resulting in better-quality answers. Contextual information allows the AI to consider various aspects of the situation, leading to more comprehensive and insightful responses.

Checking and editing is still essential

LLMs provide a good starting point, but the response still needs to be checked, edited and refined. This could be adding personal insight, refining tone or adjusting phrasing – your voice is still key. Remember you are responsible for the content you produce and ensuring that it aligns with your values and/or the values of your organisation.

Time-saving benefits

Using GenAI tools can help save time, particularly during the early stages of drafting or brainstorming. Even if the output requires refinement, starting with a first draft gives you something to edit. However, it’s important to remain cautious about over-reliance on these tools. Always critically evaluate whether the verification and editing process truly results in time saved. If we are going to use these tools regularly, we need to ensure they provide real efficiency gains.

Surprising creativity and sometimes not!

LLMs can offer different perspectives, but the content can be generic or miss the deeper nuance. We should think of LLMs as assistants, not replacements.

Think about how you refine your prompts and experiment with adding more context:

Who’s the audience?
What’s the goal of the piece?
What tone or style do you want?
Is there a prompt framework or prompt from a prompt library that could help you?

Iterate and experiment

Ask the LLM to try again or rewrite in a different style. Play around with instructions like:

Try tone descriptors, such as ‘make it more friendly but professional’.
Use a storytelling approach.
Focus on benefits instead of features.
Specify the length of the output.
Use action verbs such as ‘draft’ or ‘create’.

Use it for brainstorming, not just writing

LLMs can help you get unstuck, whether you are generating ideas for content, problem-solving or exploring different perspectives.

Be aware of its limitations

Remember that LLMs don’t have judgment, expertise, or lived experience. Fact-checking and ensuring alignment with your voice and values are essential. In the fifth course, Ethical and responsible use of Generative AI, you will learn more about responsible and ethical use of GenAI.

Useful tasks

LLMs can be useful for tasks such as proof reading – in the box below is an example of a prompt from Write with AI (Cole and Bush, 2025). This prompt can be used to ask an LLM to proofread a document. It is a very detailed prompt that reviews your document line by line and makes suggestions for change. Compare this to the prompts above and consider how much more detail is included.

Prompt

You are a perfectionist, meticulous proofreader GPT and an editor with a keen eye for detail and a mastery of the English language. Your goal is to thoroughly review and revise the provided draft text and suggest edits to improve clarity, flow, grammar, and overall quality of the document. Use English United Kingdom.

Follow this process to proofread and edit the draft text:

Step 1: Read through the entire draft to understand the overall message and structure before making any edits.

Step 2: Perform a detailed line-by-line edit, watching for:

Spelling, grammar and punctuation errors.
Awkward phrasing or sentence structure.
Redundant or unnecessary words and phrases.
Incorrect or inconsistent formatting.
Factual inaccuracies or unsupported claims.

Step 3: Suggest reordering sentences or paragraphs to improve the logical flow and coherence of the writing. Use transition words and phrases to link ideas.

Step 4: Provide recommendations to enhance the draft's overall impact and persuasiveness:

Strengthen word choice by replacing weak or vague terms with more powerful language.
Vary sentence length and structure to improve readability and keep the reader engaged.
Ensure the main points are clearly stated and well-supported.
Maintain a consistent voice and tone aligned with the purpose and intended audience.

For any major revisions, provide a brief rationale to help the author understand your thought process and learn for the future. Preserve the original author's voice and intent. Avoid making edits that change the core meaning.

Be respectful and constructive with feedback. The goal is to help the author improve, not to criticise.

Prioritise edits that have the greatest impact on clarity and persuasiveness of the writing.

Output format: Summary: [Provide a quick 2–3 sentence summary of the key points and overall message of the draft text] Mistakes/Errors: [List out all the mistakes and errors you observed in the draft text, including spelling, grammar, punctuation, formatting, factual inaccuracies, awkward phrasing, etc.] Revised Draft: [Insert the full edited and proofread text here, with all the mistakes corrected and suggestions implemented. Preserve as much of the original formatting as possible.] Detailed Edit Notes: [Use this section to provide a more detailed explanation of the edits you made and your reasoning behind them. Reference specific line numbers where these are helpful. Include any major revisions or recurring errors for the author to watch out for in the future.

The value of this prompt is that it is using an LLM as an assistant, not as a replacement. If you treat the AI like a human rather than as a tool you can get a better response.

This prompt is encouraging the user to learn and gain a better understanding of the structure of language. This might be an example of where an LLM could add value to an individual or an organisation. You might want to try this prompt and see how effective it is. You could also start your own prompt library where you keep a record of your prompts for the different tasks you use an LLM for, prompts can be both written and audio.

On YouTube there are lots of videos about creating effective prompts for the types of tasks that might be appropriate for a public LLM and the second course, Skills and strategies for using Generative AI, provided links to prompt libraries in the ‘Further reading’ section.

Session 2: Focuses on how AI solutions are being applied within the advice and legal sectors – 90 minutes

5 GenAI use cases in the advice sector

Image created by ChatGPT prompt: ‘Can you create an image for the course that represents GenAI use cases in the advice and legal sector’

Show description|Hide description

GenAI use cases in the advice and legal sector. Document drafting. Legal research. Client communication. Case analysis.

Image created by ChatGPT prompt: ‘Can you create an image for the course that represents GenAI use cases in the advice and legal sector’ ...

In this section we explore some examples of how AI solutions are being applied in the advice sector.

These typically make use of custom or purpose-built models rather than more general-purpose public LLMs. Although the models have been developed for a particular organisation, the tasks they are doing are transferable to other sectors.

Different types of models

Timing: Allow 15 minutes

Read this article Overview: General Purpose LLMs vs Purpose-Built LLMs vs Custom LLMs to learn more about different types of models

Now match each of the following terms with its correct definition.

Two lists follow, match one item from the first with one item from the second. Each item can only be matched once. There are 3 items in each list.

A general-use language model like ChatGPT or Gemini that is open to the public, often trained on internet-scale data and not tailored to a specific user.
A language model designed for a specific task or industry, such as legal, medical, or customer service applications.
A language model developed in-house or by a vendor specifically for an organisation’s needs, trained on its own private data.

Match each of the previous list items with an item from the following list:

a.Purpose Built LLM
b.Custom LLM
c.General Purpose LLM

The correct answers are:

1 = c,
2 = a,
3 = b

Answer

General Purpose LLM = A general-use language model like ChatGPT or Gemini that is open to the public, often trained on internet-scale data and not tailored to a specific user.

Purpose Built LLM = A language model designed for a specific task or industry, such as legal, medical, or customer service applications.

Custom LLM = A language model developed in-house or by a vendor specifically for an organisation’s needs, trained on its own private data.

Image created in ChatGPT teams with the prompt: ‘Please can you create a mind map, the main bubble says AI solutions and can I have an arrow off to each of these solutions which are: NSPCC Childline HelpFirst prototype, Citizens Advice SORT Caddy and Citizens Advice Scotland Extra Help Unit’

Please note: The AI-generated image above is not correct. The arrow at the top should be pointing upwards, not downwards. This is why you should always check your AI output and why the ‘human in the loop’ is important.

Case Study: NSPCC Childline HelpFirst prototype

HelpFirst worked with the NSPCC to produce AI-generated summaries. It is an example of the potential of a custom-made AI model to help organisations with routine tasks that can free up time to prioritise core delivery. Although this solution was developed in a voluntary organisation, it could be relevant for other sectors.

Approach taken to implement an AI solution

Timing: Allow 10 minutes

Read this article How AI could help Childline counsellors spend more time talking to children and young people and summarise the key messages that underline the approach taken to implementing the AI solution.

To use this interactive functionality a free OU account is required. Sign in or register.

Interactive feature not available in single page view (see it in standard view).

Discussion

There might be several points you took away from this case study including:

The importance of involving stakeholders in designing the solution.
Customised solutions often perform better than general tools.
The ‘human in the loop’ is critical, the output must be reviewed by a human. The aim is for the AI solution to support the work of the human, not replace it.

Citizens Advice SORT Caddy

Watch this video by Stuart Pearson from Citizens Advice – Stockport, Oldham, Rochdale and Trafford (SORT).

Download this video clip.Video player: course_4_interview_video_clip_1.mp4

Show transcript|Hide transcript

Transcript

FRANCINE RYAN

So, please can you introduce yourself and tell us a little bit about your job and your role?

STUART PEARSON

So I'm Stuart Pearson. I'm the head of innovation at Citizens Advice SORT, which is part of the Citizens Advice Network. My role is to look at how we can use technology to help deliver on our mission at Citizens Advice, which is to provide free, confidential, independent, impartial advice to people, whether that is through making access to advice easier or making advice-giving for advisers more straightforward and more helpful.

FRANCINE RYAN

I know it's part of your role. You've developed something called Caddy. So would you tell us a little bit about what Caddy is and how you're using it within Citizens Advice SORT?

STUART PEARSON

Yes, so Caddy… I mean, one of the drivers for Caddy was that with the cost of living crisis, demand for advice has never been higher. And also, the issues people are presented with have never been more complex. And so, like a lot of probably organisations, when ChatGPT dropped on us all in November 2022 both myself and my much smarter colleague Curtis thought, is there an opportunity where we could use generative AI somewhere within the advice process which could help us meet some of this demand?

And that was the starting point really. And one of the things we thought of is, well, how do you use it safely? How do you mitigate the risks? Because everyone will be aware now around the risk of hallucinations with generative AI, where it will provide you with an answer that is very convincing but completely made-up.

So before we wrote a line of code, we sat down with our advisors and supervisors, talked about the opportunities that generative AI might provide for us, and asked what the pain points were that we had within the service. One of the things we decided very early on was that we wanted to make sure clients spoke to people, so that any kind of AI solution would not take away from that kind of human-to-human experience that for Citizens Advice and for the clients who approach us and for our advisors is so important.

But one of the areas that we did identify was… we have kind of support spaces, chat spaces, where an advisor, when we have a person (we call them clients) approached us with an issue and maybe the advisor's not too sure or they want some validation. If it's something they've never come across again or if the issue’s extremely complex, they'll drop into one of these shared chat spaces the question for one of our kind of pool of supervisors to kind of give their expert opinion on.

So what was happening with that kind of surge of demand I was talking about is that the spaces are getting very busy with multiple inquiries from our advisors. And the result of that was it was often taking longer for supervisors to respond.

Because issues were more complex, it was taking longer for them to research and respond. And sometimes, adviser having to put the question in a couple of times because it had been missed. The impact then was that it was taking longer to get the answer back to the client.

So we thought, well, here's a space where we could potentially, to paraphrase Keir Starmer, we could inject mainline AI into the issue. So we created an LLM-powered chatbot which is now called Caddy. And Caddy sits within that supervisor space. So what will happen with the process now is again, to remember the client will always speak to a human advisor. If the advisor is not sure, they will ask Caddy the question.

Another thing we did to help improve the accuracy of Caddy and its responses and try and mitigate some risks of hallucinations was to use a system called RAG, which is Retrieval-Augmented Generation, which is a very fancy term for saying restrict the information sources. So Caddy only goes to three trusted sources. One of them is gov.uk. One of them is the Citizens Advice public site. And the third one is the Citizens Advice private back-end site, which is targeted for expert advisors. It's not available to the public.

So when Caddy’s asked the question, it will go away and look at those information sources. Then when it returns the answer, it doesn't return it straight back to the advisor, it returns it to a supervisor first. So a supervisor gets the answer, checks it's got the sources, the links are in there, the sources.

Then the supervisor approves it. Now that is a really important part of the process. Because one, we're assured that information has had human oversight, and that when that information goes back to the adviser, the advisor's confident that information is correct for them to pass on to a client.

The other reason it's important is because of compliance. What that positive approval does, when there's a button you press to say this is correct, it creates an audit trail within the system to say this has had human oversight. This is a human approval. Therefore, in terms of compliance and liability, the advice given, even though the research has been done by Caddy, the advice given has been by that human supervisor. They're the person who's approved that. So by doing that, we've got kind of human guard rails on either side.

End transcript

Download

Interactive feature not available in single page view (see it in standard view).

Stuart is Head of Innovation at Citizens Advice SORT, and his team have developed Caddy which is an AI Copilot being used by advisers to provide answers to questions that are validated and signed off by a supervisor. Caddy is an LLM, but it has been custom-built by inhouse technologists at Citizens Advice SORT and draws on content from trusted websites. It produces AI-generated advice based on existing documentation directly within the messaging interface (Google Chat, Microsoft Teams, etc).

Some key points about the Caddy use case:

Caddy answers are regulated by ‘human in the loop’ validation: answers are not presented directly to the public, but only after they have been approved by a supervisor.
The responsibility of ensuring the answer to the question is correct lies with the supervisor not Caddy.
The purpose of Caddy is to support Citizens Advice to improve its processes in order to help more people and is an example of AI implementation for public good.

Citizens Advice Scotland

Citizens Advice Scotland Extra Help Unit provides support to vulnerable domestic consumers and micro-businesses across the UK with their energy complaints. They embarked on a technology challenge via CivTech Scotland to explore ‘How can technology help quickly identify and prioritise support for the people in the most vulnerable situations’.

Watch this video which explains the work of the Extra Help Unit and what a tool could mean to the organisation: CivTech Challenge 8.1 — Identifying and prioritising support — CivTech (8:07 minutes). You can choose to watch the entire video or view the segment from the beginning up to 5:16.

The Extra Help Unit partnered with HelpFirst to develop a tool that supports the triaging and prioritisation of cases. The tool highlights the cases most at risk. The tool is integrated within the Case Management System and uses AI to scan all notes and emails that come into the service. The tool works in real time, updating the case profile within 30 seconds using an LLM that scans through all the documentation adding a score to each case and flags for any risk factors. The scores are:

S – Immediate Safeguarding.
P1 – Priority 1.
P2 – Priority 2.
U – Urgent.
C – Complaint (no risk).

Score	Risk	Definition
S	Safeguarding risk	Is the client at risk of self-harm? For instance, do they mention suicidal thoughts or ideation? Do they imply they might do physical damage to themselves or to property? Do they reference wanting to ‘end it all’ or say it’s ‘not worth living’?
P1	Off-supply risk (high)	Indicates that the client is currently off-supply, meaning there is a lack of electricity or gas supply in the property. This may be indicated by phrases such as ‘off-supply’, ‘disconnection’, or explicit mentions like ‘Is Consumer Off Supply: Y’ or ‘On/off supply: Off-supply’.
P2	Off-supply risk (medium)	Indicates that the client is at risk of being off-supply soon. This includes scenarios where the client is on emergency credit (EMCR); is in friendly credit; has less than £10 positive credit on their meter or lacks access to electricity or gas for part of the day. These scenarios might be mentioned in the text as ‘customer is on emergency credit’, ‘£3.70 EMCR on the electric’, or ‘£6 reg credit’.
U	Urgency risk	Does the text mention ‘warrant’ or ‘legal proceedings’? Does the text state the term ‘urgent’ in the notes? Does the text refer to a date on which supply will be cut off if the consumer does not pay an amount of money or carry out a task for the supplier? Are bailiffs due to attend?
C	No risk detected	If no risk is detected, the case is scored ‘C’, indicating a standard complaint

These are the LLM flags risk factors.

Age	Is the client or a member of their household over state pension age (66)?
Dependents	Are there any children (i.e. people age between 5 and 18) in the premises? Do they care for someone with disabilities?
Equipment	Does the client have medical equipment that relies on gas/electricity? This might include the following: power wheelchairs or mobility devices, ventilators, oxygen concentrators, infusion pumps, intravenous or feeding equipment, chairlifts, stairlifts, communication devices, nebulisers, CPAP (continuous positive airway pressure), BiPAP (bilevel positive airway pressure), suction pumps, dialysis machines, hoists, medical-related showering/bathing need, need to wash clothes (washing machine), electrolarynx, hospital bed. Medication that must be kept in the fridge such as insulin, eye drops, antibiotic liquids, etc.
Under 5	Is there a baby present in the premises? Do they mention children under the age of 5?
Health	Is there evidence that the client’s mental health is poor (e.g. depression, schizophrenia). Is there evidence of physical health needs? This may include any physical health conditions or disabilities mentioned.
Safeguarding	Is there an active safeguarding issue? Please only output risk present if the case directly mentions domestic violence or states the words ‘active safeguarding issue’ or ‘safeguarding: y’
Immediate safeguarding	Is the client at risk of self-harm? For instance, do they mention suicidal thoughts or ideation? Do they imply they might do physical damage to themselves or to property? Do they reference wanting to ‘end it all’ or say it’s ‘not worth living’?

The scores appear on the dashboard and on the list of cases, it then helps the service with the prioritisation and triaging of cases.

It is important to stress that the normal processes used by the Extra Help Unit have not changed, the AI tool is only an aid.

Benefits seen by CAS include:

It allows customers that are most in need to be triaged more quickly and provides those working in the service with the detail to target cases based on priority rather than on date order which was not available before.
Managers can use the scores to allocate cases, and it will also highlight P2 and urgent cases at the list level – giving the service much greater insight into the cases.
In terms of the AI, it is returning over 80% accuracy on most scores and alerts.
The tool has helped the team get a better understanding of the cases they are receiving. It supports them to identify the most critical cases, which has helped to improve the mental wellbeing of staff.

AI has given the Extra Help Unit a safety net that they didn’t have before that protects consumers and staff.

Implementing AI tools

There is a lot of interest within the voluntary and advice sector around how GenAI can support organisations to help more people, but it is essential to think carefully before integrating a tool as there can be significant reputational risks if technology implementation goes wrong and the risk of harm to service users.

Citizens Advice SORT had the technical capability within its innovation team to build a custom-based model inhouse. Citizens Advice Scotland and the NSPPC worked with HelpFirst to build their solution. Wyser is another company that is working with advice and public sector organisations to integrate AI solutions, and you can read some of the case studies on their website.

Reputational risk when implementing AI tools

Timing: Allow 20 minutes

Watch this video from Stuart Pearson where he discusses the importance of organisations being aware of reputational risk when implementing AI tools.

Download this video clip.Video player: course_4_interview_video_clip_2.mp4

Show transcript|Hide transcript

Transcript

FRANCINE RYAN

So you've really baked in a lot of these responsible AI principles in there, and how important is that for organisations to really think about?

STUART PEARSON

I think for where we are with technology at the moment, particularly in the field we're in like the legal field, we can't get it wrong. You know, we have to be right. Because you've got the one, the potential reputational risks or damage to the individual, the detriment they might suffer from wrong advice. And then more selfishly, the organisation’s reputational risk. You know, that if they were to get it wrong using AI and it was kind of unfiltered advice given and that was wrong, it could ruin an organisation like Citizens Advice because trust is a key part of our brand. And if people think, well, we can't trust the advice you've given us because it's just AI, you know, then that could potentially be the end of Citizens Advice.

So it is… not just for us, I think for any organisation, it's really important at this moment in time with AI that there is a human-in-the-loop process baked into it, and that there is human oversight and that AI is there to augment the process and not seeking to be a shortcut to replace the process.

Another element as well as them being guard rails, one of the other things we wanted to protect is the human experience. So a key part of our service for a particular cohort of our clients is, they approach us and they're often vulnerable. They're often distressed. They're often angry. And what they want is to be able to speak to somebody and for that person to have the empathy and the emotional intelligence to be able to not only deescalate a potential challenging call, but also relay sometimes quite complex and sometimes not the answer the person wants to hear, in a way that that is gonna land.

And again, that human-to-human contact is so important. So that's why AI is there. What the AI does is help the advisor to be more present in that advice session, because Caddy’s doing the kind of grunt work of the research, and they can focus on delivering that message.

FRANCINE RYAN

Excellent. So in terms of top tips… I think you've touched on some of them, but what would be your top tips to other organisations thinking about using AI within their service?

STUART PEARSON

One, involving the people impacted by it. So the people who will be using it and the people whose job will be changed by using it, making sure they’re a key stakeholder in that design process, that you're getting their thoughts and they feel involved. So again, acting with a kind of transparency.

Having guardrails as well, you know, having human oversight. I think at the moment, having that human-in-the-loop process is just so important, to have human oversight over any kind of generative AI system. I think that is essential.

But also just kind of thinking about what we're trying to replace here and what are we replacing it with. I think that's the important things. And making sure that AI is there to amplify the things that you want to do, you know, rather than replace things that might be essential.

End transcript

Download

Interactive feature not available in single page view (see it in standard view).

What principles does Stuart identify as important when thinking about implementing AI tools?

To use this interactive functionality a free OU account is required. Sign in or register.

Interactive feature not available in single page view (see it in standard view).

Discussion

Stuart discusses the importance of organisations being aware of reputational risk and thinking about how AI can assist, not replace people. Organisations should consider their existing practices, and ensure they establish guard rails. They should involve stakeholders, act with transparency, and make sure there is human contact and oversight.

Citizens Advice SORT has adopted an approach to integrating AI tools that aligns with the strategies of Citizens Advice Scotland and the NSPCC. These organisations have placed strong emphasis on the responsible use of AI, ensuring that the technology supports and does not undermine the core work and values of the organisation.

People’s Panel for Artificial Intelligence

Image created with ChatGPT 4 Teams with prompt: ‘Please can you create an image with people form a diverse range of backgrounds including ethnicity and disability with a heading People’s Panel for Artificial Intelligence’

Advice and voluntary organisations that are considering adopting AI solutions might want to think about how they involve their communities in their decision-making.

The People’s Panel for Artificial Intelligence (PPfAI) is a project led by Dr Annabel Latham and Professor Keeley Crockett at Manchester Metropolitan University and funded by The Alan Turing Institute. It brings together a diverse group of citizens to learn about, discuss and share their views around the use of data and artificial intelligence in people’s everyday lives. The Panel explores both the positive and negative consequences of new research ideas, products and/or services presented to them by businesses and researchers, and they give their opinions and feedback.

The Panel is designed to give a voice to citizens in communities and empower them to have the confidence to question the use of technology.

Through a Public Engagement Grant award 2022 from the Alan Turing Institute they set up the Greater Manchester People’s Panel on AI, and a follow-on project funded by Manchester City Council Digital Strategy Team (2024) expanded the panel which Citizens Advice SORT presented Caddy to in July 2024.

Watch this video from Stuart Pearson where he talks about the Citizens Advice SORT presentation to the Greater Manchester People’s Panel on Caddy in July 2024.

Download this video clip.Video player: course_4_interview_video_clip_3.mp4

Show transcript|Hide transcript

Transcript

STUART PEARSON

Yeah, I mean just one other thing that we did within Caddy – and something that we'll do with with some other areas we're looking at is – we were really lucky that Manchester Metropolitan University and Manchester City Council, their digital strategy team, had just done a project called the People's Panel for AI, which was where Manchester Met had done some roadshows around the city to regular people asking if they're interested in being part of a People's Panel for AI.

Obviously, people self-selected. They were then put on a boot camp, a crash course on what AI is and what AI isn't. And then the People's Panel act as a critical friend to any organisation or local authority that are thinking about using AI as a potential part of their service.

And we were really lucky because we were able to present Caddy to the People's Panel and get, really, our potential clients’ input to see what they think about Citizens Advice using AI in the advice process. Were they happy the way we'd done it? Were there things we could do better? And I think again, making sure you've got as many stakeholders involved in the design of these solutions to make sure they’re as inclusive as possible when you're developing them, I think that is a key part as well.

So involving people. Openness, transparency. Ensuring that you're doing it for the right reasons. So obviously, we're looking at it through the lens of helping more people. You know, we're not looking at in terms of, you know, reductions or staff reductions. So it does make it easy for us.

The other areas we're looking at – again, focus on that kind of amplifying the kind of human skills that we have – is case note automation. It's a huge pain point in our service. So for what we call tier one advice, it might take 20 minutes maybe on the phone with a client. That could sometimes result in a 20 minute write-up. So it's taken us as much time to write about the interaction with the client as it is the actual interaction with the client, which is the most important part. So if we can automate the write-up part, that would free us up to help more people.

End transcript

Download

Interactive feature not available in single page view (see it in standard view).

The panel raised questions and conducted consequence scanning based on the project presentation and then provided feedback. Consequence scanning is a method where people discuss openly and provide their opinions on the intended/unintended consequences, potential harms or disadvantages of the technology being created.

The panel fed back with questions and comments on:

Who are the stakeholders?
What are the intended consequences of using this technology?
What are the unintended consequences of using this technology?
What are the positive consequences of using the technology that the presenter focused on?
What are the consequences of using this technology that the presenter wants to mitigate?

These are important questions that all organisations should consider when implementing new technologies. Engaging with the People’s Panel offers a valuable opportunity to involve citizens in their communities, helping to ensure that the adoption of AI solutions is informed, inclusive and transparent.

6 GenAI use cases in the legal sector

Created by ChatGPT: ‘Can you produce an image of a newspaper with this information as a story: LexisNexis produces a report each year that surveys lawyers to explore the use of AI across the legal sector, in the 2025 report it states: “Our survey of 800+ lawyers reveals that law firm leaders must act now. Not just to adopt AI, but to prove its value, optimise its use and measure its impact”.’

AI within law firms is not new. Large law firms have been using AI for due diligence, document management and contract review for a while, but GenAI has created new opportunities, which may enhance efficiency, reduce costs and allow lawyers to focus on higher-value tasks.

According to Bloomberg Law (2025) these are some of the tasks that GenAI tools are supporting lawyers in legal practice with.

Application area	Description
Drafting correspondence	Assisting with client emails, letters and internal communications.
Legal research	Retrieving and summarising case law, statute, and legal commentary.
Summarising	Condensing legal documents, judgments and contracts.
Retrieving documents	Analysing legal documents for key terms, risks, consistency and data extraction.
Due diligence	Scanning and summarising large volumes of material during transactions.
Discovery	Identifying relevant documents in litigation.
Negotiating contracts	Suggesting edits and alternatives during contract negotiation and clause comparisons.
Preparing case papers	Assembling and drafting case bundles, skeleton arguments and briefings
Estate planning	Assisting in drafting wills and trusts

In addition, Smokeball (2025) says that time tracking, client intake, legal calendaring, knowledge asset search and workflow, matter management, and invoicing and billing are activities that could be supported by GenAI tools. LexisNexis produces a report each year that surveys lawyers to explore the use of AI across the legal sector. In the 2025 report it states:

‘Our survey of 800+ lawyers reveals that law firm leaders must act now. Not just to adopt AI, but to prove its value, optimise its use and measure its impact.

Without clear AI success metrics, firms risk wasted investment, inefficiency and even an exodus of skilled professionals.’

The report is based on 807 responses from the UK, which is a relatively small survey sample size when compared with Solicitors Regulation Authority (SRA) data on the number of law firms which in February 2025 was 9,178 (SRA, 2025).

The report’s ‘Must act now’ suggests that law firms should act urgently, but the quotes within the report suggest a more pragmatic approach, acknowledging AI’s potential value whilst emphasising the need for careful implementation and clear success criteria.

Large law firms are adopting GenAI tools: Mishcon de Reya has developed deReyAI – an inhouse tool – and are combining it with Legora, Simmons and Simmons use Percy, Linklaters have Laila and some law firms are now using Harvey AI. Harvey AI will be discussed in more detail later in the section.

However, as Laura Hodgson, AI Lead at Linklaters highlights in the LexisNexis (2025) report: ‘Law firms are yet to publish hard data on the efficiency gained through using new AI tools because getting tools to a point of adoption where real value can be measured has been a challenge’.

The Legal Trends for Mid-Sized Law Firms, published by Clio (2025) in the US, found that mid-sized firms were using AI-powered legal research platforms, generic non-legal AI tools, document drafting or automation tools, eDiscovery solutions and predictive legal analytics. Mid-sized law firms were more likely than smaller firms to be using these solutions. Time saving and efficiency were considered to be the biggest drivers for AI adoption.

Other research from the US (Smokeball, 2025) states that firms are cautious about AI adoption because of cost, training and ethical concerns. They also note that smaller firms are behind because of cost and integration barriers.

Research from Macfadyen and Bish (2025) on behalf of the Bar Standards Board has looked at the use of technology to support legal work at the bar. The research found that some barristers were using ‘everyday’ AI tools such as Microsoft Copilot and ChatGPT. They found the adoption of technology at the bar was mixed, but that barristers could see the benefits of further adoption. The research highlighted some of the barriers which included the fragmentation and independent practice nature of the bar.

While these new tools may offer benefits, it is important that firms approach any investment with caution, ensuring they have a clear understanding of how the tool will add value and establishing success metrics from the outset.

Legal use cases

Created by ChatGPT Team with the prompt: ‘Please can I have a mind map with Use Cases in the centre and then off that can I have an arrow for each of these: Document drafting, Due diligence, Legal research and Client communication’

Document drafting

AI can assist with drafting legal documents using predefined templates and input data. With the correct prompts, AI can draft legal briefs, contracts, waivers and memos as a starting point, which a human is then able to review and amend. This function streamlines the drafting process and also helps to develop legal drafting skills alongside consolidating understanding by reviewing and checking the information provided by AI.

Most importantly, these documents can only be created with human input and guidance, otherwise the risk of incorrect information is significant. AI may include prohibited provisions or incorrect contract clauses. For example, AI may create a contract which has a clause to exclude liability for death or personal injury as a result of negligence – which is prohibited under the Unfair Contract Terms Act 1977.

Due diligence

AI can automate the review of large volumes of documents, identifying potential risks, issues or any information which stands out. AI tools are also able to generate due diligence reports, summarising findings and highlighting the key points.

In the ‘Further reading’ section later in this course, there is a link to a research paper, The Rag Report: Large Language Models in Legal Due Diligence (Addleshaw Goddard, 2024), which explores the development of an inhouse LLM for mergers and acquisition transaction due diligence and commercial contract review.

GenAI can make mistakes and therefore the onus is still on the human providing the prompt to ensure that the summary is conducted effectively.

Legal research

Listen to this audio clip in which Henry Sturm talks about legal research tools.

Download this video clip.Video player: course_4_workshop_video_clip.mp4

Show transcript|Hide transcript

Transcript

HENRY STURM

For all of us who are trained as lawyers or who work on legal research, you know, you go onto Westlaw and search up your case law and things like that. It's just taking that a step further in terms of being able to really find those, you know, citations. And whether it's good or not is another question.

End transcript

Download

Interactive feature not available in single page view (see it in standard view).

GenAI has the potential to revolutionise legal research by automating the analysis of vast amounts of case law and legal precedents. This technology enables users to quickly find relevant information, and save time by summarising lengthy legal documents, and making complex cases more accessible and easier to understand. However, it is important to remember that LLMs can hallucinate and make mistakes. Damien Charlotin has created a website with a database that is seeking to track the use of AI hallucinated cases in legal decisions. If you are using LLMs you are responsible for checking the output.

Click on each box below to read more.

Active content not displayed. This content requires JavaScript to be enabled.

Show description|Hide description

1. Lexis+AI. Lexis+AI is a GenAI platform integrated with LexisNexis’ legal database, enabling conversational legal research and automated drafting. It delivers summarised case law, answers legal questions in natural language, and recommends cited authorities. It combines traditional legal research tools with AI-enhanced capabilities such as citation validation (Shepards), AI-brief analysis, and search visualisation. Westlaw Precision is a similar platform. There have been concerns raised about the accuracy of these tools, therefore the results must still be checked and verified. 2. Amicus CaseMine. Amicus CaseMine is another GPT-powered, legal-AI chatbot that conducts legal research and can respond to enquiries. Tools like Co-Counsel can review, compare and summarise documents, which support document review and contract drafting.

Interactive feature not available in single page view (see it in standard view).

Client communication

AI powered chatbots can play a vital role in client communication by answering common legal questions and providing initial guidance. They can provide instant responses and are available 24/7 so can offer immediate assistance to clients.

Some law firms have integrated them into their website to explain legal processes and firm services to clients in accessible language, automating responses to FAQs about legal timelines, document requirements, and case updates. They can provide consistent, branded responses while escalating complex matters to human lawyers.

Here are some examples of tools.

Harvey AI

Harvey AI is an advanced artificial intelligence platform specifically designed to enhance the efficiency and accuracy of legal professionals. By leveraging the power of machine learning and natural language processing, Harvey AI assists with a wide range of legal tasks, including legal research, contract analysis, document drafting and regulatory compliance. By automating these time-consuming processes, Harvey AI enables lawyers to focus on higher-value activities such as strategy development and client interactions.

Harvey AI is an advanced AI-driven legal assistant that utilises LLM's to assist lawyers, law firms and corporate legal departments. It provides powerful tools for legal professionals to draft documents, analyse contracts, conduct legal research and generate insights from case laws quickly and efficiently. Developed by a team of AI researchers and legal experts, Harvey AI integrates cutting-edge technologies like OpenAI’s GPT-based models to understand complex legal queries and provide relevant, reliable and actionable responses. The AI aims to enhance decision-making, reduce manual workloads and improve accuracy in legal documentation.

Harvey AI operates through advanced machine learning algorithms and natural language processing capabilities. It processes vast amounts of legal texts, case laws, regulations and statutory materials. By understanding and interpreting legal queries in plain language, Harvey AI allows lawyers to interact with it naturally. Using deep learning Harvey AI refines its accuracy by learning from past cases, user interactions and historical legal data. It can quickly scan through legal databases to provide relevant case law and citations, assist in drafting contracts and legal agreements, and review legal documentation for inconsistencies or potential risks. Additionally, Harvey AI helps firms stay compliant with legal regulations by analysing policies and detecting any areas of non-compliance.

Several prominent UK law firms have adopted Harvey AI. These firms utilise Harvey AI to perform tasks, such as due diligence, litigation support, and compliance analysis, significantly improving productivity and reducing manual workloads.

AI Matter Assistant (Archie)

Archie automates routine legal tasks. It has been designed to support law firms with summarising documents, drafting letters or emails, answering matter-specific questions and conflict checking and risk analysis. It has been created by a US company as part of its document management system.

LegalMation

LegalMation is a US company that brings together lawyers and technologists to design solutions for law firms. It has a series of tools that use AI to support litigation and dispute resolution and is being used by law firms, corporate legal departments and insurance companies. To learn more about this tool watch this short video.

Automated contract review

There are a proliferation of AI contract review tools like Kira Systems and Luminance that can analyse contracts to identify key clauses, risks and compliance issues. These tools use machine learning algorithms to quickly process large volumes of documents, saving time and reducing human error. Kira Systems can extract and analyse clauses from contracts, providing insights into potential risks and obligations. Consequently, these tools result in increased efficiency, accuracy, and consistency in document review. Other examples include Juro, Spellbook and Lexion, and this article suggests the nine best AI contract review tools.

Predictive analytics

AI tools like Lex Machina and Premonition can analyse historical case data to predict litigation outcomes. These tools use advanced algorithms to identify patterns and trends, helping lawyers make informed decisions. Lex Machina provides data-driven insights into case outcomes, judge behaviour, and opposing counsel strategies – improving decision-making, strategic planning, and risk assessment.

AI in judicial decision-making

The University of Ulster is working on research into the role of AI in judicial decision-making and the research will develop ethical guidelines and policy recommendations about the adoption of AI across the courts system.

For organisations considering the adoption of GenAI tools, the final sections of this course explore a practical framework for AI adoption, along with processes to support effective change management.

Strategic framework for AI adoption in law firms

Created by ChatGPT teams prompt: ‘Please can you create an image that is titled: Strategic Framework for AI Adoption in Law Firms and then it has four stages, can you add some visuals to make it engaging and all refer to the four stages below: Stage 1: Education and exploration, Stage 2: Strategy development, Stage 3: Implementation planning and piloting and Stage 4: Adoption and integration’

GenAI tools can be applied to both legal and non-legal tasks. When considering which tools to adopt:

Engage with different vendors.
Conduct thorough due diligence to evaluate risks and capabilities.
Develop a clear implementation plan that aligns with the firm’s overall strategy and professional obligations.

The third course, Key considerations for successful Generative AI adoption, explored how to create a GenAI strategy. Below is a framework that organisations can adapt to guide their thinking around AI adoption.

Stage 1: Education and exploration
Objective	Key activities	Success criteria
Learning about GenAI, experimenting, and understanding capabilities.	Organise educational workshops for staff on GenAI fundamentals. Create cross-functional working groups with both legal and technical staff. Allocate dedicated time for hands-on exploration of GenAI tools. Review industry case studies of AI implementation in legal settings. Speak to vendors. Identify internal champions and sceptics to ensure balanced perspectives. Document initial impressions, concerns, and potential opportunities.	Knowledge assessment scores pre/post training. Percentage of staff participating in AI workshops. Quality of internal discussions and questions raised. Initial inventory of potential use cases identified. Initial inventory of potential vendors.

Stage 2: Strategy development
Objective	Key activities	Success criteria
Creating a tailored approach that aligns with firm priorities.	Identify and prioritise specific use cases based on: – Potential efficiency gains – Current pain points – Client pressure/interest – Competitive landscape. Map legal and regulatory requirements for AI adoption: – Client confidentiality implications – Data protection compliance – Professional responsibility considerations – Disclosure obligations Develop governance framework: – Oversight committee structure – Decision-making protocols – Risk management approach – Ethical guidelines for AI usage and policy documentation. Create preliminary resource allocation plan: – Budget requirements – Personnel needs – Technology infrastructure assessment – Training requirements.	Completion of comprehensive use case analysis. Documentation of legal/regulatory compliance framework. Establishment of governance structure. Sign off and approval of strategic direction. Clear definition of ‘minimum viable success’ for initial implementation.

Stage 3: Implementation planning and piloting
Objective	Key activities	Success criteria
Targeted testing of AI solutions in controlled environments.	Select specific tools for pilot testing based on strategy. Design pilot projects with: – Clear scope limitations – Defined timelines (typically 3–6 months) – Representative but low-risk matters – Participation from various seniority levels Establish robust measurement frameworks: – Time savings (hours saved per matter type) – Cost efficiency (reduction in billable hours while maintaining quality) – Accuracy rates (compared to traditional human review) – Client satisfaction scores for AI-assisted work – Lawyer satisfaction and adoption rates. Develop feedback mechanisms: – Structured questionnaires for lawyers using the tools – Documentation of specific successes and failures – Tracking of technical issues and workflow disruptions – Measurement of learning curve and training requirements – Calculate true ROI including training, implementation, and licensing costs.	Completion of planned pilots within timeline. Quality of data collected from pilot projects. Specific improvements identified in workflow efficiency. Documented Return On Investment (ROI) calculations for each use case.

Stage 4: Adoption and integration
Objective	Key activities	Success criteria
Scaling successful implementations and embedding into firm operations.	Select tools and processes for firm-wide deployment based on pilot results. Develop comprehensive training programs: – Role-specific training modules – Ongoing support resources – Best practices documentation – Knowledge sharing mechanisms. Create integration plans for existing systems: – DMS connections – Billing system integration – CRM integrations – Knowledge management incorporation. Establish continuous improvement mechanisms: – Regular usage audits – Performance benchmarking – Feedback collection systems – Update and iteration processes. Develop client communication approach: – Value proposition articulation – Transparency about AI usage – Billing implications – Success stories and case studies.	Adoption rates across practice areas. Efficiency gains (time/cost). Client feedback on AI-assisted work. Staff satisfaction with AI tools. Impact on recruitment and retention. Revenue implications (cost savings and/or new service offerings). Competitive positioning enhancement.

Best practice throughout all stages

Click on each label below to read more.

Active content not displayed. This content requires JavaScript to be enabled.

Show description|Hide description

1. Focus on assistance, not replacement. Identify which parts of legal work benefit most from AI assistance. Look for opportunities where AI elevates the lawyers' work rather than substituting it. Track how AI changes the nature of paralegal/junior lawyer work and training. 2. Maintain clear communication. Regular updates to all stakeholders. Transparent sharing of both successes and challenges. Active management of expectations. 3. Prioritise security and ethics. Regular security assessments of AI tools. Clear protocols for handling confidential information. Ethical guidelines for appropriate AI usage. 4. Iterate based on evidence. Make decisions based on gathered data, not hype. Be willing to pivot if initial assumptions prove incorrect. Learn from failures as well as successes. 5. Build for the long term. Focus on sustainable adoption rather than quick wins. Consider how AI fits into broader innovation strategy. Develop competencies that will transfer across tools and generations.

Interactive feature not available in single page view (see it in standard view).

7 Change management

An organisation’s decision to implement AI tools should be embedded within its broader strategic plan.

However, even with a sound strategy, technology projects can still often fail because the tools are not effectively adopted by users within the organisation. Change is inherently difficult particularly when it involves adopting new technologies. Without effective change management processes in place, organisations may encounter a range of issues, including staff resistance, reduced productivity and, ultimately, failure to successfully implement the technology.

Change management processes can help to:

Minimise resistance: staff should be informed and involved in the change process and understand and see the benefits of the new technology.
Focus on people: staff should be part of the change process to ensure their engagement and adoption of the new technology.
Transition processes: change management processes should ensure that the transition to the new technology is planned, ordered and managed to minimise disruption and risk.

Change management models

Image created by ChatGPT Team using the prompt: ‘Can you create a visually interesting diagram of the ADKAR Model: (A) Awareness of the need for change, (D) Desire to support and participate in the change, (K) Knowledge of how to change, (A) Ability to implement the change, (R) Reinforcement to sustain the change’

There are different change management models. One example is the ADKAR Model which is a 5-step model developed by Jeff Hiatt in 2003 (Angtyan, 2019):

(A) Awareness of the need for change.
(D) Desire to support and participate in the change.
(K) Knowledge of how to change.
(A) Ability to implement the change.
(R) Reinforcement to sustain the change.

This model emphasises individual change, aiming to support each person in the organisation through the transition processes. One of its key advantages is that it provides managers with a clear, outcome-focused framework shifting the emphasis from simply completing tasks to trying to achieve meaningful change.

As organisations move toward implementing GenAI solutions, effective change management will be essential. A thoughtful, well-planned approach is critical to maximising adoption and ensuring long-term success. If you want to explore other change models, you can read about the top ten change management models.

Next steps

To move forward, organisations should:

Set and clearly communicate goals.
Build a cross-functional team.
Develop a well-defined strategy.
Create a detailed implementation plan.
Plan internal and external communications.
Establish mechanisms for continuous improvement.

Throughout this course, we have explored the importance of having strategies, frameworks, and models to guide the adoption of GenAI tools.

While time has only allowed us to provide an overview, our goal is to raise awareness and help organisations think about the questions they need to ask. This includes identifying potential use cases, understanding the implications of implementation, and planning thoughtfully for roll-out and adoption.

By taking a structured and reflective approach, organisations can lay the foundation for successful and responsible use of GenAI solutions. To learn more about change management you can enrol on the free Open University course, which explores managing and leading change.

8 Conclusion

This course has explored potential use cases for GenAI, and we hope it has provided some insights into how AI tools might be adopted.

The second course, Skills and strategies for using Generative AI, explored skills and strategies for GenAI and when it comes to thinking about use cases the importance of playing with tools cannot be overstated.

It is essential to encourage people in your organisation to play with the tools to develop skills. Playing with these tools allows individuals to explore their capabilities, understand their limitations, and discover innovative ways to leverage them. This hands-on approach fosters creativity and problem-solving skills.

Although GenAI has the potential to transform the workplace, it is important to remember that skills like critical thinking, empathy, creativity, and collaboration are becoming increasingly valuable because AI cannot replicate them (Randa, 2024).

Rather than viewing AI as a replacement for humans, the better model is augmented intelligence – AI that enhances rather than substitutes human input. The World Economic Forum (2025) stresses that uniquely human skills must be developed in tandem with technological adoption to avoid skill gaps and displacement.

If AI takes over routine cognitive work, the competitive edge will shift toward emotional intelligence, communication, and ethical reasoning, which is discussed in further detail in the eighth course, Preparing for tomorrow: horizon scanning and AI literacy.

In the next course, you will explore Responsible and Ethical use of Generative AI.

Moving on

When you are ready, you can move on to the Course 4 quiz.

Website links

Actionable Intelligence

AI Hallucination Cases Database – Damien Charlotin

AI is transforming the world of work, are we ready for it? | FT Working It

Alan Turing Institute

CivTech Challenge 8.1 — Identifying and prioritising support — CivTech

Consequence Scanning – an agile practice for responsible innovators – doteveryone

Contract Review – Powered by AI

Embracing AI: How charities like us can benefit | The Herald

Generative AI: Beyond The Theory Webinar Series | Addleshaw Goddard LLP

Generative AI for legal professionals: Its growing potential and top use cases

How do people feel about AI?

How to train and configure GenAI tools | ICAEW

Introduction to_AI_Assurance.pdf

People’s feelings about AI: an evidence review

People's Panel for Artificial Intelligence

Percy – Our unique Generative AI tool, purpose-built in house

The Rag Report: Large Language Models in Legal Due Diligence

Refreshed AI Guidance published version – website version

Report on AI and the public good

Report on a survey of AI researchers

Research: Greater Manchester People’s Panel for AI

Responsible AI Toolkit – GOV.UK

Tech Horizons report 2024 | ICO

The tipping point: Measuring the success of AI in legal services

Tackling a new legal landscape with AI

Wherever Computer Meets Contract

World’s Largest Litigation Database

References

Addleshaw Goddard (2024) The RAG Report: Large Language Models in Due Diligence. Available at: https://www.addleshawgoddard.com/globalassets/insights/technology/llm/rag-report.pdf (Accessed 14 April 2025).

Angtyan, H. (2019) ‘ADKAR Model in Change Management’, International Review of Business and Management Research, 8(2), pp. 179–182. Available at: https://www.irmbrjournal.com/papers/1560753273.pdf (Accessed 16 April 2025).

Bloomberg Law (2025) AI for Legal Professionals. Available at: https://pro.bloomberglaw.com/insights/technology/ai-in-legal-practice-explained/#legal-ai-tools-and-technology (Accessed 14 April 2025)

Clio (2025) Legal Trends for Mid-Sized Law Firms. Available at: https://www.clio.com/wp-content/uploads/2025/01/2025-MM-Legal-Trends-for-Mid-Sized-Law-Firms.pdf (Accessed 14 April 2025)

Cole, N. and Bush, D. (2025) Write with AI. Available at: https://writewithai.substack.com/ (Accessed 16 April 2025).

Financial Times (2024) AI is transforming the world of work, are we ready for it? | FT Working It, 2 December. Available at: https://www.youtube.com/watch?v=hQX_wIW9Nh0 (Accessed: 17 June 2025).

Harvey AI (n.d.) Professional Class AI. Available at https://www.harveyai.com (Accessed: 17 June 2025).

Litera Kira (n.d.) Contract Reviews Powered by AI. Available at https://www.litera.com/products/kira (Accessed: 17 June 2025).

LexisNexis (2025) The tipping point: Measuring the success of AI in legal services. Available at: https://www.lexisnexis.co.uk/insights/measuring-the-success-of-ai-across-the-law/index.html?utm_source=All%20Subscribers&utm_medium=email&utm_content=Main+CTA&utm_campaign=603983_EM2+-+Generative+AI+H1+2025_RET_ACA (Accessed 14 April 2025).

Macfadyen. J. and Bish, D. (2025) Technology at the Bar Research: Report for the Bar Standards Board. Available at: https://www.barstandardsboard.org.uk/static/370d1003-9533-4316-87ba75f68a41c357/Tech-at-the-Bar-2025.pdf (Accessed 22 May 2025).

Randa, K. (2024) ‘The human edge: Why soft skills matter more than ever in the AI era’, Fast Company, (29 August). Available at: https://www.fastcompany.com/91178670/the-human-edge-why-soft-skills-matter-more-than-ever-in-the-ai-era (Accessed 31 March 2025).

Smokeball (2025) State of Law: Adapting for resilience and growth in a tech-driven age. Available at: https://go.smokeball.com/2025_state_of_law_report?utm_campaign=sol_download&utm_medium=email&utm_source=content (Accessed 14 April 2025)

Solicitors Regulation Authority (SRA) (2025) Breakdown of solicitor firms. Available at: https://www.sra.org.uk/sra/research-publications/regulated-community-statistics/data/solicitor_firms/ (Accessed 14 April 2025)

World Economic Forum (2025) Elevating uniquely human skills in the age of AI. Available at: https://www.weforum.org/stories/2025/01/elevating-uniquely-human-skills-in-the-age-of-ai/ (Accessed 31 March 2025).

Acknowledgement

Grateful acknowledgement is made to the following sources:

Every effort has been made to contact copyright holders. If any have been inadvertently overlooked the publishers will be pleased to make the necessary arrangements at the first opportunity.

Important: *** against any of the acknowledgements below means that the wording has been dictated by the rights holder/publisher, and cannot be changed.

Course banner 557600: Creativa Images / Shutterstock

557951: Created by OU artworker using OpenAI. (2025). ChatGPT 4

557959: Created by OU artworker using OpenAI. (2025). ChatGPT 4

557995: Created by OU artworker using OpenAI. (2025). ChatGPT 4

558046: Created by OU artworker using OpenAI. (2025). ChatGPT

558058: Created by OU artworker using OpenAI. (2025). ChatGPT

558125: Created by OU artworker using OpenAI. (2025). ChatGPT 4

558132: Created by OU artworker using OpenAI. (2025). ChatGPT

558260: Created by OU artworker using OpenAI. (2025). ChatGPT 4

558273: Created by OU artworker using OpenAI. (2025). ChatGPT

558245: Created by OU artworker using OpenAI. (2025). ChatGPT

558355: Created by OU artworker using OpenAI. (2025). ChatGPT

558363: Created by OU artworker using OpenAI. (2025). ChatGPT

Use cases for Generative AI