4 Model selection and customisation

As organisations begin to engage with GenAI, it is helpful to distinguish between ‘everyday’ AI tools, those integrated into existing software, and more specialist AI tools designed for specific tasks. For many, starting with everyday tools offers a practical first step before considering the adoption of more advanced or bespoke AI solutions.

Pre-trained models are more cost effective, quicker to implement, but might not be as useful for specific-use cases. Some models might be free as part of your existing IT package such as Google Gemini or Microsoft Co-Pilot, other models are likely to be subscription based. If you are a charity, you might be able to access Nonprofits offers, for example, both Microsoft [Tip: hold Ctrl and click a link to open it in a new tab. (Hide tip)] and Google provide technology grants and discounts to nonprofit organisations.

Once you have considered model selection you should consider consequence scanning which is a method that allows you to explore the following three questions about your solution:

• What are the intended and unintended consequences of this product or feature?
• What are the positive consequences we want to focus on?
• What are the consequences we want to mitigate?

These questions provide an opportunity to share insights, raise concerns, and have structured conversations about potential impacts. You should document the answers and assign clear ownership for any follow-up actions. These outputs can then be integrated into your planning. If you want to learn more about how to do consequence scanning you can read this guide.

In the fourth course, Use cases for Generative AI, there are examples of purpose-built and custom-built models which have either been developed for the sector or in-house or in partnership with technology companies. Customer built and purpose-built models can be tailored to a business or an organisation but are more resource intensive and require technical expertise in model training and prompt engineering.

Before deciding on any tool, ensure that you trial and demo it first to determine its effectiveness and relevance to your purpose.  It is essential to carefully read and fully understand the contractual terms associated with any GenAI tool or service.

Terms and conditions for GenAI providers can typically be found on their official websites, often linked in the footer or during the sign-up process. It's crucial to review these terms to understand the scope of services, any restrictions on use, and the legal implications of using the GenAI tools.

These terms typically cover critical issues such as:

• Data rights.
• Intellectual property ownership.
• Acceptable use policies.
• Geographical location of data storage.
• Liability clauses.

If you are a charity or non-profit organisation, consider pro bono legal advice to review the terms and conditions before making any commitments. A law firm can explain the terms and conditions in plain English, ensuring that you are fully aware of the implications and potential risks involved in using the technology. Taking this step can protect your organisation and the people you support. LawWorks brokers legal advice to small not-for-profit organisations on a wide range of legal issues.

The sustainability of the technology is also an important consideration – you need to think about not only the initial costs but also the ongoing costs of using the tool. The adoption of GenAI needs to be part of long-term planning and you need a clear understanding of what the exit plan is for any tool that you acquire.