How do you understand what you are buying when you procure AI? This was the topic of the first panel discussion at The AI Agenda with Lewis Silkin tech lawyers Rory Campbell and Roch Glowacki talking to Jaeger Glucina from Luminance and Janine McKelvey from BT Group.
Some key themes were:
- Understanding what you are buying, what you need the AI to do and how it does it. There is often a misalignment of expectations between the supplier and the customer. However, this can be overcome by proper planning and due diligence, for example, is the supplier well-known and what are the pain points that you want the AI tool to solve. Also think about what productivity means to you – is it just doing more with less or do you want to mitigate risk, unlock revenue or futureproof your business.
- The difference in approach between private practice lawyers and in-house lawyers. Law firm clients are interested in trust, liability and insurance, whereas in-house lawyers are interested in what the AI is, where the data is hosted and how accurate it is. Lawyers need legal grade AI rather than general solutions.
- Explainability is key and vendors must understand what they are selling – it's a red flag if they can't tell you how their tech works.
- It's not just another piece of tech and it's just about (personal) data and IP – and if you are regulated you need to consider sector regulation as well as the wider regulatory context.
- Train your staff and make sure they are tech and AI literate. And look at your AI governance procedures and policies generally.
There are four elements to how you approach procuring AI: procure, deploy, generate and sell. The latter three depend on the first.
Market practice is evolving and the approach to contracts is gradually changing. There's a place for a universal standard such as from the ISO which would give both vendors and users comfort that the tech meets appropriate standards. Such standards are more agile and useful than heavy handed and slow-to-react regulation. There is also a nascent market in insurance products which will shift the dial on contracts. However, contracts are very important as they provide certainty and there are model clauses which can help to address some of the key risks, although they need to be adapted for the context. Think about your perfect position, and the position you can live with. Finally, assurance is key in addressing risk and being able to report to the Board about the success or otherwise of the AI.
So what were the takeaways?
- Avoid template contracts – be specific and talk to the right people in the business.
- Keep your contracts in one intelligent system so that you can track what you have.
- Understand what you are buying – spot risks, work out your risk appetite and mitigate those risks. Do your due diligence and control your exposure.
- Equip your stakeholders to have informed conversations about AI and what forms of AI you need. Model clauses and FAQs help to contextualise use cases.
Useful resources:
- How to navigate the AI procurement maze
- AI Implementation Pathway - The key legal, regulatory and business issues you need to think about when you are navigating the AI product lifecycle
- Society for Computers and Law AI Group launches
Artificial Intelligence Contractual Clauses - Society for Computers
& Law
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
