How we approach contracting for AI

One of the most fascinating aspects of our artificial intelligence (AI) journey at Pinsent Masons has been the dual role of advising clients on AI contracting while also reflecting on these opportunities and risks in our own operations.

In general, businesses adopt AI for three reasons: achieving efficiency, new business propositions/channels, and simply remaining competitive. To achieve these goals, it’s imperative they consider:?

• What sets us apart from the competition?
• How can we help our people use AI responsibly?
• Do we have the necessary resources—infrastructure, data access, and expertise—to promote AI adoption in the workplace?

Through our engagements so far we’ve learned that AI can revolutionise contracts more than almost any other technology, similar to how cloud-based computing defined a new contracting approach 15 years ago. Due to the complexity of AI and the unique needs and risks at each stage of its lifecycle (like training and deployment), having a solid contract and governance is crucial to manage risks and set clear expectations.

In this issue of AI in Action, we delve into our approach to AI contracting, covering the essential steps before, during, and after acquiring AI.

Strategic thinking

Before any business dives into AI, it’s essential to have a clear strategy. To ensure AI is not used in a high-risk way, strategic thinking must determine the course of action prior to contracting for AI. Even when focusing on low-risk, high-return areas like marketing and coding, even the most cautious businesses are creating board-sponsored manifestos to guide their AI journey before wide-scale adoption.

As an innovation-first firm, we've been experimenting with AI for years. The development of generative AI technology refuelled existing conversations at Pinsent Masons about the opportunities and risks associated with AI, as well as initiating newer conversations focused on generative AI.

Our AI Steering Committee (SteerCo) shapes our strategic thinking on the opportunities and risks associated with AI, and the governance around our practical rollout of AI. They prioritise which AI projects to tackle first and give our people a central contact point for any questions. The SteerCo itself works with other teams, like our data team, to ensure our AI efforts are effective and we can leverage our AI tools in the best way possible. Meaning, we’re ready to safely seize opportunities as they present themselves.

Policies, guidance and education

For any organisation exploring AI, clear policies and guidelines are the guiding light. These rules help with both acquiring AI and using it safely within the company.

We created our own AI policies early on to ensure everyone understands the opportunities and risks (and how to handle them). Policies should generally contain simple rules and guidance. For example, if you include a rule requiring your people to validate the outputs of an AI tool because it can make mistakes, it is sensible to accompany that rule with guidance on how to perform that validation.

These rules and guidelines build confidence and understanding, leading to greater engagement. Whenever we introduce a new tool, we provide specific training and policies to ensure everyone feels empowered and safe. Our Procurement and Legal teams also use these policies to keep our supplier checks up to date and ensure smooth contract negotiations.

Education is key to building trust and excitement about AI. That’s why we also established an AI Forum a couple of years ago, a junior-led group where colleagues and external speakers discuss AI developments in an accessible and collaborative way.

Contracting for AI

When dealing with AI products and services, it’s important to understand how and where AI is used, what models will interact with your data, and where this interaction happens. This is part of the supplier due diligence process.

In contract negotiations with AI suppliers, it’s crucial to get key personnel in the room early. This helps build mutual understanding around proper testing and monitoring of AI services to ensure we get the best service and maintain accountability for errors. Transparency from suppliers is key to ensuring data safety, including where data is stored and if it’s used to train the supplier’s existing model. We validate AI outputs to ensure they meet our goals before finalising contracts. Without these things, we won’t proceed to contract.

For guidance on how buyers of AI can impose contractual controls on use, please see our recent Outlaw article and watch our latest webinar on the topic.

Learn by doing

As we discussed in our last edition of AI in Action, the integration of M365 Copilot taught us a great deal of vital lessons, which we have since applied to our usage of Lexis+ AI and Ask Practical Law.

The next thing we're going to do is use the cutting-edge AI platform V7 Go to improve our creativity and analysis. As we investigate these innovative tools, we'll keep sharing our findings.