GenAI in-house: Cambridge University Press GC on how to increase engagement

Catie Sheret, general counsel at Cambridge University Press & Assessment, admits that she was hesitant to take on a GC role. Having switched from Linklaters to Charles Russell at 2 years’ PQE to get more commercial work and exposure to clients, she became aware of a buzz around in-house IP roles. Her interest drew her to her first in-house role at educational publisher Pearson in 1999.

Going on to manage a team and becoming Pearson’s chief privacy officer in the lead up to GDPR, Sheret was then headhunted for her current role.

Speaking of her time in-house, Sheret comments: “It felt like my hands were on real work straight away. I negotiated a massive contract with Disney one year in, taking it from start to finish.”

Introducing AI to the organisation

As with all lawyers, things took a turn for Sheret’s in-house team at Cambridge following the formation of genAI platform ChatGPT. First, it prompted the organisation to work on a safe use process of dos and don’ts. Little did they know the potential of these foundations.

Sheret mentions: “There was a very interesting drafting activity on the guidance, where people argued with each other on the legal issues of genAI.

“While some were trying to push the boundaries and allow more to happen than is really safe, others jumped in by saying, for example, ‘absolutely not, we can’t have personal data going in without it being anonymised’, which was brilliant because I didn’t have to be the ‘bad cop’.”

Collaboration across the business also gave the 50-strong legal team early exposure to the technical teams who were really engaged with AI. The organisation was made as aware as possible on evolutions in tech through a strategy and central team, who coordinated activity in each of the business areas, prioritising pilots within the legal team. The legal department has also been granted access to a private copilot and a web app so it can upload confidential information in a safe space.

Taking up a fair chunk of Sheret’s diary at the moment is the hub committee. This provides a weekly sense-check on use cases and a final check and balance before going to the production and full deployment of software. Sheret believes: “It’s about getting that balance of governance right and not stifling experimentation and innovation”.

Tech in the legal team

Cambridge’s legal team has grown as responsibilities have grown. Legal operations has eight in its team, with six in Manilla, Philippines and two in the UK. To rally engagement, which can prove difficult for GCs, Sheret kickstarted a competition to ideate the best way to implement AI within the team. The winners each had an element of their suggestion related to contract reviews.

Navigating the minefield of software options then became a game of trial and error. Cambridge didn’t have the capability internally to do anything other than use a non-private version of ChatGPT, and they weren’t comfortable putting contracts in there. Sheret employed legal engineering firm Simmons Wavelength to install a pilot to compare three different versions of different LLMs.

“We kind of knew it wouldn’t work because the LLMs weren’t built for what we need, but the pilots were able to show us the outputs and the pros and cons of each software,” says the GC. “It was a learning experience and we ended up with a great report about the jargon and limitations of the systems.”

Thankfully for Sheret, the legal team were not starting from scratch; they have some AI capability inside the tools they already use. There’s an eDiscovery tool that cuts review time on large datasets and the Beta enhanced AI functionality is switched on in the team’s contract management system.

Sheret explains: “We are just starting to see what insights that can give us, so we are moving towards more comprehensive obligations management on our contracts, which is very exciting for our procurement team, but it’s fairly early stages.”

Finally, for its contract management system, Cambridge has a user community that sits within Teams. They get multiple questions every day, many of which have been asked before. When you’re in Teams, it’s difficult to find questions and answers further up the chat, which can lead to hours of scrolling through irrelevant chats. So, a copy of the chat is being cleansed and put into the chatbot. As Sheret details, though, this is in the initial stages: “We will see if that can answer the questions as a first line support which will take some of the load off our legal operations team.”

Sheret attends workshop sessions between general counsels and lawyers in their organisations. They work through and share what they are doing about AI, a level of collaboration, which Sheret likens to the introduction of GDPR.