Addleshaw Goddard tested legal AI with 400 people before giving it to 3,400
Software selection in a large firm is usually a procurement exercise with a demo at the end. A shortlist, a few partners in a meeting room, a scoring sheet and a decision the rest of the firm hears about by email. Addleshaw Goddard ran it the other way round. Over three months, 400 partners, lawyers, legal technologists and business services professionals tested leading legal AI platforms on real legal and business tasks before the London-headquartered firm chose Harvey.
What the firm committed to
From January, Harvey goes to the firm's entire workforce of more than 3,400 lawyers and business services professionals across Europe, the Middle East and Asia. The 80-strong Innovation Group will work directly with Harvey's product and legal engineering teams to shape and test priority capabilities, build consistent ways of working for adoption across the firm and create new AI-enabled services. The firm and Harvey also plan to meet selected clients every quarter to identify shared problems and, where it makes sense, co-create solutions.
Managing Partner Andrew Johnston called it "the most significant technology transformation in AG's long history as an innovator" and said it "goes well beyond a firmwide deployment". The co-development has already started. During the assessment, the firm's feedback shaped Harvey's Review Tables, which help lawyers organise, compare and verify large volumes of information, with changes to source checking, to preserving work lawyers have already reviewed, and to sharing and performance.
Why the selection process matters
Four hundred testers amounts to an early adoption cohort. Those people have used the tool on real work, formed views about where it helps and where it does not, and seen their feedback land in the product. When access reaches everyone else in January, they become the colleagues people ask first. That is a stronger starting position than launching to the whole firm at once with a training video and an FAQ.
Kerry Westland, Partner and Head of the Innovation Group, was careful about what the rollout is for. "This is a decisive next step, but it is not simply a technology rollout," she said, describing the goal as better outcomes, greater quality and consistency and new ways of working with clients "while keeping human judgement, supervision and accountability at the centre."
That last clause is where the work sits. Review Tables can organise a data room, but somebody still has to decide what the table means, which rows to trust and when to go back to the source. Supervising AI-assisted work is a skill in its own right, and it deserves the same deliberate teaching that drafting and research have always had.
What this means for L&D
The transferable lesson is the sequencing. Involve hundreds of people in choosing the tool, let their feedback change it, then use them to carry the habits into the wider rollout. Learning teams can borrow that pattern without a vendor partnership: run structured trials on real work, turn what testers learn into shared ways of working, and teach supervision and verification as deliberately as prompting. Short, structured trials on real work are the idea behind our AI Design Sprints, and they beat handing 3,400 people a login and hoping consistency follows.
Sources:
News and insights for innovation, digital transformation, future of work and L&D leaders.
Stay ahead of learning and development, corporate innovation and digital transformation news. Plus the future of work. For leaders in AU, NZ, HK, SG, the US, the UK and Canada.
