Legal Geek Insights - Joanna Goodman

Spring forward

As the clocks go forward and agentic AI takes over from prompt engineering as this year’s legal tech buzzword, there is a new elephant in the room: the extent to which digital colleagues can replace humans in legal services. In both corporate and consumer work AI automation is a trade-off between the speed/efficiency of an AI tool trained on relevant, up-to-date data, and the knowledge, experience and emotional intelligence of a human lawyer. The optimal position on the artificial/emotional intelligence scale will always depend on context. 

On the client side, if a person or organisation is dealing with standard legal agreements, straightforward transactions, or minor disputes, they will appreciate instant online access to information and guidance, quick answers and a cost-effective solution. Some of these workflows are semi-automated already, and deploying agentic AI to make them mostly/entirely self-service will reduce lawyers’ work– and should also reduce costs. However, parties involved in an emotive dispute or a high-value, bet-the-company transaction will be looking for more personalised service, so agentic AI will be confined to internal processes and workflows, potentially replacing business support professionals. Either way, once it is up and running, agentic AI introduces digital workers into legal teams, changing the legal services business model in a more direct way than the perennial discussion about the demise of the billable hour.

Fast and slow followers

Firms in the vanguard of GenAI are already investing in agentic AI – with March announcements including several new wins for Legora, which launched in the US just last week, making no secret of the fact that they are competing head to head with Harvey, which also announced ‘next generation agents’. And incumbent legal tech vendors are also introducing agentic capability. However, some firms may be hesitant about delegating legal workflow management to GenAI because of the implications of replacing people with technology. While legal has invested a lot into ‘embracing GenAI’ and the technology landscape has indeed transformed, the underlying narrative, culture and people are broadly the same. The legal procurement process is long, and the biggest influencers are not first movers, but fast followers.

GenAI adoption in legal is also about slow followers! There is still a lot of discussion about AI readiness, use cases and prompt engineering 2.5 years after the launch of ChatGPT – even though a lot of firms and most people (not just lawyers) are already using GenAI. While agentic AI is a new direction for legal, because it is about replacing people, not augmenting them, ultimately it is about efficiency – finding better ways to automate processes and workflows. So the starting point has to be to look at each process in a workflow, and identifying logjams and sticking points where the AI agent should have human supervision may help alleviate concerns about keeping a human in the loop.

What seems to be missing from the many recent reports published on GenAI in legal is the key first element of the classic MBA strategic plan – vision, mission, objectives. While the mission and objectives around GenAI are clear, around boosting lawyer productivity and client service, keeping up with peers in the market, etc, the vision of an AI-powered law firm, where human and digital colleagues work together, and how this will improve on the current law firm operating model, is still pragmatic. This suggests that legal (thought) leadership is eating its own dogfood and applying GenAI to the problem – and, because large language models are trained on (existing) current and past data, they often lack clarity when it comes to building out future scenarios.

Innovator or mirror

In an interview with Alan Alda on his Clear+Vivid podcast, Shannon Vallor, author of The AI Mirror, argues that because large language models (LLMs) learn from currently available data, their output is predicated on past values. This is less relevant for business processes and workflows, but it is important when it comes to making value judgments, because it mirrors dominant patterns in previous judgments and decisions, without always recognising emerging factors, or changes in the business environment. These are at the heart of legal work, and failing to take into account new developments in tech and shifting political and societal values and priorities is also counter-innovative. 

This may be one reason why consolidation seems to be outpacing innovation in legal tech right now, with global cloud-based legal tech provider Clio following the growth by acquisition strategy of other large incumbent vendors and acquiring UK case and matter management software vendor Sharedo. More unusually, US law firm Cleary Gottlieb acquired GenAI product development company Springbok AI. In the last couple of years the direction of innovation has shifted from large firms and vendors developing, investing in and incubating legal tech/legal AI start-ups and then spinning them off, to the big North American vendors and law firms boosting their innovation credentials by acquiring start-ups that have already achieved market recognition. 

Catch the vibe

The last couple of weeks have seen a plethora of new tools and capabilities around GenAI, both generalist and legal specific, but how do you find the best solution for different use cases? Maybe try them all!

In his Brainyacts newsletter, Josh Kubicki, who lectures on AI & business of law, proposed a way of leveraging multiple GenAI models to speed up legal research which he describes as vibe lawyering (based on Wharton Associate Professor and Co-intelligence author Ethan Mollick’s concept of vibe working). He explained how he built a cross-model legal research toolkit by applying (in this order) Chat models (ChatGPT, Claude, Gemini etc), Reasoning models (Grok, Gemini’s Deep Research, OpenAI’s o1) and Agentic models (ChatGPT’s Code Interpreter etc). The key is the vibe – rather than prompt engineering, Kubicki suggests using voice mode to talk to them naturally, and letting the AI adjust to you

Beyond creating cool vibes, what do lawyers actually want from GenAI? This was one of the discussions at a workshop for in-house lawyers I attended last week. The corporate lawyers on my table agreed that the key objective was cost-effective growth, and the hope was that agentic AI would enable them to cover more work without needing more people. One delegate who travels regularly for work has created an alter-ego chatbot to deal with queries on a specific guidance document in his absence – effectively providing interim cover. This digital colleague is temporarily replacing a person. Is the next step, a permanent digital assistant? When it works, agentic AI is about doing a lot more with the same number of human lawyers. 

 

Legal Geek are hosting four conferences this year, learn more on our events page.

Written by Joanna Goodman, tech journalist

Photo credit (Joanna): Sam Mardon

share
Addleshaw Goddard Workshop

Level up your prompting game: Unlock the power of LLMs

A workshop intended to dive into the mechanics of a good prompt, the key concepts behind ‘prompt engineering’ and some practical tips to help get the most out of LLMs. We will be sharing insights learned across 2 years of hands-on testing and evaluation across a number of tools and LLMs about how a better understanding of the inputs can support in leveraging GenAI for better outputs.

Speakers

Kerry Westland, Partner, Head of Innovation Group, Addleshaw Goddard
Sophie Jackson, 
Senior Manager, Innovation & Legal Technology, Addleshaw Goddard
Mike Kennedy, 
Senior Manager, Innovation & Legal Technology, Addleshaw Goddard
Elliot White, 
Director, Innovation & Legal Technology, Addleshaw Goddard