Following the money
Now that lawtech, and specifically legal AI, is recognised as a mainstream investment, money is changing the shape of the sector at least as much as advances in technology. In February, legal AI saw record fundraising, with multiple announcements revealing a complex network of funding interrelationships. For example, OpenAI backed GenAI assistant Harvey raised $300m in a Series D funding round with investors including REV, a VC partnership backed by LexisNexis parent company RELX Group. Meanwhile LexisNexis announced its own (separate?) partnership with OpenAI, to fine tune OpenAI models for legal workflows. Legal AI pioneer Luminance raised $75m in Series C funding, with investors including its original backer Slaughter and May.
It’s not just easier and faster for AI start-ups and scale-ups to raise money. Legal tech is also experiencing consolidation, with February’s deals including US legal tech giant Litera acquiring UK legal workflow provider Peppermint Technology. Rapid advances in technology – such as what’s happening with GenAI – highlight the value of legal design, which makes both legal terminology and legal tech tools and applications more user friendly, and last month integrated contracting ALSP Factor Law acquired high-profile US legal design and development agency, Theory and Principle.
Not so secret agents
Agentic AI could potentially be a game changer, introducing digital co-workers into legal, alongside human lawyers and (human, for now) legal engineers and technologists. The main opportunity and challenge is that agentic AI is scalable – by configuring multiple AI agents to run and integrate routine workflows and processes – so it is only a matter of time before GenAI starts replacing human roles and some ALSPs. Several agentic AI platforms for legal are rapidly gaining traction, notably Harvey, in-house legal agentic AI start-up Eudia which raised $105m Series A funding in February, and Legora (recently rebranded from Leya) which is being rolled out across Bird & Bird global offices and has established a collaborative partnership with Mishcon de Reya. Mishcon de Reya also invested in agentic start-up Ctrl AI which supports in-house legal operations at large and medium enterprises.
Performers or reformers
Although it feels like intelligent GenAI models with new competencies are released every other week, the one thing they don’t seem to be doing is increasing productivity – in legal and elsewhere. Azeem Azhar’s Exponential View newsletter identified AI’s productivity paradox as the factor that is preventing AI from becoming a general purpose technology – i.e. a societal game changer like the smartphone or cloud computing (which enables apps). Reading between the lines of the various legal AI reports, although AI can handle routine tasks and improve efficiency, so far it isn’t boosting productivity and its value is predicated on accuracy rather than speed (because of the need to check its output). Nor is it replacing current lawyer or technologist roles or incumbent legal tech tools.
So is GenAI for legal more performative than game-changing? i.e. is it looking for attention or efficiency? In response to this mostly unasked question, both sides of the Atlantic have seen a move towards AI benchmarking. The latest buzz is around the Vals AI benchmarking study, which compared the performance of Harvey Assistant, Thomson Reuters’ CoCounsel, vLex’s Vincent AI and Vecflow’s Oliver against each other and against a Lawyer Baseline – a control group of human lawyers – in seven tasks set by a consortium of eight law firms. The vendors signed up to the tasks that matched their products’ capabilities, and the results varied by task. The AI tools outperformed the Lawyer Baseline on four tasks related to document analysis, information retrieval and data extraction and matched the Lawyer Baseline on chronology generation. The Lawyer Baseline beat the AI tools on two tasks, EDGAR research (which involves multiple research steps and iterative decision-making) and redlining. Vals AI is a US study, but it acknowledges support and guidance from UK organisation LITIG’s AI benchmarking team.
However, while benchmarks may help to guide procurement decisions in an evolving and confusing market, they may equally encourage some law firms to revert to legal’s traditional fast follower tactics, which is less risky than innovation and requires fewer resources.
The lawtech start-up sector too risks becoming a victim of its own success. While thanks to AI attracting a huge influx of funds, it is growing faster than ever, strong M&A activity in the first quarter of 2025 shows that scaling up may be the biggest challenge – for the start-ups themselves, and also for legal innovation. Classic management education says that all business plans should include an exit strategy, but if all or most of the game-changers are acquired, lawtech innovation is likely to slow down.
Talking of game-changers, as it’s International Women’s Day this Saturday, I’m jumping on a familiar soapbox – gender equality – noting that several leading lawtechs with female founders have been acquired in recent months. That’s great news for them, but they leave
an important gap in the lawtech start-up ecosystem. Last year, when I joined the UK government’s legal tech trade mission to the US, I met only one female founder and one female co-founder. It’s important to attract more female founders to lawtech and this means rather than bringing the same token representatives to every new initiative, the lawtech community needs to consider new entrants as serious investment prospects and give them air time to promote their ideas and products.
GenAI predictions are as easy as AI BaZi
And finally, given the expanding list of legal use cases for GenAI, a feature in this week’s MIT Technology Review made me consider the possibility of using large language models to generate lawtech predictions. In China, Gen Z are prompting DeepSeek R1 to tell their fortunes, using the traditional Chinese BaZi system (the Four Pillars of Destiny) that uses people’s birth date and time to predict career, relationships and financial fortunes by analysing and interpreting the balance of the elements in their charts. As with other large language models, the reliability and credibility of GenAI fortune-telling depends on the quality of the prompts and the level of detail. This has led to fortune-telling platforms like Fatetell which combines prompt engineering with multiple AI models to generate BaZi readings, and whose founder Levy Cheng has a background in legal AI. I wanted to use it to make predictions about current lawtech start-ups and scale-ups based on when they were founded (obviously we have the date, but not the time) but as yet I haven’t been able to get further than the landing page…
Legal Geek are hosting four conferences this year, learn more on our events page.
Written by Joanna Goodman, tech journalist
