With generative AI becoming a common fixture of law firm tech in 2025, how will this shape law firms – and their lawyers – in 2026?
As year-end approaches, it is worth looking at current trends and where they might lead. This was the year when GenAI became a key component of the legal tech stack. It is likely that by next year, legal services delivery will have changed quite dramatically in just a few years. A significant factor shifting the profession is the close connection between Microsoft and OpenAI. It is worth noting that since OpenAI’s restructuring last month, Microsoft owns 27% of OpenAI, worth about $135bn, and receives 20% of its revenue.
Pretty much every discussion about GenAI in law firms and legal departments cites Microsoft Copilot as an entry point for lawyers and business support professionals. Nearly all law firms
deploy Microsoft 365 and therefore have easy, secure access to Copilot. ChatGPT is the most popular GenAI platform. Most firms and legal departments which have developed their own in-house GenAI capability have effectively created a wrapper for OpenAI’s GPT series to give their lawyers a secure ChatGPT behind the firewall. OpenAI and Microsoft mean that lawyers can start working with GenAI in a familiar, secure environment.
Microsoft and OpenAI also underpin the latest breed of legal GenAI unicorns. The OpenAI start-up fund was an early investor in agentic AI platform Harvey, founded in November 2022. Harvey recently expanded its integrations with Microsoft applications, including Copilot and Word. Harvey’s competition for the legal market, agentic AI platform Legora, which was founded in Stockholm in June 2023, is built on Microsoft Azure and can also operate within Microsoft applications.
Funding boom – or bubble?
As venture capital funding for legal tech start-ups surged in the third quarter of 2025, there have been murmurs about whether this is a watershed moment or a bubble about to burst. Both Harvey and Legora, which have seen spectacular growth, announced $150m funding rounds. Harvey’s Series F, led by Andreessen Horowitz, increased its valuation to $8bn. Legora achieved unicorn status with its Series C, which was Europe’s largest-ever legal tech VC round, and increased its valuation to $1.8bn. And as LawSites’ Bob Ambrogi reported last week, specialist US venture capital firm The LegalTech Fund closed its second fund at $110m, with investors including major law firms and vendors. While The LegalTech Fund is based in the US, its recent investments include London-based Wexler, an AI platform for complex litigation workflows.
However, not all legal AI start-ups are thriving. At the end of October, it was reported that Anthropic-based Robin AI was listed for sale on a bankruptcy website following its failure to secure $50m in Series C funding. It later emerged that it is facing a winding-up petition from HM Revenue & Customs. There has been much discussion over whether this is an outlier or a harbinger, but it is surely a sign that the legal tech start-up space is becoming crowded, and not every investment is profitable.
Force for change – and a giant step
Harvey and Legora focus on the top end of the legal market, targeting large law firms and government institutions, and have quickly established an international client base. TechLawFest in Singapore showcased Harvey’s partnership with the Singapore Supreme Court and Legora’s partnership with the Singapore Academy of Law. Both companies participated in the start-up event Slush last week.
'The vertical software opportunity today is building a deeply connected suite of applications that work seamlessly together, are deeply integrated, eliminate context switching and the friction of moving between applications'
Jack Newton, Clio
GenAI has also seen significant developments in the small to-mid-size market segment. Jack Newton, founder and CEO of legal tech unicorn Clio, spoke at Web Summit in Lisbon just two days after closing legal tech’s biggest deal ever – the $1bn acquisition of Spanish-owned AI legal intelligence research company vLex. On the same day, Clio closed $500m in a Series G investment round, along with a $350m debt facility, boosting its valuation to $5bn. This is a giant step towards the democratisation of legal GenAI. It creates a company with $400m in recurring annual revenue and a customer base of 400,000 legal professionals in 130 countries.
Newton’s session was about Clio’s strategy of ‘taking a vertical approach [to building software] and going really deep on a specific industry and building technology that deeply addresses the needs of that industry and focusing on a strong product market fit’, which is also the approach taken by the agentic AI vendors who are dominating the market.
Another key point was interoperability. ‘People increasingly want an integrated suite of tools. The vertical software opportunity today is building a deeply connected suite of applications that work seamlessly together, are deeply integrated, eliminate context switching and the friction of moving between applications.’ Newton referenced venture capital partner
Sarah Tavel’s 2023 article AI start-ups: Sell work, not software. This underlines that rather than helping people work more efficiently, AI software is about helping people work differently and potentially replacing labour.
When AI decides
Although there were relatively few legal tech sessions and start-ups at Web Summit, sessions that looked beyond agentic AI provided food for thought. Tao Zhang is co-founder and chief product officer at Singapore-based start-up Manus AI, which launched the world’s first general AI agent. He was interviewed on stage on the topic of what happens when AI starts to decide. Zhang demonstrated how Manus AI uses a multi-agent architecture to find the best way to execute end-to-end tasks and processes without predefined workflows – or prompts. Manus works autonomously. You simply give it a high-level goal and it deploys a series of specialised sub-agents (planner agent, execution agent, knowledge agent and verification agent) to break down and complete complex instructions, and check its own work, with minimal supervision. While Mazur precludes the possibility of an AI lawyer, it seems there is now the technology to build one.
GenAI is replacing people
Almost exactly three years after OpenAI launched ChatGPT, AI is replacing people in law firms. While it is not replacing lawyers yet, GenAI is making steady inroads into the back office. Last Friday’s Financial Times reported that Clifford Chance is cutting 10% of its business services staff in London due to increased use of AI. Earlier in the year, there was a similar announcement from BCLP affecting 8% of back-office staff, with its global chief operating officer referencing ‘new ways of working, including investing in… digital solutions and market-leading technology’. Last week, Irwin Mitchell abolished the litigation assistant role, again citing AI (possibly also influenced by the Mazur decision). While qualified lawyers’ roles are indeed protected by regulation, AI is upending the law firm hierarchy. With GenAI and agentic AI being deployed for (administrative) legal and business processes, key business support roles ranging from marketing and business development to financial and risk management have evolved too.
AI-first firms need AI-fluent lawyers
And while there are already AI-first firms, such as Garfield AI in the UK, transitioning to AI-first across the sector requires AI-fluent lawyers. One dilemma arising from AI taking over routine tasks, such as reviewing contracts, is that this work was traditionally used to train junior lawyers who still need to work through straightforward legal processes to ensure they understand them. The solution, unsurprisingly, is more AI. Last month, Kennedys announced a strategic collaboration with GenAI platform Spellbook to help junior lawyers build the skills they need ‘to thrive in a market where many traditional entry-level tasks are increasingly automated’. According to Kennedys’ website, the ‘training programme will use simulated scenarios and AI-assisted drafting exercises to replicate learning opportunities that may start to disappear in the workplace’.
Other firms are taking a proactive approach to AI fluency. US firm Ropes & Gray is encouraging junior lawyers to allocate billable time to training on AI tools.
Another step forward is GenAI vendors working with law schools to encourage AI fluency even earlier. This is not a new idea. Law schools have trained their students using research tools such as LexisNexis and Westlaw, and practice and case management software such as Clio or LEAP. Last week, it was announced that four prominent UK law schools are partnering with Harvey, integrating its AI tools into their teaching programmes. Harvey has similar arrangements with more than 25 US law schools.
Still in transition
Notwithstanding the rapid trajectory of GenAI, legal is still in transition. Here in the UK, while adoption is continuing apace, there is still a risk that introducing automation can open up justice gaps. For example, delays caused by the government’s road traffic accident portal, which was introduced to facilitate/speed up the claims process, is one of several factors discouraging accident victims from claiming the compensation to which they are entitled. However, when automated processes work smoothly, they should reduce claims backlogs. This is happening in Singapore, where some of the Singapore Supreme Court’s self-service capabilities, including one that provides guidance around RTA cases, are built in partnership with Harvey AI.
While AI is replacing humans in some roles and functions, another factor making AI ‘business as usual’ [normalising AI?] is the emergence of new career paths as firms and enterprises invest in AI leadership and AI champions. Simmons & Simmons has an international network of AI champions. Just this week, Linklaters launched its first cohort of 20 dedicated AI lawyers. While 2025 saw the emergence of AI-first law firms, if these are going to become a market force rather than a category of specialist boutiques, they will have to address the perennial software challenges of integration and interoperability.
No comments yet