LexBeat #1: AI agents in the legal market – and who controls them

This article launches the column LexBeat in the German Legal Tech Summit magazine. It is expected to appear every two weeks and puts current developments in the legal market into context. The focus is on the seven theme areas of the Summit: AI agents, AI governance, digital justice, cybersecurity and sovereign AI, legal operations, future skills, and start-ups and investment. The first issue looks at AI agents in legal work.

The new division of labour

In recent weeks, almost all the major providers in the legal market have unveiled so-called AI agents. These are no longer chatbots that simply answer a question. They are systems that read a statement of claim, propose a strategy, carry out research and draft the pleading themselves. Humans come into play at defined checkpoints: approving the plan, reviewing the research, signing off at the end. Control hasn't disappeared – it has simply shifted elsewhere.

That’s exactly the problem hardly anyone talks about. Until now, the work between checkpoints has been the work of newcomers to the profession. One in three HR managers plans to hire fewer junior staff by 2030.¹ But if you’ve never spent 40 hours working on a brief yourself, how will you recognise, ten years from now, when the AI agent has got it wrong? Oversight requires experience. And experience comes from exactly the work that is now being automated.

There is a second imbalance on top of this. Law firms are upgrading their capabilities, while the judiciary is not keeping pace. Courts are already receiving AI-generated submissions running to hundreds of pages in simple civil proceedings.² When agents are doing the writing on one side, and judges are reading with a highlighter on the other, that is no longer a level playing field.

None of this is an argument against AI in law. It can make processes faster and the law more accessible. But we need a vision of what training, everyday law firm life and the judiciary should look like in three to five years – before technology decides for us.

Promise and reality

Anyone wanting to understand how legal work is changing right now needs to look to the US. On 17 September, OpenAI unveiled “Astra for Law”: its most powerful model, tuned for legal work, with its own search index for judgments, statutes and procedural rules. At launch, there are 26 plugins from partners such as Thomson Reuters, iManage, Harvey and Legora. On 200 research tasks involving US law, OpenAI says the model performed 40 per cent better than the same AI using a standard web search. However, there is no comparison with traditional legal databases.³

Harvey and LexisNexis demonstrated what this means in practice in early October. Their new workflow for motions to dismiss works like this: you upload the complaint. A planning agent proposes up to eight possible points of attack and rates each for strength. The lawyer approves the strategy, further agents carry out research, and a final agent drafts the document. Until now, an associate would typically need 40 to 50 hours for such a motion.⁴

That same week, US law firm Cooley announced it was building an agent with Google that searches court files for confidential data and suggests redactions. What actually gets redacted is still decided by lawyers.⁵ And AT&T is having its new in-house legal agency, LegalEdge, develop AI workflows together with OpenAI.⁶ The major AI labs are no longer just selling to software providers. They're going straight into law firms and legal departments.

The trend has reached Germany too, just more quietly. BMW is rolling out Legora across its entire legal department: for research, contract review, process management and first drafts based on its own standards, directly in Word. Legora opened an office in Munich at the end of September. According to the company, more than 13,000 lawyers in Germany now work with the programme – up from around 3,000 at the start of the year.⁷ Legal publishers are also restructuring: Wolters Kluwer has integrated international content on arbitration, competition and IP law into its AI workspace Libra, across eleven European countries.⁸ The logic behind it: an agent is only as good as the sources it can access.

The truth is that there's a gap between press releases and everyday life at the firm. The industry publication Artificial Lawyer notes that most large law firms have long since adopted AI tools, yet their way of working has barely changed.⁹ One obvious reason is the business model: if you bill by the hour, you have little incentive to turn 40 hours into four. Law.com also reports that the interfaces promised between tools often don't work as advertised. Gaps like these increase the risk of data leaks.¹⁰

The pressure is therefore coming from clients. AT&T plans to cut its list of outside law firms from around 30 to 18–20 within two years, favouring firms that use AI themselves.¹¹ And at the other end of the agents stands the judiciary. Federal Minister of Justice Stefanie Hubig recently told the Rheinische Post: “It cannot be right for courts to be literally buried under unstructured AI-generated text.”¹² Her ministry intends to put forward a reform within six months so that courts can give parties clear guidance on how to present their submissions.¹³

The solutions are on the table – they’re just uncomfortable. Control points that actually check rather than just rubber-stamp. Training in which young lawyers learn to review an agent instead of simply being replaced by one. The US law firm Ropes & Gray has first-year associates spend around 20 per cent of their billable-hour target on AI training.¹⁴ And a judiciary equipped with powerful AI of its own. The agent drafts in minutes. The signature remains human. The question is whether the person behind it still knows what they are signing.

Outlook towards the Summit in December

Today and tomorrow the industry is meeting at Legal Geek in London, on 25 and 26 October ClioCon follows in Boston, and on 4 and 5 November Legal Innovators UK. It will be interesting to see whether law firms there show what agents really deliver in everyday practice, or whether it stays at the demo stage.

On 19 November, the Conference of Justice Ministers meets in Hamburg. The next federal-state digital summit on the judiciary is also expected in the autumn. The federal government intends to provide a total of €210 million for digitalising the judiciary between 2027 and 2029.¹⁵ That may sound like a lot, but it amounts to €70 million a year for the judiciary across the federal government and sixteen states. The BMJV’s reform targeting AI-generated court filings is expected by spring at the latest.

The German Legal Tech Summit comes to the HCC Hannover on 3 December. Three of the solutions presented here can be experienced up close there: Harvey, Legora and Libra, Wolters Kluwer’s AI workspace, will be exhibiting on site. In the Agentic Legal AI topic area, the Summit showcases concrete applications for research, contract analysis and document drafting – an excellent setting to explore the questions raised in this article directly with providers and users.
The judiciary, which plays a central role in this article, also has its say: Lower Saxony’s Minister-President Olaf Lies speaks about digital justice, democracy and access to the law.

Figures of the week

$15.5 billion – that’s the valuation US company Harvey has reached following its latest funding round. On 9 September it raised $550 million, after $200 million in March at an $11 billion valuation. Annual revenue stands at around $400 million, and the valuation has more than tripled in just over a year. What’s the money for? According to COO Katie Burke, mainly two things: people and computing power, as Harvey expects a race for data centres next year. Whether this is a bubble is openly debated in the industry.¹⁶ One thing is certain: whoever controls computing power will help determine what legal work costs in future.

79 per cent – that’s how many business executives cannot reliably explain why their contracts contain the clauses they do. This is according to a Censuswide survey commissioned by Luminance in the UK, the US, France and Germany. 98 per cent report financial or operational consequences when contract knowledge is lost. Only 26 per cent can search their entire contract portfolio quickly and reliably.¹⁷ This is exactly where Document Intelligence comes in. In fairness: the study comes from a provider that sells precisely this kind of solution.

33 per cent – that’s how many experienced legal professionals trust the results of their AI. At the same time, almost half of legal organisations have already embedded AI firmly into their workflows.¹⁸ Every second organisation uses it, yet only one in three professionals trusts it. This isn’t a technology problem. It’s a trust problem, and trust isn’t built by a press release.

36 – that’s how many senior partners the US law firm Vorys has now recreated as AI personas, in collaboration with a lab at Stanford Law School. Colleagues can have these digital partners review documents, suggest amendments and flag contractual risks. Next up: personas of clients’ legal departments.¹⁹ The knowledge of experienced lawyers is being preserved. What remains unclear is where the next generation of experienced lawyers will come from.

Reading tips and short news items

For those who want to dig deeper: Legal IT Insider has written up the launch of Astra for Law in detail, including the question of how much of a workflow the AI handles before a human needs to step in.²⁰ Artificial Lawyer explains in “Legal AI vs The Adversarial Incentive” why large law firms still work much the same despite AI.²¹ And LTO covers the whole debate about AI-drafted pleadings and the planned reform of procedural law, with voices from every parliamentary group in the Bundestag.²²

A few more recent developments in brief. Legora has launched “Skills”, a library of more than 250 legal work instructions for AI agents²³, and “Intake”, an AI entry point for legal departments in Slack.²⁴ Ironclad has unveiled its own contract agent that draws on the company’s contract knowledge.²⁵ Ivo has released “Sage”, a freely available model for contract work that law firms can customise themselves.²⁶ And in China, Z.ai has launched its own language model dedicated solely to legal tasks.²⁷