AI Strategy for Law Firms: The Path to Becoming AI-First

AI-first means this: for every recurring task, AI support is the default option, not the exception you experiment with when you happen to have spare time. For an established law firm, that doesn’t mean reinventing itself from scratch. It means making AI the standard for specific, well-defined tasks.

63.6 percent of German law firms now actively use AI, mostly ChatGPT or a comparable generative tool for research and first drafts. Internationally, the latest Legal Industry Report from 8am shows a similar pattern: more than four in ten firms have no AI policy, and more than half offer no training at all. The gap isn’t between firms that use AI and firms that don’t. It’s between individual lawyers quietly working with ChatGPT on their own and a firm that treats AI as part of its strategy. That gap is exactly why “AI-first” has landed on the agenda of partners who waved it off as hype two years ago.

What “AI-first” actually means for a law firm

“AI-first” often gets confused with “we have AI tools.” Those aren’t the same thing. A firm with five ChatGPT licenses and no rule for what they’re allowed to be used for isn’t an AI-first firm, it’s a firm with an unresolved governance problem. AI-first means this: for every recurring task, AI support is the default assumption, not the exception you try when you happen to have time. That applies to deadline checks as much as to the first market research on a new client. The difference between “we use AI” and “we are AI-first” is the difference between a tool in the cupboard and a tool on the desk.

Can an established firm actually become that?

Worth being honest here. Some firms today were built as AI-native businesses from day one: no hourly-billing culture, no paper deadline diary to digitize first, no partnership structure that has to sign off on every process change. For those firms, AI-first isn’t a transformation, it’s the default setting they started with.

A mid-sized firm with long-standing client relationships, liability exposure, and a partnership model is a different animal, and it shouldn’t measure itself against whether it ever looks like one of these startups. It won’t, and it doesn’t need to. What’s realistic is making AI the default option for individual, clearly defined tasks, without reinventing the entire firm. The ambition “we’re going to become fully AI-native” is the wrong benchmark for most established firms, and it’s usually also the reason these projects lose steam once the initial enthusiasm wears off.

The business case for partners

Firms that actively use AI mostly report two effects: higher efficiency and, trailing a bit behind, higher profitability. The two are related but not identical, because efficiency gains only turn into more revenue if the billing model keeps pace. A firm that bills by the hour and finishes the same task in a quarter of the time earns less at first, not more. That’s why many firms are rethinking flat fees or value-based pricing alongside their AI rollout. On top of that comes pressure from clients who use AI themselves and increasingly ask why a routine review still takes their law firm three days.

The legal guardrails in Germany

For German law firms, the legal framework isn’t a side issue, it’s the starting point for every decision. Section 203 of the German Criminal Code (StGB), which protects client confidentiality, doesn’t ban AI use, but it demands more than a standard GDPR data processing agreement: the AI provider also has to sign a confidentiality commitment backed by criminal liability, because Section 203 is criminal law, not data protection law. The German Federal Bar Association’s December 2024 guidance (in German) sums up the obligations in three points: independent final review of every AI output, maximum anonymization of queries and documents, and a duty to disclose AI use to clients and courts.

How seriously this is taken became clear on 20 November 2025, when the Berlin Court of Appeal cautioned a lawyer (case no. 17 WF 144/25, decision in German) for citing two court rulings in a family-law appeal that simply don’t exist, which the court described as apparently the result of a “fantasizing” AI. The court issued a formal warning and pointed to the professional duty under Section 43 of the German Federal Lawyers’ Act (BRAO) to check every citation before using it. Liability stays with the lawyer. The EU AI Act generally doesn’t classify legal AI use as high-risk, but it does require transparency toward clients, and it effectively rules out standard consumer tools like ChatGPT without a data processing agreement once client data is involved.

The roadmap: the first six months

A transformation of this size rarely happens in one move, and it isn’t finished after the following six months either, it’s just getting started. One approach that has worked in practice splits the initial rollout into four phases:

• Pilot (days 1 to 60): Two or three lawyers, one clearly defined task with high volume and low risk, such as research memos or document summaries. One person owns it and measures time before and after. Two months instead of the often-cited 30 days, because the first few weeks usually only show whether the tool fits the task, and only after that whether it actually sticks in daily practice.
• Infrastructure (from day 31): This phase deliberately starts while the pilot is still running, not after it ends. Individual use gradually becomes a shared system: prompt libraries get shared, and two or three related tasks get added once the first pilot results are in.
• Governance (days 61 to 90): Rules before scaling, not after: who reviews AI output, how AI-assisted work gets billed, which data is allowed into which tool.
• Specialized platforms (days 91 to 180): Tools matched to the specific practice area and task, not to whichever brand has the biggest marketing budget.

What doesn’t fit neatly onto a start date in this framework: culture, training, and communication. They don’t kick in after the governance phase, they run alongside all four phases from day one. A firm that only starts explaining why this is happening, and what it means for each person’s job, after the pilot is done has already lost trust that’s hard to win back. Short, regular formats tend to work better than one big kickoff event: a ten-minute slot in the weekly team meeting where someone shows what’s working and what isn’t.

And even if all four phases go as planned, the firm isn’t AI-first after six months. It has just started becoming one. Getting to the point where AI use is genuinely routine, not the exception still being discussed separately in partner meetings, takes more like two to three years than six months, based on comparable digitization projects. A firm that doesn’t tell its partners and staff that honestly from the start risks disappointment at exactly the moment the first win should be getting celebrated, when instead it becomes clear how much of the road is still ahead.

The most common pitfalls

Most failed AI projects at law firms don’t fail because of the technology. They fail because nobody was clearly responsible, because AI ran alongside the old process so nobody could tell whether it actually saved time, or because governance questions only got asked once ten lawyers had already built their own uncontrolled workflows. And one point is consistently underestimated: without training, AI use stays a matter of chance. If you want AI use to become a habit, you have to budget time for it the same way you would for any other new piece of software in the building.

Culture change and the partners’ role

Partners set the tone, whether they mean to or not. When a partner is open about using AI and explains what they use it for and what they don’t, associates are more likely to do the same instead of hiding it. One issue that doesn’t get raised often enough: if junior lawyers only ever sign off on AI output instead of working through the underlying research themselves at least once, they never develop the instinct for when a result is wrong. The question every partner should ask with every new AI application isn’t just “does this save time,” it’s also “what skill is nobody learning right now.”

Case study: KUCERA Rechtsanwälte

KUCERA Rechtsanwälte, a mid-sized German firm specializing in real estate law, tracked deadlines until recently with a paper deadline diary that the front office checked daily and that got read aloud again in the weekly team meeting. Using the no-code platform e! by Lexemo, an in-house, three-person legal engineering team built four bots themselves, without outside development: for document analysis, commercial register searches, notarial client intake, and automated lease drafting. The result: more than 20 percent time savings on the affected tasks and over 1,000 documents generated automatically.

Matthias Fridriszik, an attorney and legal engineer at the firm, put it this way: “With e!, we’ve turned repetitive manual steps into digital assistants, so we can get back to doing what we actually became lawyers for.” That’s the AI-first principle from this article, scaled down to a single firm: don’t reinvent everything, automate clearly defined, recurring tasks in a way that’s auditable and ISO 27001 compliant, and put the time you free up back into the actual legal work.

Frequently asked questions

What does “AI-first” mean for a law firm?

AI support is the default option for recurring tasks, not the exception. What matters isn’t how many tools a firm uses, it’s whether using them is routine.

Can an established firm actually become AI-first?

Becoming fully AI-native like a startup is rare for an established firm. What’s realistic is making AI the default option for clearly defined tasks, rather than reinventing the entire firm.

Is using ChatGPT at a law firm compatible with client confidentiality?

Section 203 of the German Criminal Code doesn’t ban AI use outright, but on top of a standard GDPR data processing agreement, it requires the AI provider to sign a confidentiality commitment backed by criminal liability.

How long does it take before a firm is actually working AI-first?

A first structured rollout can be done in around six months. Getting to the point where AI use is routine firm-wide takes more like two to three years.

Does the EU AI Act classify law firms as high-risk?

Generally not. The EU AI Act usually doesn’t classify legal AI use as high-risk, but it does require transparency toward clients.

Conclusion: the next step

A firm doesn’t become AI-first through a policy decision at the next partner meeting. It becomes AI-first through one well-chosen task that gets automated first, with one person responsible for it, and a time measurement before and after. Governance and legal safeguards belong in the process from the start, not as an afterthought once it’s scaling.

If you want to figure out which task in your own firm is the right place to start, and what a no-code rollout with e! by Lexemo could actually look like, that’s a conversation worth having with Lexemo, no strings attached.

As of: September 2026