What Everyone Is Still Figuring Out
Last week I was at Future Lawyer UK, three days of interviews, conversations, and a floor full of legal tech vendors and lawyers trying to figure out what to do with all of this. One question kept coming up in almost every conversation I had, whether it was with a solo practitioner, a mid-sized firm’s innovation lead, or a legal tech founder.
Do I build something custom around Claude skills? Do I buy something “off the shelf”? Do I stitch together what I already have? Or do I hire someone to put the whole thing together for me?
And it’s not just the conference, this is pretty much the thread running through most of the conversations in the space currently.
Nobody I spoke to feels very confident about the answer, and I think there is a good reason for that. The market is moving fast, the tools keep changing, and between vendor pitches, online commentary, and case studies there is a lot pulling you in different directions.
What I have been finding useful when I walk through this with firms is stepping back from the tool decision itself and starting from what you are actually trying to solve.
The Three-Step That Tells You What to Do
First, see where your existing setup is actually getting you. Before bringing anything new in, it helps to push what you already have as far as it will go. Claude and Claude skills, whatever vendor product you already have a seat on, your CRM, your document management system. When you know where the ceiling is, the next decision becomes a lot more informed.
Second, get specific about what is not working. This is the step that makes all the others easier. “Claude can’t do this” is usually too broad to act on. Is it a capability gap, a data access issue, or a reliability problem on a type of work where one wrong output matters? Each of those leads somewhere different. A reliability gap points you toward a different tool. A data access issue can often be solved with a connector. A capability gap sometimes calls for custom work, though there is often a simpler path too.
Third, and only then, pick your move. When you know where your existing setup stops and exactly why, the choice between buying, stitching, building, and hiring becomes a lot clearer. You have a specific problem in hand, which is a very different starting point from trying to pick the right tool off a list.
Where The Sequence Gets You
The pressure to act on AI right now is real. Vendors are moving fast, your peers are sharing wins on LinkedIn, leadership wants something to show, and nobody wants to feel like they are falling behind. That pressure is a big part of why so many firms end up at a tool decision before they have really looked at what they already have. It is completely understandable, and I have sat with plenty of lawyers and firm leaders in exactly that position.
What the sequence does is slow the decision down in the right places. It surfaces the capability you already have before you buy more of it. It sharpens the problem you are actually trying to solve before you go looking for solutions. And when the answer does turn out to involve custom work or a new vendor, you walk into that conversation knowing exactly what to ask for, which usually means the project moves much faster and the result actually gets used.
Where Stitch, Buy, or Build Actually Sit
Once the sequence is clear, the four moves start to separate out naturally. Here is what I have been seeing work in practice.
Buy when someone has already solved your problem commercially and the vendor has done the compliance and engineering work. If you are doing heavy contract review, a product like Legora is going to outperform what you can put together on your own. The vendor’s team becomes your leverage.
Stitch when the capability already exists across your current tools and the only real gap is that they are not talking to each other. Connecting Clio to Claude through an MCP server is cheap, fast, and reversible. So is routing Claude through your DMS. Stitching does not get talked about as much as building, but for a lot of firms it is the move that actually sticks.
Build custom when you have a workflow that is very specific to how your practice runs, will not change the next quarter, and the existing tools cannot get you to the reliability or the specificity you need. Custom work also comes with a maintenance commitment, which is worth factoring in from day one.
Hire an advisor when you don’t know where to start, when you don’t want to do it yourself and maintain what you have on your own, or when the decision making involves judgement calls that nobody on your team has made before. A good advisor is with you through the whole thing, from the decision tree at the start, the rollout in the middle, and the follow-through once it is live and being used.
When firms take the time to sit with their limits first, the decision across all four becomes a lot clearer. Everything else downstream gets easier from there.
Legal AI in Action
🎥 From the Conference Floor
Last week I attended Future Lawyer UK in London as a media partner. Three days of interviews, conversations between sessions, and more content than I could possibly send out in one newsletter.
On day one I caught up with eight legal tech vendors - LexisNexis, Spellbook, Legora, Harvey, Aloi, Clio, Draftwise, and PointOne - and gave each three questions. Their responses together sketch out where the market actually sits right now.
On day two, I spoke with attendees between sessions about what they were taking away from the conference.
And then, on day three, I sat down for longer conversations with:
- Timo Karakashev, Founder & CEO of Cosmonauts - the team behind Future Lawyer UK
- Elgar Weijtmans, Head of Technology at HVG Law, CTO and AI Researcher at LegalBenchmarks and lecturer on AI at Universiteit Leiden and on Sdu’s Certified AI Strategy & Implementation Legal program
- Pranav Anand, EMEA Digital Forensics, Incident Response and Cyber Investigations Lead at Gravity Stack
- Melina Efstathiou, Strategic Advisor on AI Governance & Legal Tech, LDI Architect at Legal Data Intelligence, and Board President of the ACEDS UK Chapter
Each of those interviews is dropping over the coming days on YouTube and my website below:
All the coverage from Future Lawyer UK 9 →
🎬 Claude’s Compliance Gaps Were the Question. Here’s the Answer.
What are the three different compliance profiles of three main Claude products, how to build the monitoring and governance layer that Anthropic doesn’t ship with two of them, and what the full rollout sequence looks like from a 10-lawyer firm to 300+.
🎙 Enhancement Over Efficiency: The Mindset Shift for Lawyers Using AI
Kyle Bahr is AI Innovation Manager at Cleary Gottlieb and a lawyer with 23 years across federal clerkship, litigation, and now AI adoption at one of the world’s top global firms. We talk through why enhancement is a sharper goal than efficiency when lawyers work with generative AI, what the 1,000+ cases in the hallucinations database actually reveal about who is making the mistakes (not who you would think), and what the new cognitive load of managing AI agents means for a profession that was never trained for it.
Coming up!
🗓 Front Foot Webinar: May 13
On May 13, I am invited to speak on Front Foot’s CPD webinar to talk through where AI agents actually are in legal - what’s possible, what’s advisable, and where the risk sits. Less “will AI take your job” and more “here’s what your firm can actually start doing now”.
If you can’t come live, catch it on demand.
👇 Register below:
CPD Webinar: Are the robots ready?!
🎙 Next Tuesday at 2pm CET!
Next week’s guest on Rok’s Legal AI Conversations is Damien Charlotin, a lawyer, data scientist, and lecturer whose AI hallucinations database has become the most referenced resource on the topic in the legal world.
We discuss what 1,300+ hallucination cases actually reveal about who is getting caught (and who isn’t), why specialized legal AI tools can sometimes produce more hallucinations than the generic ones, how AI is opening up kinds of legal work that used to be too expensive or time-consuming to attempt, and why the predicted consolidation of the legal market may not play out the way everyone assumes.
Each edition of Legal AI Brief brings practical lessons from firms using AI safely.