To help cover what’s going down at ILTACon this year, the team at Draftwise has kindly offered to check out some of the sessions and startups there.
Key Session: ‘Charting Your Search Journey in the Age of AI’
Speakers:
Douglas Freeman, Enterprise Search Manager, Simpson Thacher
Yannic Kilcher, Chief Technology Officer, DeepJudge
Ilona Logvinova, Director of Practice Innovation, Cleary Gottlieb
Oz Benamram, Chief Knowledge & Innovation Advisor
Key topics: The current and future state of AI search in legal tech, importance of good, conceptual search on precedent, explaining the value and differences of search engines and LLMs.
Highlights:
LLMs are not search engines. Need to pair an LLM with good search and good inputs to get good results out. LLM model good for well-known facts, but not things not well-known or if it is changing daily.
Search engine will narrow down to 10 relevant answers, then LLMs can analyze. The ability to supply the correct context at the correct time to the LLM is what makes the difference.
Why search? Precedent is everything – without it there is nowhere to start. And search tools turn needle in a haystack into a ‘haystack full of needles’.
Legal tech now allows for ‘dynamic relationship with static documents’ – to learn more, uncover more from your past precedents.
Q&A – lots of focus on data security and concerns that their clients are resistant to using AI technology on their data (discussion of prompt insertion, locking down DMS, techniques to getting clients on board).
‘Precedent is everything in legal work – you can’t start anywhere without finding something first. Used to be like finding a needle in a haystack. Now good search gives you a ‘haystack full of needles’ (Ilona Logvinova) where you actually have multiple relevant examples instead of struggling to find one.
LLMs aren’t search engines. They’re great at well-known facts but terrible at finding specific documents in your system. You need search to find the 10 most relevant documents first, then feed those to the LLM for analysis. Getting the right context to the LLM is what makes all the difference. (Oz Benamram)
Oz also said clients will probably solve 60% of their questions with AI without ever calling you. Do you want them using ChatGPT or something built on your firm’s actual work?
Also interesting: we can now interrogate our old documents in ways that weren’t possible before. Instead of just storing precedent, you can actually extract intelligence from it. (Ilona Logvinova)
The Q&A got into security concerns and client pushback on AI:
Hallucinations and security risks with LLMs:
LLMs can hallucinate and output incorrect information if not provided the proper context and safeguards.
Security and compliance are critical when integrating LLMs, as they can access sensitive data if not properly restricted.
The security needs to be built into the tools/systems, not just the LLM itself, to prevent issues like prompt injection attacks.
Organizing and “cleaning” data for search:
As search capabilities improve, the need to meticulously organize and “clean” data decreases.
Connecting data sources and enriching document profiles is more important than rigid folder structures.
Proper matter/client profiling and closing processes are key to having well-structured data.
Lightweight LLMs vs. large frontier models:
Lightweight LLMs hosted within the security perimeter can be a viable option, as the key is providing the right context, not just model size.
Educating clients on how LLMs work (not learning from sent data) can help address concerns about data leaving the firm.
Applying LLMs to extract intelligence from past data:
Larger, more consistent data sets (e.g. insurance claims) make it easier to find patterns and derive insights.
Even smaller data sets can benefit from LLM-powered analysis to uncover correlations.
The overall theme was balancing the power of the technology (search, LLMs, agents) with the necessary security, governance, and data management practices to unlock the full potential for law firms.
Seems like a lot of firms are overthinking data protection when most clients actually want their lawyers to have institutional knowledge from past deals.’
—
Startup Alley
And then the Draftwise team spoke to two startups at the event. Below is a transcript of the interviews with someone at each company.
Lexara
Q: Alright, tell me about who you are and your company.
A: My name is Amy Swanner and I practiced law for a number of years. And my company, Lexara, I started because I saw how AI could benefit law firms.
One of the things that was the hardest when I was a lawyer was losing billable hours interviewing potential clients. And so I thought, this is a pain point that AI is perfect for. This is safe. This is compliant. But what we do, so we started off helping lawyers adopt AI safely and effectively in their law firms. But what we’re focusing on now is an AI tool called Lexara Engage. And it’s a potential client screener. So it uses AI to walk potential clients through a very guided interview.
And it gets relevant information from them after warning them that there’s no turning client relationship or anything. And it gets their information and then it does a high quality conference check in real time.
So you can tell if there’s a potential client and then it will quickly, if there’s a really high risk, it will guide the conversation to a close. If there’s not a high risk, conflict risk it will keep asking questions until it meets all of its goals or objectives and they’re different per practice area but you can also customize those based on what you want at your firm so if you want if you want personal injury clients but you don’t want somebody with head trauma it will screen that out you can make that a question runs a high quality confidence check and then on the back end the lawyers can look at see a summary of the case and then the whole transcript that they can download or archive or do whatever they want to with.
–
Crimson
Q: Can you tell me a little bit about what you do?
A: We’re Crimson. We’re a YC-backed start-up, currently in the A&O Sherman Fuse batch. And we’re focused exclusively on litigation, so we’re an AI platform for litigation arbitration teams, focused on the UK and the US. We’re based in London and Wales.
Q: What’s your customer base like?
A: Customer base? a wide range from litigation boutiques to regional law firms to firms like A&O Sherman. So it’s a real mix. Awesome. And then what are you looking to get out of Build2Con this year? Expand our footholds in the U.S. market and just learn about what customers want. Litigators in the U.S. care about what they’re looking for in AI solutions. Awesome.
Q: Why’d you found the company? I’m assuming you’re the co-founders.
A: We’re the co-founders. We’re two of the three co-founders – ex-Clifford Chance, Wilmer Hale, and Wilkie litigators. I experienced the problem every day for eight years and teamed up with David and Amin, who are both AI engineers to build the platform that I always wanted to have, essentially.
—
That’s all for now. Many thanks to the Draftwise team!
—
Legal Innovators Conferences in New York and London – Both In November ’25
If you’d like to stay ahead of the legal AI curve….then come along to Legal Innovators New York, Nov 19 + 20, where the brightest minds will be sharing their insights on where we are now and where we are heading.
And also, Legal Innovators UK – Nov 4 + 5 + 6
Both events, as always, are organised by the awesome Cosmonauts team!
Please get in contact with them if you’d like to take part.
Discover more from Artificial Lawyer
Subscribe to get the latest posts sent to your email.