A legal brain for every business team
London
Joined August 2018
- Tweets488
- Following661
- Followers2.5K
- Likes232
Three legal AI launches. Three products called Intake.
How much should AI be allowed to do on its own?
As much as your rules allow. You set the rules and the escalation points. Only the calls that need a person reach one.
Who sets the rules when AI acts?
AI can draft and negotiate faster than ever. Knowing it followed your playbook every single time is the harder part.
Prompts aren't policy.
We're at Legal Geek Wed and Thu, Old Truman Brewery. Rafie's on stage. Booking link below.
#LegalGeek
Book 20 minutes with the team: calendly.com/d/d2pz-r2x-bxv/…
In law, "mostly right" is a liability.
Imad Nazar (ex-Skadden M&A, now a legal engineer at GenieAI) and our CTO Nitish Mutha on why legal AI is harder than code.
Out of the Bottle, Episode 7 🎧
Full episode here: youtu.be/zlLVwEskHpY
We're at Legal Geek next week: The Old Truman Brewery, Wed 14 and Thu 15 Oct.
Bring us the AI question, challenge or idea you're working through; we'd love to hear it.
Grab 20 minutes with Naz, link down below. #LegalGeek
"We have to abstract ourselves away from the day-to-day work faster than AI progresses."
Rosie Dent-Erdoğan (Product Creator, @GenieAI) on the death of the Figma, five agents and one human brain, and what tokens actually cost.
EP4 of Out of the Bottle:
youtu.be/ZNQF2qAhKTk
"You wouldn't hand this volume of company know-how to anyone else. Except in this current state of hype."
Dr @AbbyGKGilbert (@FutureWorkInst) on economic sovereignty, good work vs bad automation, and the AI risk we're not talking about.
Check the link below👇
New episode of Out of the Bottle is out now.
Here is the link: youtu.be/n3drtixsEr4
#AI #FutureOfWork
Episode 2 of Out of the Bottle out now 🍿
@RafieFaruq sat down with Sophie Catto, Joint Managing Director of AllBright everywoman, the largest ecosystem of women in business, 500,000+ strong, to discuss storytelling, AI, and the future of work.
1/5 Introducing Live Steering for Agentic loop. Most AI agents give you one way to change course mid-task: stop, and lose the work. With long-running agents that is brutal: a single workflow can run for twenty-plus minutes, and aborting it to add one instruction means paying for all of it again. In legal the tradeoff is worse still.
4/5 Stop, too. Halt a workflow any time. It stops gracefully on request, and force-terminates within a minute if it is mid-operation.
5/5 The line steering will not cross: it can change what you ask for, never how rigorously it is produced. Push for a shallower review and the system holds full depth, with a mandatory quality gate before anything is delivered. That is what agentic architecture is for. A system you can direct, interrupt, and trust to push back.