&AI, a new legal tech startup focused on patent workflows and AI agents, has raised $6.5m in a seed round, with Y Combinator backing the San Francisco-based company. Artificial Lawyer spoke to co-founder Caleb Harris about it all.
First, what is &AI? It’s ‘a secure workspace engineered for patent attorneys, leveraging AI to accelerate end-to-end prosecution and litigation workflows’. It handles prior art searches, claim chart drafting, and much more besides, (see table below).
Central to this is ‘Andy’, the company’s AI agent, which ‘collaborates with attorneys’.
The funding, led by First Round and supported by Y Combinator, as noted, plus SV Angel, BoxGroup and angels Kulveer Taggar, JJ Fliegelman and Rich Aberman, will be used to ‘enhance Andy’s functionality, broaden integrations and grow the team of experts in AI and patent law’.
Harris and co-founder Herbert Turner both attended MIT and became friends. Harris later worked as a technical advisor to law firms on patent issues, where he saw first-hand the challenges of manual workflows for patent due diligence, and decided to build a solution.
Harris commented: ‘&AI allows patent teams to replace manual workflows with AI so they can do due diligence in a fraction of the time and with much higher accuracy. In the process, we free up valuable attorney time for less mundane work.’
So far it handles:
- ‘Invalidity – Understand the complete picture of patent validity in minutes with global prior art search and litigation-ready claim charts.
- Preparation + Prosecution – Save time for the complex work. Quickly check the novelty of invention disclosures, draft robust claims, and automate responses to office actions.
- Infringement – Pinpoint new opportunities. Obtain detailed insights into how products overlap with your patent or portfolio and assess monetization potential.
- Portfolio management – Scale your analysis without limits. Evaluate entire portfolios against prior art, standards, or product specifications of interest.
- Workspace – Securely manage siloed projects. Ensure seamless collaboration while maintaining the highest level of security for sensitive material.’
Agent Andy + More
AL then asked Harris some more about Andy. Harris explained that in simple terms ‘it does tasks on your behalf. It can both forward plan a set of tasks and start executing those tasks. And it’s specifically designed for patent workflows, such as litigation.’
Harris stressed that they’re not a generalist genAI tool, or even a general legal AI tool. Instead they’ve really analysed very specific workflows that are part of what is itself already a niche subject, i.e. certain aspects of a patent lawyer’s work, and then started from there.
(And interestingly, this is the ‘flow chart’ approach taken by another new AI company, SmartEsq, re. fund formation – i.e. really understanding the intricacies of complex workflows within the activity that very niche lawyers do, and then building specific AI tools for those very precise needs.)
So, to agents. There’s a growing number of definitions, (e.g. OpenAI often talks about screen control agents for making bookings on websites, AKA operating in ‘the pixel space’, as opposed to running tasks set within a series of legal documents and a collection of emails). What are the agents in this case?
Harris says that agents for &AI are basically something that can aggregate steps, with a planner and an executor function. I.e. you break a project into a set of tasks, attempt to do each step, but….and it’s a key point…..the agent can ‘dynamically update the plan’, and may even need to go back to the start to get it right.
And what about accuracy? Harris explains that there is a very low risk of hallucination errors as the agent is designed to carry out very specific tasks that mimic those of a human. I.e. if you really can keep very tightly to those human lawyer actions then you’re on safe ground.
The company also adds: ‘Claim charts, search results, and other work products made on the &AI platform are only composed of ground-truth information, not generated content. This eliminates the possibility of hallucination.’
In terms of the bigger picture, Harris notes that ‘the state of the art for LLMs changes every three weeks’, and it’s clear they are moving along with this frenetic evolution. He’s also impressed with what general foundation models can do now – and that will likely only get better. In turn, that should make what they can do in the patent space even more impactful.
Overall, this looks like a very promising start – and it’s got some big-hitter backers. And as Harris concludes, what they’re doing now with AI wasn’t even possible a couple of years ago. Who knows where &AI will be in another two years?
Congrats to them on the launch and funding.
—
P.S. On the subject of legal AI and California – check out the Legal Innovators California conference, SF, June 11 and 12.
If you found the above of interest, then come along to the Legal Innovators California conference in San Francisco, June 11 and 12, where over two days – one for law firms, and one for inhouse – we’ll be exploring the cutting edge of legal tech with some of the top lawyers and technologists in the sector.
More information here about the two-day conference, which is at the CJM, 736 Mission Street, SF, on June 11 and 12.
For information on how to sponsor, please contact: robins@cosmonauts.biz
And for information on speaking opportunities (if you are at a law firm, or working inhouse), please contact: phoebe@cosmonauts.biz