In this week’s Law Punx episode we hear from Richard Mabey, CEO of Juro, on the key subject of AI and the automation of legal work. In this blast, Richard argues that by the end of 2027 around 80% of legal tasks could be automated. And that’s a radical position….!
Without further ado, here is the pod. Press Play to watch / listen, or go directly to the AL TV Channel.
This episode and four other Law Punx podcasts are on Spotify, which also features Horace Wu, Jake Jones, Todd Smithline, and Electra Japonas
The Spotify Law Punx site with all episodes is here – where you can listen/watch all of the episodes in one place.
And the same, but just an audio version, is available on Apple Podcasts.
Several more Law Punx blasts have been recorded are now in the studio, including Oz Benamram, Tara Waters, and Jerry Levine, among others. Drop AL a line if you’ve got something to say!
There is also an AI transcript below.
—
Legal Innovators Conferences in London and New York – 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 and also, Legal Innovators UK – Nov 4 + 5 + 6, where the brightest minds will be sharing their insights on where we are now and where we are heading.
Legal Innovators UK arrives first, with: Law Firm Day on Nov 4th, then Inhouse Day, on the 5th, and then our new Litigation Day on the 6th.
Both events, as always, are organised by the Cosmonauts team!
Please get in contact with them if you’d like to take part.
—
AI Transcript
Hi everybody, Richard Tromans here at Law Punx and not Artificial Lawyer TV. Our new guest is Richard Mabey, the CEO and co-founder of Juro, a company that you probably know well. He is going to propose a radical idea. Richard, what is your idea?
Richard (00:32.878)
So I want to argue that 80% of legal work, 80% of legal tasks, by the end of 2027 will be executed without lawyers.
Richard Tromans (00:43.276)
And just to clarify, when you say legal tasks, you mean like discrete things like summarize this, check this. We’re not talking 80 % of all lawyers being replaced by AI. We’re talking about specific tasks.
Richard (00:57.024)
Specific tasks, yeah, I go as bold as to say that the tasks will be automated, but I wouldn’t predict what that will mean ultimately for legal headcount.
Richard Tromans (01:06.874)
Alright, okay, you have the floor, take it away. Make your point.
Richard (01:10.862)
Awesome. So I’m going to argue today that by the end of 2027, 80 % of legal work will be done without lawyers. And when we started Juro, which was some time back in the dark ages of SaaS before AI was really taking hold, the vision that we had was that we would make executing legal tasks accessible to teams and businesses that were outside of the legal team.
And that was back then quite a controversial opinion. So the idea would be that if you give sales teams the right parameters and guardrails, that they would be able to take on the creation of documents, the negotiation of documents and so on and so forth. And that opinion, you we fast forward now to 2025, where AI is unlocking still all of
or the potential to do more and more complex work is still a relatively controversial opinion. And most thought leadership and content says that actually AI is going to net benefit lawyers and we’re going to still do this, do the stuff, but we’re going to do it faster and more efficiently. So.
Going back in time, when I worked at Freshfields, I was in &A, I worked in corporate law, and one of the observations I had was that even though that law firm work was complex and engaging, A, the vast majority of tasks that were done by lawyers there were not legal.
Richard Tromans (02:34.62)
Mm-hmm.
Richard (02:35.406)
they didn’t have much to do with the law. So a lot of project management, like really complex sort of drafting of, you know, waterfall clauses in SPAs, but actually very little to do with the law. And the second thing is that actually that for those tasks that had something to do with the law,
So not particularly complex, after the basic contract law, you didn’t really necessarily need to go to law school for multiple years to kind of figure out. was ultimately a test of where you smarten it and the problem solver.
And that was the main reason why we kind of founded Juro and why we wanted to get folks who are not lawyers to go and take on some of this work. And initially, if you look back in time to 2018 to 2021, the kinds of tasks that software providers could automate were fairly basic. So being able to fill in the blanks in the document, being able to have a clause library with fallback positions. And yes, it helped. And some marginal tasks were taken away from the lawyers, but actually not that much.
And then of course, with the rise of Gen.ai, in particularly in use cases for us like contract review, you’re taking on increasingly complex work. you know, initially in that co-pilot phase, you would say, well, keep on redlining the documents in Word. We have this plugin, it’s going to help you do that faster. Still see a lot of usage for that. But actually at the bleeding edge, you see a lot of legal teams now deploying Juro agents, which will autonomously redline documents within Slack and Teams.
And so we’ve become really interested in this question of, know, where does this go and what is the natural conclusion of that wave? And I think we’ve got the first signs now in early adopters that work that we would never have said in 2018 would ever fall into the hands of the business is. So a specific example might be a finance team uploading a supplier agreement to Slack and having an agent redline it.
Richard (04:30.802)
was unthinkable, but that’s already happening in those sectors. And we see that’s a bit of a radical shift. of course, right now, you see that
agentsic work going into simple tasks, NDA markups is the classic one, but we believe that that will go into increasingly complex contracts. And at the same time, what we’re seeing is that the models are getting smarter. So yes, GPT-5 has a lot of problems, but at the end of the day, it is actually spitting out better accuracy and evals on legal work. And we don’t think GPT-5 is the ceiling, we think it’s the floor.
So all of this stuff is kind of ongoing. And we’re kind of saying to ourselves, well, if AI and agent can execute to the level of, say, a junior associate, plus actually in some respects has better context if you give it the memory of contracts or documents or past negotiations, it must be only a matter of time before the intelligence level actually outpaces humans. And if it does,
then you ask within a business, well, how will that legal work get done? And we think it’s basically going to be deploying agents which the business self serves on rather than being legally supervised every time. So that’s the context within kind of contract automation. But what’s actually interesting is you take that outside of the stuff that we normally talk about. So like think about consumer law, for example.
So I love this new product Garfield, which I think you featured Richard on on your show. And again, you know, not much happening in consumer versus kind of business to business.
Richard (06:03.906)
But actually you see already consumers taking law into their own hands, literally. So, you we saw this in the past with do not pay and Garfield, you see simple use cases where folks are saying, either I can’t afford a lawyer or I don’t want to spend a thousand bucks on a lawyer and therefore I can self serve. Again, imagine what that means today, but more importantly, where that’s gonna go.
And in disputes, for example, it’s the same thing, right? So will we get to a case where you can simply submit your own pleadings? You can have an AI agent who can ask you intelligent questions, crunch that information and start to make mediation type judgments.
Richard Tromans (06:26.648)
Interesting.
Richard (06:44.29)
And so a lot of people will say to that opinion, well, this all sounds kind of a little bit far fetched and actually we’re very worried about regulation. We’re worried about privacy and we’re worried about having tasks that aren’t overseen by lawyers. And if this is true, and I’m not sure that that is a valid concern, but if it is a valid concern, the argument that I have is that it is only a matter of time before it isn’t. Models get smarter.
context gets richer, legal intelligence rises, and therefore the universe of legal tasks, which really require human judgment, is gonna shrink. And our estimate, just a gut-feel estimate, but our estimate is by the end of 2027, that would be the 20 % of really high value tasks, and the 80 % would be commoditized and automated. And…
the kind of governmental aspect to that is also kind of stark, right? So if you look at the way regulation works for lawyers, regulation is built entirely on the premise that only humans can deliver advice. So legal advice, protected activity. We can’t have people do it themselves. They’re not smart enough. They’re not trained enough. You need trained lawyers.
Richard Tromans (07:37.244)
Thanks.
Richard Tromans (07:55.514)
And ironically the entire basis of legal regulation in all countries is also based on the idea that lawyers will make mistakes.
Richard (08:03.51)
Right, right.
Richard Tromans (08:05.084)
which is why they regulated because it’s tough job and even the people who are good at it will make mistakes.
Richard (08:13.42)
Yeah, which is really sad. And that’s right. mean, until, say, a year ago, you would say, well, yes, mean, their law is clearly outperform machine. So actually, that’s the best we’ve got. What I would argue is, if it is the case that that sort of level of super intelligence is going to rise, then actually, you know, there’s also kind of policy and governmental questions of how do you regulate that? Because it’s not the same, right?
Richard Tromans (08:36.892)
Well, let’s say that for another law, punks, on how to regulate the future. I’ve just got, because we’re running out of time, because I want to try and keep this punchy and short. I’ve just got one key question, which I’ve related to, know, through my article a few times, is what about the accuracy aspect? You know, I totally agree with you on the broad spread of this. Can generative AI get a long way into many, many, many, many, many tasks? The critical thing for me is…
What about the accuracy? Okay, so an agent does X, Y, and Z for you, pings you a document, et cetera, et But if it’s not totally, totally accurate, if you’re not 100 % happy to trust it, have you really automated it? You know what I mean? It’s like, is a self-driving car that you have to stop every 20 minutes really a self-driving car?
Richard (09:28.43)
Well, I think it’s a question of scale, right? So I mean, on one end of that spectrum, you have the heavy lawyer in the loop thing, right, which is something like a first pass review of a contract.
Yes, it saves you time because the issue is spotting, but you are actually reading every single word of that document. So to your kind of self-driving car and actually it’s not self-driving, it’s just driving with a sort of, you one of those inbuilt autopilots that keeps you in lane or whatever it is. And I think, you know, on that scale, like what we’re starting to see, and I think this is the early stage of it, but we’re starting to see people say, well, I want massive leverage. So the question is, to what degree are you in the loop? So for example, let’s say you have an AI
Richard Tromans (09:50.98)
Okay.
Richard (10:08.024)
agent which is deployed in Slack and you as a lawyer are defining the playbook, you’re taking exceptions, you’re fine-tuning, it might be that you could have a million contracts that just choose as crazy example a million contracts that are being agreed with that playbook that you have honed and refined and set the guardrails of.
And that’s very different from jumping in each time. So I think that the direction of travel will be lawyers gradually stepping out of the loop, still in a supervisory role, but that’s very different. That’s like being the engineer of a wave self-driving car, as opposed to being the of the autopilot that you need to turn off and on.
Richard Tromans (10:44.856)
Right, so just to wrap up, so you really believe that we’re going to get to a point quite rapidly, very rapidly, where the lawyer really does just take themselves out of the loop in multiple areas.
Richard (10:58.274)
I think so. And it’s not that we won’t need lawyers like me, but we will. But we’ll be doing, we think, 80 % fewer tasks.
—
Discover more from Artificial Lawyer
Subscribe to get the latest posts sent to your email.