Legal AI is coming for your admin - and your hiring plans

In this issue: Clueless, coding, prompting, and the (probable) three phases of generative AI in legal

When I was a trainee lawyer, I sat in a windowless room reading 30,000 pages of contracts to find the change of control clauses. Now you can get the output with an LLM in about 10 seconds. That’s a pretty big shift. If Clueless had been shot in 2024 rather than 1995, the most famous legal discovery scene in movie history would never have happened, thanks to AI:

Nowadays, you’d feed all those documents to an LLM, the August 28th files would have been analysed in a matter of seconds, and Josh and Cher would probably never get together. It doesn’t bear thinking about.

We’ve all read the stat from Goldman Sachs that 44% of legal tasks are capable of being automated by AI. How they came up with such a precise figure, I have no idea, but they are basically right and we’re seeing the impact every day.

At Juro we’ve been talking as much as possible to lawyers about how they actually use generative AI, right now (do give it a read - we love to see forward-thinking legal teams winning time back with technology). But where does this all lead?

Over the next few years, I think there will be three waves in this transformative AI era for legal:

Phase 1: Legal teams drive the AI

Some, but by no means all, lawyers are riding this wave. Legal teams - especially lean legal teams at mid-market or small companies - are solving the lack of resource/time by leaning on ChatGPT, CoPilot, legally-trained assistants or similar.?

The goal here is to solve repeat problems quickly. Laura Jeffords Greenberg shared a great example in the blog referenced earlier - she wanted to draft “a mutual non-disclosure agreement that we could give to the sales team so that they could run with it and then we would never see it again”. ChatGPT obliged when it came to the drafting:

... and voilá, that’s one repeated low-value task moved off her desk forever, and a business team enabled. For a scaling company with a lean legal team, that’s better than priceless - if we model everyone’s salaries, it has a distinct monetary value, which is scalable and helps with capital efficiency. Perfect ??

Phase 2: Legal sets the rules - the business uses AI to follow them

The second wave, at least from our point of view, is illustrated by my experience in building and using Juro’s AI Assistant. We aimed to go much further than ChatGPT or CoPilot and built an AI playbooks function, as well as a whole bunch of privacy guardrails. Because it lives within Juro’s editor, you also don’t need to switch to a different tool to use it.

And we use it every day. Looking at our internal processes: we’re a scaling B2B SaaS company that agrees a lot of sales contracts, and we have a lean legal team.?

Yes, Michael Haynes could spend every month-end running around the sales team hand-holding on negotiations. Instead he uses AI playbooks to set guardrails for the team; conditional logic to give them fallback clauses; and shortcuts to give them one-click solutions to common redlining problems, like “make this mutual” or “increase indemnity”. He’s scaled himself, without begging for more headcount.

The conclusion of this is that increasingly teams outside of legal (I hate the term ‘non-lawyer’) will conduct what were previously legal jobs-to-be-done. They will be able to negotiate their own contracts - yes really - but within the guardrails set by legal. Now everyone’s in the legal team ??

The problem this solves, and the solution it presents, could have consequences that might scare us. The idea that AI is coming for your job is another worn-out meme, but is it coming for your admin? Absolutely (and even the EU has moved quickly to make it easy for companies to test it)

And probably it is coming for your hiring plans. Like any diligent startup, we’ve been gathering case studies on our new AI features from our customers; and one of the common themes we hear, when we ask for the tangible benefits of AI Assistant, is that it means you can hire less, or later.

Smaller legal teams might not be the outcome that GCs want to achieve with their software budget. But for CFOs looking at the cost of hiring from private practice, it’s a great result. And for GCs ready to embrace it - ready to delegate big chunks of their work to junior team members or business colleagues driving LLMs - it unlocks some pretty cool outcomes.

Like Lucy Ashenhurst said: “Not only does automating these tasks save a vast amount of time, but it also helps to avoid decision fatigue - you don’t spend all your time on boring decisions that don’t matter. If you can automate non-critical, non-material decision-making, then you enable yourself to be better equipped for big decisions."

That kinda sounds like … you can do more with less. We did it!

Phase 3: AI solving legal problems at the point of requirement

So where do we go next if business teams are conducting legal jobs now? Well, probably the same old challenges will come for AI as they did for workflow software. Is it flexible? Is it safe? Most importantly, will anyone use it?

Yulia Andrusiv, in the legal team at 3Commas, said that having used AI Assistant, "the draft clause I get back is on-the-money. It’s faster and more convenient than Chat GPT because it analyses all the context from the contract itself, and we can stay within our contract platform without having to switch between different tools. It definitely helps us get to the outcome we want in negotiation faster.”

That’s cool, but what if you didn’t even have to switch to Juro? If legal AI is to be operated by folks outside of legal, then it needs to live in the tools that business teams already use.

We made a step in the right direction last week when we enabled customers to iFrame a contract view of Juro into Salesforce and use AI Assistant from within Salesforce ??:

What this means is that, as an Account Executive in the sales team (with minimal training) you can:

• Generate a pre-drafted contract directly in Salesforce
• Send it for approval / negotiation without leaving Salesforce
• Have AI rewrite clauses based on redlines, within legal’s guardrails, without leaving Salesforce
• Send it out for signature once you’re done ??

We see lawyers shifting from front-end users of AI to back-end users. They configure the playbooks, templates, and guardrails and then sit back to see contract negotiations on routine contracts be conducted independently of the legal team

In a different context, this is something Flank does amazingly well - serve people intelligent legal answers, right where they need them, in a way that doesn’t disrupt their regular working patterns at all.?

If this is where we’re headed - and of course, I could be wrong - then falling behind the curve here could be essentially fatal for lawyers. And for those who embrace it??

A future where legal teams are back-end engineers, rather than a front-end service desk, is a rosy one. The asset that’s never been more precious is your judgement, based on experience, the creativity you have in solving problems, and your storytelling ability.?

Thanks for reading this newsletter - somehow 1000+ people subscribed to the first edition. Do pass it on, we have a big week at Juro next week!?

And if you’re interested in the reality of legal AI, follow the likes of Andrew Cooke and Electra Japonas who are truly cutting through the noise.?