AI is already super useful in the legal world: including for guidance and advice, as well as preparing/ reviewing and helping negotiate contracts.
Does it get everything right? No. Even if you get the prompts right, there are often gaps or weaknesses. We know, because we’re actively using and testing legal AI tools.
So what should you use AI for when it comes to legal work, contracts, etc?
The answer is – anything, provided you do a couple of things first:
- A basic risk assessment review. What’s the potential cost if the tool gets something wrong? Does it justify being checked by someone who knows what they’re doing and will stand behind their advice (and has professional indemnity insurance).
- Read: Have you checked how your team is using AI? (and why you should care)
- Be very careful if you’re involved in a dispute that could turn into litigation. There’s some doubt over whether legal privilege exists/ is lost if you use AI tools.
The main problem with using AI advice/ docs is similar to if you use cheap templates/ platforms/ Google – i.e. if something goes wrong, and it doesn’t work as expected, you have absolutely no come back. Because there’s no-one to blame, no-one to stand behind the advice and documents that knows what they’re doing and can double check that they work as you need them to.
Here’s what the tools and platforms say:
- “I can’t give legal advice.”
- “I can help you create a draft contract, but a qualified lawyer should review it if the stakes are high.”
- “You should obtain professional advice.”
- “This is for general information only.”
We understand, your appetite for risk is usually/ needs to be high when you’re a start-up. But as your business grows, your approach to risk needs to shift. Which means key contracts/ docs need to protect you. Because at the end of the day they are an insurance policy, which you’ll need to rely on if something goes wrong. So use tools/ templates for short term, low cost deals/ contracts, low risk guidance, etc. NOT for key deals/ contracts, managing key IP, your cap table, etc. Because not only might they fail when you really need them, but if you’re fundraising or preparing for an exit, you’ll get a red flag in the due diligence exercise, which could slow the deal progress, impact price/ value, or even kill it.
We’ve been actively trialling a range of AI tools internally at LegalEdge – and the results have been mixed. The outputs often look impressive on the surface, but when you dig deeper, there are some clear limitations and some real concerns. We’ve found that AI can miss key issues or protections, or even produce drafts that don’t work in practice – which without human oversight could lead to serious risk. And sometimes it gets confused about the perspective it’s meant to take; for example, when asked to review a contract from a customer’s viewpoint, it might make further pro-supplier changes instead. It also tends to ignore commercial realities, suggesting terms that are simply unworkable (like expecting a local council to agree to multiple uncapped indemnities for minor things that are simply not needed or unworkable). AI tools work from general patterns rather than understanding the nuance or strategic position of a particular business or deal, so a detailed human review from someone who understands/can spot the issues is critical. And while humans naturally build negotiation strategies (“let’s start at A but fall back to B”), AI tends to take a black-and-white approach, losing that tactical edge. Where we have seen it often work well is in summarising large data sets – if you give it very specific prompts/instructions. These are general trends we’ve seen across different technologies and platforms (not issues with any one provider in particular) but they’re worth keeping in mind when deciding how and where to use AI in your legal and other processes.
If you’re not convinced your contracts/ legal docs/ processes are doing what they should please get in touch. We won’t over-lawyer or reinvent the wheel – we just want to help you to recognise risk and deal with it appropriately.
