We still need the humans: Where AI contract analysis falls short and what it needs from us
I was more than a bit skeptical when a client called this week to walk me through an AI-generated analysis of a contract he was considering.
I'll start by saying I'm not anti-AI, and my skepticism had nothing to do with this client. In a year, I've gone from unimpressed to experimenting to using AI almost daily for specific tasks, with guardrails and human review.
But analyzing a contract is the kind of nuanced task I would expect AI tools to struggle with, and I'm not just saying that because I practiced law for almost 10 years and want to make sure my former colleagues still have work. Contract analysis needs to reflect how parties understand their relationship and how it will actually operate. That's where AI models often struggle, and where the analysis fell short here.
What the AI got right
To be fair, the AI tool got some things right. It understood the referral relationship context enough to recognize that the contract's definition of customer data was critical to the protections and limitations it spelled out. It also correctly recognized the third party's incentive to obtain customers' opt-in to further engagement directly from the third party — which would cut my client out of the loop on future referral fees. Those were good catches.
What AI missed
Once my client walked me through the customer referral journey, I could see what the AI analysis missed. Three blind spots stood out.
There are different ways to structure a referral handoff. Each impacts responsibility for disclosures and consents, control of customer data, the customer journey, and how the contract ultimately should be drafted. The AI analysis only considered one possibility.
The walk-through also raised questions about the customer journey that were missing entirely from the AI analysis. At what point would customers know they were moving from my client's platform to a third party? How would the different handoff options impact the customer experience with my client vs. the third party?
AI tools can't fully account for negotiating dynamics. What changes would the other party likely accept or reject? Could an ask jeopardize the relationship? What alternative asks would preserve the relationship while still protecting one or both parties and furthering their interests?
The referral handoff, customer journey, and negotiating dynamics need to inform the contract analysis, not just the other way around.
Where this leaves us
In this case, the walk-through and a more nuanced understanding of the relationship between the parties were critical context the AI model never had. Which means as far as the AI technology has come, we still need the right people to complete the picture.
What's your experience been using AI tools in high-stakes situations — where has it helped, and where has it fallen short?
(Obligatory ex-lawyer disclaimer: I no longer practice law, this post isn't legal advice, and anyone with legal questions should consult a practicing attorney. But you know that already.)
