The Faster Horse Problem in Legal Tech

If you've spent any time around legal innovation panels, you've heard the line.

Henry Ford supposedly said that if he'd asked customers what they wanted, they would have said "a faster horse."

Clients hesitate around GenAI pricing — they question the cost, worry about their data, want the benefits without paying for what produces them. The easy response from firms: clients don't value what we're building.

Maybe.

But here is the flip side: a lot of legal AI work still centers around the horse.

Faster contract review. Cleaner drafting. Better search. Smoother reporting. More predictable delivery. Less friction in the same familiar places.

Useful? Absolutely.

A car? Not yet.

That is the faster horse problem in legal tech: it is not that firms are improving the horse, but that they sometimes mistake those improvements for the car.

And that is where the client reaction becomes more interesting. If the work still looks like an improved version of what they already buy, it is not surprising that clients ask who should capture the value.

The cliché is harder to use on clients when the offer still resembles the thing the cliché describes.

So what would an actual car look like?

I want to be upfront: I'm not an attorney. What follows is not a roadmap or a set of polished product ideas. Think of it more as a set of primers, deliberately rough, meant to provoke rather than prescribe. The value is not in whether these specific ideas are right. It is in whether they get a lawyer to say, "No, but here's what that would actually look like" — because that reaction is worth more than these ideas ever will be.

Catching the dispute before it becomes a dispute

AI arbitration tools already exist. They can help resolve disputes faster once they happen. Useful, but still close to the horse: the dispute exists, and now the work is to process it more efficiently.

The more interesting question is what happens before that.

Imagine a system that doesn't wait for a contract to be breached, but continuously watches how the relationship is behaving against what was agreed — a supply arrangement, a partnership, a licensing deal — and flags the moment conduct starts drifting from the terms, while the issue is still small enough to address without becoming a formal dispute.

Not: you have been sued. More like: this relationship is beginning to move outside the agreed boundaries. Here is where. Deal with it now.

That's a different question than "how do we resolve disputes faster?" It asks whether some disputes could be prevented from becoming disputes at all. I have no idea what this would require legally or technically — that's the point. The useful reaction isn't "interesting idea." It's someone saying, "No, that wouldn't work like that — but here's what might."

Contracts that are not frozen in time

A contract is, in a sense, a guess about the future written in language that can't update itself. That's not a criticism — it's part of what makes contracts useful. They fix obligations, create certainty, give parties something to rely on when circumstances change. But they're also static instruments in dynamic relationships, which is why force majeure clauses, change-control provisions, and renegotiation rights exist in the first place.

So the car question isn't "how can AI read or draft this contract faster?" It's: what would the agreement look like if it were built for a relationship that keeps changing? Maybe a contract that continuously monitors performance data. Maybe provisions that escalate or require human review when defined conditions shift. Maybe the agreement becomes less like a document and more like an operating system for the relationship.

I'm not pretending to know how you'd govern that, or how a court would think about it. But it hopefully is a more interesting question than how to draft contracts more efficiently.

Precision, but only when it's earned

A related question, and I'll admit upfront I don't fully know where it leads. From the outside, one striking thing about contract drafting is how much precision gets spent on futures that may never arrive — language negotiated for scenarios that may never happen, because once something is signed, ambiguity becomes expensive.

What if some precision could be deferred until reality made it necessary? Not sloppier drafting — something more deliberate: an agreement that knows which of its own terms are intentionally unresolved, and why. The moment one of those areas becomes relevant, the system flags it and pulls the right people in, applying expensive precision to the thing that actually turned out to matter — grounded in facts, not a guess made years earlier across a negotiating table.

Maybe that creates enforceability problems. Maybe it just recreates the same disputes in a different place. But it's not a question you get by asking how to draft faster — it's the question you get by asking where and why precision creates value.

The one that should make firms uncomfortable

The most provocative car may not require speculative technology at all. It requires a firm to change what it's willing to stand behind.

Right now, most firms sell effort, expertise, or a defined project. What if a firm sold certainty instead? Not "we drafted your compliance policy," but "we built and continuously run the system that watches your business for this category of risk — and we are financially exposed if something it should have caught slips through."

That's not a tooling upgrade. It's a different relationship with the client, the work, and the risk — and it's the kind of thing clients might plausibly pay much more for, because it's closer to what they actually want. Not more legal activity. Fewer unpleasant surprises, caught earlier, by someone with enough skin in the game to care.

This is where the faster horse problem becomes commercial, not technological. A firm can use GenAI to do the same work in fewer hours and argue over who captures the efficiency. Or it can use GenAI to support a different promise altogether — one that touches pricing, insurance, professional responsibility, and the firm's appetite for being wrong. Which is exactly why it looks less like a better horse.

And if you want to keep stretching

Agents negotiating on behalf of each party, with humans stepping in only when they get stuck. Regulations written as machine-checkable logic, so compliance is tested continuously instead of interpreted after the fact. Some of this may be unrealistic, or impossible for reasons a lawyer could explain in five minutes — that doesn't make the exercise useless. The point is asking what the client is actually trying to accomplish, before the familiar legal service appears in the middle.

Comfortable innovation is not transformation

Here's the part that makes all of this hard. The obstacle isn't only technology, client hesitation, or budget — those are real, but they're often the visible edge of a deeper one: most firms are willing to rethink the work before they're willing to rethink what they stand for and what they sell.

The current model has pricing structures, staffing patterns, and liability boundaries built around it. Improving it is easier to approve, govern, and explain than questioning it — and often, that's the right call. But it isn't transformation.

There's a moment in Rise Above AI Chaos where a senior partner, reflecting near the end, recognizes that she'd wanted transformation without the discomfort transformation would require. The line lands because it isn't really about technology — it's about the very human desire to change the outcome without changing the posture.

Firms want the language of transformation. Clients want the benefit of innovation. Everyone wants the upside. But the car asks a harder question: what are you prepared to change about how value is created, priced, delivered, and guaranteed?

If the answer is "not much," that's fine. Just be honest that you're still working on improving the horse.

The actual point

None of this is a pitch for any of these specific ideas. It's a pitch for the exercise.

Most legal AI conversations begin with the thing the firm already does — drafting, reviewing, researching, reporting — and ask how GenAI can make that thing faster, cheaper, or more scalable. Useful questions. Horse questions.

The car question starts somewhere else: what is the client trying to avoid, achieve, or stop worrying about — and could the firm deliver that more directly than it does today?

I can't tell you what your car looks like. But I'd bet you already know where the horse is getting tired.