Vespas vs Ferraris: What to Build in this Age of Uncertainty
We dedicate a small portion of our time to Syntheia Services, where we work with a handful of law firms to build custom software. For the last few years, the qualifying test for whether we would take on a project was simple: build if the customer needs a Vespa, not if they want a Ferrari.
We wanted to strike a balance for adoption — build too far ahead, and people resist the change; build too conservatively, and the technology is obsolete by the time it ships. Stay just one or two steps ahead of the market. That was where humans felt safe, and the technology still felt progressive.
Except now, with AI's capability continuing to accelerate, we have formed a different view: building Vespas doesn't get our customers far enough, or fast enough.
The bottleneck was always human
The natural question is then, "what is worth building?"
Before we try to answer that, three constraints we learned over the years:
People can't always see the target, even when we point at it. Tell a firm "this workflow is perfect for technology enhancement" and the answer isn't always yes. Readiness isn't blocked by the technology. Readiness is emotional.
Capacity isn't always easy to reallocate. When a process improvement turns a four-hour task into a one-hour task, the people doing that task don't automatically know what to do with the freed-up three hours.
Neither the data nor the plumbing is ready. Many agentic, autonomous workflows already work today, but they need structured data behind them. Most firms don't have their data ready, and even the ones that do are often still missing the plumbing.
Most lawyers were only ready to ride a Vespa. They would have crashed if they were given a Ferrari.
Technology was never the problem. The difference between then and now is that when the technology moved slowly, getting those three human constraints wrong was not excessively costly mistakes; but now, with AI moving so fast, those mistakes feel much more expensive.
What to build instead
With the rate of AI progress, Vespas feel outdated by the time they are released into production. If the frontier moves fast enough, there is no reason to build most things. Whatever gets built becomes obsolete by the next model release. Today’s evidence certainly seems to support that AI is progressing faster than traditional engineering efforts — but, we still have to place a bet on the future — what is left to be built?
As many in the tech industry have said — building is no longer the hard part, judgment is. Deciding what's worth building is the scarce resource now (though some would say it always was).
For legal services, the right people to decide what to build are the lawyers speaking directly to their clients, because they are the ones who actually understand the problem.
(Note: there is possibly one general exception: technology that captures a firm's own expertise and structures its data. So far, it appears the value of that data appreciates as AI models improve. It appears that data is the one bet that will probably pay off, regardless of whether the frontier continues to race ahead or stalls.)
Place your bets, people
For many, the worst case scenario is that the simple chatbot from OpenAI or Anthropic can get a firm the same measured efficiency gain as a six-figure custom software project. In that case, what was the point of paying so much money in the first place?
We believe the six-figure custom project is an investment to answer the question: given where technology is all going, where can our law firm differentiate from everyone else? In a market where the commoditized layer is being priced to zero, whoever differentiates is the one who will be left standing. Racing your peers and surviving the technology shift are the same race.
So, each technology project is betting on the basis of one of these assumptions:
If you believe AI capability will plateau for the problem you want to solve, then you should build ahead of where you are forecasting the frontier will stall. What is left for lawyers is whatever sits above that ceiling, and your technology will push you ahead of your peers.
If you believe AI capability will keep growing quickly, but not without limit, then you should get embedded with your clients instead. Bet that AI leaves behind a human-shaped gap — the trust, the accountability, the counsel — and race against the clock to earn the gap more quickly than your peers.
If you believe AI capability is headed toward no limit at all, then build to capture your own expertise and data, because there is nothing else left. Your only remaining value is to improve the inputs into the model, and try to go deeper than others so your AI generates a better answer than the raw models.
Refusing to pick one of these three is not the safe, neutral option. The non-choice is the fourth bet — a wager on the plateau case, except this does not even give your firm a differentiating technology.
The worst case scenario is not losing the dollars you have invested in building custom software, but losing the existential bet on the survival of your law firm.
