Fluent in the Tool, Illiterate in the Tradeoff

Most conversations I have about AI these days have the same shape. A leadership team can describe, in impressive detail, everything the tool can produce. Then I ask who checks the output, how long that takes, and whether the time they saved actually survives the review — and the room goes quiet.

I've come to think of that silence as the real problem. Not the model, not the budget, not the roadmap. The gap sits one level up, in the thinking. And I don't have a cleaner way to name it than the way my guest on this episode did: most organizations are fluent in the tools and illiterate in the tradeoffs.

Frank Bravo has earned the right to say that. He's a Bay Area technologist who spent more than three decades leading technology through disruption — 32 of those years as CTO of a newspaper company, where his team published an entire paper's editorial content to the web in 1994, before Google, before Yahoo, before most people had a browser to read it in. He now runs technology at ProLegal, a legal-services startup, keeps a consulting practice on the side, and hosts his own show, Your Tech Makeover. What makes him worth listening to right now isn't the résumé. It's that he has already watched one industry get completely rewired by technology, so when he talks about this AI moment, he isn't guessing at what's coming. He's seen the movie before.

Being early was never the advantage

Here's the lesson from that first disruption that I can't stop thinking about. The media industry was often early. It rushed to adopt each new tool as it arrived. And being early, Frank pointed out, rarely turned into the lead everyone assumed it would.

That should sit uncomfortably with anyone treating AI adoption as a race. Speed to adopt is not the same as clarity about what you adopted. You can be first to a tool and still lose, because the advantage was never the tool — it was understanding the tradeoff well enough to use it in a way your competitors couldn't copy. Frank's whole career is a case study in the difference. He was at the front of the web in 1994 and it mattered, but not because he clicked "publish" before anyone else. It mattered because he understood what the shift actually did to how a newspaper worked.

That distinction — between owning a capability and merely owning the tool — is the thing I keep trying to get leaders to sit with. It's easy to buy the tool. It's hard to build the understanding that makes the tool yours.

He builds faster than ever. He reviews everything.

What I appreciated most was that Frank refuses the caricature. He isn't an AI skeptic slowing his team down out of fear. Just the opposite. As a developer, he's building software faster than he ever could before, because AI has genuinely collapsed the time it takes to get from idea to working tool.

And nothing goes live without a human reading it.

Those two facts are supposed to be in tension, and for a lot of teams they are: you're either fast or you're careful. Frank is both, and he's blunt about why. AI, in his words, is a tool that does some things really well and some things really badly — and the job is knowing which is which. In his world, that isn't an abstract concern. He's the CTO of a legal-services company, where getting AI wrong doesn't just mean an embarrassing output; it can get a lawyer sanctioned. When the cost of a confident, wrong answer is that high, "who checks it" stops being a footnote and becomes the whole design question.

He was honest, too, about the limits of leading with the flashy version. ProLegal launched an AI product aimed at law firms back at the start of the year, and what they found was that firms often already had tools in place, or that other services were further ahead. So the company kept the AI where it delivered real, quiet value: internally. When someone is in an accident and has a judgment coming but needs cash now, ProLegal can advance money against that future payment. Deciding how much to advance means reading a pile of case documents and forming a judgment about how strong the case is. That's where their AI earns its keep — boiling those documents down into something a human can decide on. Not forward-facing. Not a demo. Just useful.

His diagnosis of where companies go wrong is one line I'd frame on a wall: businesses treat AI as the end-all, let it run rampant, and end up with a pile of bad software that does something simple in a needlessly complicated way. The failure isn't too little AI. It's too little thinking wrapped around it.

A portfolio that holds itself up

There's a second throughline in Frank's story that regular listeners will recognize as close to my own obsession: he never bet everything on one thing. He runs three lanes at once — his role as a startup CTO, his consulting practice, and the podcast — and they feed each other. The consulting lets him do work his day job can't. The podcast grew out of the consulting during COVID, when he couldn't sit with clients in person and figured the questions he kept answering one-on-one deserved a bigger room. Each piece makes the others stronger, and none of them is load-bearing alone.

That's not hustle for its own sake. It's the same instinct that makes ProLegal itself interesting. They aren't trying to sell a single feature; they're building a suite that law firms can plug into — the kind of thing others can stand on. The newest piece, ProLegal Talent, is a concierge hiring service for firms, and the design choice tells you everything about where they compete. Anyone can hand a law firm a hundred résumés. ProLegal hands them two people it's confident will fit. Fewer, better, matched by the technology in the background. That's a niche you can defend, because it's built on judgment, not volume.

And they can move. That product went from a hallway conversation to a live launch in under three months. Which brings me to the part of the conversation I found genuinely instructive as an operator.

Autonomy is what makes a small team dangerous

ProLegal is fifteen people. Frank has one full-time report and a couple of contractors. On paper, that's a rounding error next to the institutions he used to work inside. In practice, it ships faster than companies ten times its size, and the reason is culture, not headcount.

When the CEO hired him, he said something Frank clearly hasn't forgotten: think of ProLegal as your company. Build it into what you think it should be, and don't let anyone tell you that you can't. After decades inside a world where every decision moved through committees and working groups, that kind of autonomy was new — and it's the engine behind the speed. Everyone on the team has the standing to walk in with an idea. No idea gets dismissed out of hand; the good ones get built fast, precisely because there's someone inside championing them. ProLegal Talent existed because two people had a conversation and were allowed to run with it.

That's the quiet lesson for anyone envious of startup velocity. The speed doesn't come from working people harder. It comes from giving them enough ownership that a good idea has a short path from someone's head to something real.

The unglamorous frontier

Ask Frank what he'd build with unlimited resources and he doesn't reach for anything sci-fi. He points at the least glamorous problem in the room: law firms are excellent at law and disorganized about their own information. The dream tool isn't a robot lawyer — it's something that could actually take a firm's scattered files, systems, and knowledge and run it, whether the firm is two people or two hundred. Tools that gesture at this exist, in his view; none of them do it well. That's a green field.

He said the same thing about the smallest client on his consulting roster — a solopreneur drowning in files spread across iCloud and OneDrive who simply can't find anything, because nobody built her a way to keep it all in one place. Different scale, identical problem. It's the same instinct that drives his podcast: help people understand the technology in front of them well enough to make a good decision with it.

Which is where he'd want me to land this, and where I want to land it too. He's a committed open-source believer, not because free is cheaper but because he cares about standards — data you can actually move from one system to another, instead of being quietly locked in. And he's insistent that education isn't a nice extra bolted onto a product. If people can't understand the thing, the thing has failed, no matter how clever it is under the hood.

That's the whole argument, really. The hardest part of any technology shift was never the technology. It's getting people to understand it well enough to choose wisely. Every organization I've watched struggle with AI has been fluent in the tools and illiterate in the tradeoffs.

Clarity isn't a nice-to-have on top of strategy. It is the strategy.

Listen to the full conversation with Frank Bravo on What Comes Next with Arun*, wherever you get your podcasts. If it shifts how you think about AI in your organization, share it with one person who needs it — and tell me the one tradeoff your team keeps skipping.*