James Deller explains why machine learning has quietly become a basic requirement for running a serious company, and why most founders still treat it like a bonus feature instead of infrastructure.
James Deller has spent enough time inside machine learning products to notice something most founders miss: the technology stopped being impressive a while ago. Nobody claps for a company that uses electricity. Machine learning is heading the same direction, and a lot of businesses haven’t updated their expectations to match.
“The minute something works well enough to be boring, it’s already become mandatory,” James Deller says. “That’s exactly where machine learning is right now, and most companies are still treating it like a party trick.”
From Marketing Line to Load-Bearing Infrastructure
James Deller built his company, 1Touch, around a matching engine that paired vendors with influencers — a task that had to survive real privacy limits and real money on the line. There was no room for the system to be a slogan. If the matching got something wrong, an actual business relationship took the hit, not a slide in a pitch deck.
That kind of pressure teaches a founder the gap between decorating a product with “AI-powered” language and actually depending on the thing to work. James Deller puts it plainly: a feature you announce and a system you’re judged on are not the same animal, even if they use the same words to describe them.
His read on where this leaves the market is blunt. Companies using machine learning as plumbing, not as a press release, are the ones that will still look sharp in a few years. Everyone else will look like they missed a turn nobody warned them about.
“A feature you announce and a system you get judged on are two different things wearing the same name.”
The Bar Moved and Nobody Sent a Memo
According to James Deller, most founders outside of hardcore technical companies haven’t clocked how far things have shifted. Personalization, churn forecasting, pattern spotting — five years ago these were genuine edges. Now they’re just what a functioning business does.
He frames the stakes simply: if a competitor is flagging churn before it happens and another team is still waiting on a quarterly report to find out after the fact, that second team isn’t behind on innovation. They’re behind on the basics of running the company.
James Deller believes boards are asleep at the wheel on this specific point. “Boards grill founders about runway every single meeting,” he says. “They should be asking about model adoption with the same seriousness. It’s not a nice-to-have question anymore — it’s a survival question.”
“You’re not behind on innovation. You’re behind on operations — and those are much harder to fix in a hurry.”
It’s Rarely a Model Problem. It’s a Data Discipline Problem.
Every founder James Deller talks to about stalled machine learning adoption assumes the holdup is technical — not enough engineering talent, not enough model sophistication. In his experience advising companies through exactly this wall, the real blocker sits further upstream. The data is messy, unstructured, or nobody trusts it enough to act on it.
You can’t drop a sharp model into a business that doesn’t already run on evidence instead of opinion. The model doesn’t fix bad inputs — it just repeats them faster and with more confidence.
The founders who actually get value out of machine learning, in James Deller’s observation, are the ones who treated clean, trusted data as a precondition for the work, not a side project to patch up later while the model runs anyway.
“Feed a smart model a messy business, and you don’t get intelligence — you get confident nonsense.”
The Model Doesn’t Replace Judgment. It Relocates It.
James Deller is careful to push back on a common misread of the technology: machine learning doesn’t remove the need for human judgment. It just moves judgment to a different spot in the process.
At 1Touch, he says the hardest calls were never about which algorithm to reach for. They were about what the model should actually be optimizing for, which tradeoffs were acceptable, and — this is the part people underrate — recognizing when a confident-sounding output was quietly wrong in a way only a person would catch.
Founders who treat the model as a stand-in for their own judgment, rather than a tool that still needs judgment wrapped around it, tend to get burned by their own confidence long before the market gets a chance to burn them.
“The model doesn’t remove judgment from the room. It just changes where judgment has to show up.”
Privacy Isn’t a Speed Bump — It’s Part of the Blueprint
Building a matching system that respected user privacy while still doing its job taught James Deller something counterintuitive: privacy and performance aren’t actually fighting each other the way people assume.
The constraint, in his experience, produces better design. Building with privacy baked in from day one forces more honest feature engineering and a harder look at what data is actually needed versus what’s being collected out of habit or laziness.
His warning to founders is direct: treat privacy as a legal checkbox and you’re building something that will need a full rebuild the moment regulators catch up — and they always catch up.
“Privacy isn’t the fence around good machine learning. It’s part of the blueprint for building it right the first time.”
James Deller’s Advice to a Founder Starting Today
If machine learning isn’t already baked into your core operating assumptions, James Deller’s view is that you’re not being careful — you’re accepting a structural disadvantage against competitors who already made the shift. That doesn’t mean bolting AI onto every screen for a press release nobody asked for.
It means going function by function and asking an honest question: where should pattern recognition at scale be doing work that a person or a spreadsheet is currently doing slower and less reliably?
“Machine learning stopped being the interesting part of the pitch a long time ago,” James Deller says. “It’s just what running a serious company looks like now. The founders who get that distinction are the ones building companies that will still matter once the novelty is completely gone.”
“The novelty already wore off. What’s left is just whether you built something real underneath it.”
