At ProfilePrint, a former Air Force commander is proving that AI in agribusiness works best with humans in charge

Most decisions about food quality are made without anyone ever tasting the food. They unfold across spreadsheets and shipping schedules, across assumptions of sameness that rarely hold, in a global system where variation is treated as an inconvenience rather than a truth to be reckoned with.

It is in this submerged layer of trade — far from the sensory theatre of tasting rooms and cupping tables — that ProfilePrint operates.

Yet, when its CEO and founder, Alan Lai, began building the company seven years ago, he resisted calling it “AI”. “We were really just using data to help users make better decisions,” he says. The phrasing feels almost understated now, at a time when intelligence has become a sexy (and lucrative) marketing adjective.

“If you look at our tagline, ProfilePrint provides the fastest insight so that our users can make better decisions,” Lai tells me. “That is essentially what we do. They may use us to inform their decision, but they make it based on human context and the other knowledge they already have.”

He repeats that distinction deliberately. ProfilePrint accelerates clarity; it does not replace judgment. In an era where many AI companies equate speed with autonomy, Lai draws the line elsewhere. Insight, he insists, belongs to the machine. Decision remains human. For Lai, restraint is the feature.

The work beneath the surface

“Think of it as two parts,” Lai shares when I probe for even more clarity about how ProfilePrint works. “One,” he explains, leaning slightly closer to the camera, “is the online software platform, the other is the custom-developed hardware that acquires a molecular signature.”

A client then places a sample into the hardware and “within 10 seconds, the signature is acquired and transformed into a stable digital fingerprint that goes into the software”. The AI reads that fingerprint and produces results “as if someone had tasted it or tested it in a lab.”

While about 80 per cent of Lai’s clients are in the food industry, anything with molecular variance can, in theory, be assessed. “Textiles, cotton, pharmaceuticals, or even bio-products in the healthcare industry.” 

At its core, ProfilePrint challenges not testing itself but the fiction of homogeneity. “The industry functions by the ‘lot’,” Lai explains. “A lot typically ranges from 100g to 1kg, depending on the industry.” The assumption is that the lot is uniform. “But the reality is that it’s not.” 

To account for that, he takes multiple scoops — “perhaps two for rice, but up to ten for highly variable ingredients like coffee beans — to accurately represent the homogeneity and acquire the digital fingerprint.” Ultimately, variation is acknowledged rather than smoothed away.

That acknowledgement matters because of who sits on each side of the transaction. Lai divides ProfilePrint’s clients into three groups. “The first is ‘traders’ — people who buy to sell,” he says, naming firms like Cargill, Louis Dreyfus, and Olam. “There are eight large food traders that command about 80 per cent of the world’s food trade, and more than half of them are our clients today.”

The second group is the buyers — big brands, CPG companies, manufacturers — who can now assess quality before goods ship. The third is the sellers: cooperatives, exporters, and farmers. “They use the technology to better sense their own quality and prevent rejection.”

By arming every player in this chain with the same molecular map, the traditional lag between shipping a product and knowing its worth begins to vanish. The effect is a radical compression of the supply chain — the digital fingerprint now replaces the need to ship physical samples.

Play

Lai likens the shift to “wargaming in the op centre”, a nod to his former military background. “Instead of waiting for soldiers to arrive to report their strength, you already know the status before they get there.”

Trust, responsibility, and refusal to decide

It would be easy to frame all of this as a triumph of quality control, but Lai resists that, too. “When we started, we saw that most people in the supply chain genuinely want to do the right thing, but they are working with incomplete or delayed information.” Someone might taste the middle of a bag, “but they don’t know the quality of the top or bottom”. The problem, as he sees it, is not dishonesty but opacity.

“Better data doesn’t replace trust; it supports it,” Lai adds, insisting that better data gives everyone a clearer, shared view of the reality of what is being bought and sold. Still, responsibility, in Lai’s view, never migrates to the machine. “It is always with the human, because the decision is made by the human.”

His analogy for this turns to his role as a father of two. “Think of a parent with a child who has a fever. The thermometer provides accurate data, but the thermometer doesn’t decide whether to go to the doctor or keep the child home from school — the parent does.” 

Lai’s fluency in the subject matter is surprising, given the path he took to get to where he is today — one that appears, at least on paper, to have little to do with the work he now leads. He describes himself as “a typical engineer by training”, a government scholar who studied aeronautical engineering, then served in the Air Force.

When his bond ended, he rejected a promotion, left the military, and spent the next eight years in commercial roles across China, Paris, London, and Africa. 

It was in Uganda that the misalignment became visible. Producers were “price takers”, he says, trapped in a system where quality and price barely correlated — grow a baseline crop and receive a fixed price, improve the quality and receive the same price, fall below an invisible threshold and face outright rejection.

“I realised this could be solved with a tool that embeds industry intelligence.” He observed, built an MVP, raised early-stage funding, and moved— slowly and deliberately. 

From Air Force to Ag-Tech

Play

That MVP now sits among ProfilePrint’s eight patents. At the time, he was told it was impossible. “People said you needed million-dollar lab instruments or expert taste panels.” He disagreed, not out of stubbornness but analysis. “By looking at the data and the inconsistencies of human assessment, I believed a faster, better tool was possible.” 

Parenthood sharpened the stakes. Lai founded the company the same year his first child was born, after which the work quickly became about sustainability and about making the food system “more effective and efficient for the sake of the next generation.

Responsibility then expanded again, to colleagues and clients. “Today, the three angles of importance are: for the next generation, for my clients, and for my staff.”

His philosophy on AI follows the same arc. AI, to him, sits alongside Google Search and factory automation — a productivity tool that creates capacity. “The real issue is what humans do with that enhanced capacity.” Waste it, or invest it. The technology enables conversation, but “the value is the human interaction”.

“I do have fears about extreme dependency where humans no longer understand things themselves.” It’s perhaps why he’s not coy to admit that ProfilePrint flags ambiguity deliberately. And when that happens, it asks for human judgment. “This interaction strengthens the AI while keeping the human’s skills sharp.”

This, I believe, is the unexpected core of Lai’s AI: a machine that refuses to make the final call. In an industry increasingly defined and sharply shaped by automated certainty, ProfilePrint succeeds by doing the one thing no one expected from a neural network — insisting that the human remains the most important part of the equation.