While Silicon Valley chases flashy AI demos, a new guard of founders is quietly rebuilding the world’s plumbing

The current conversation around artificial intelligence (AI) is suffering from a terminal case of “main character syndrome”. We are obsessed with the flashy demo — the AI that writes poetry, paints masterpieces, or promises to replace the C-suite.

But after looking under the hood of the industries that actually keep the world turning — from underwater power cables to chronic disease management — it is clear that we are looking at the wrong thing. 

If we want to understand the future, we need to stop looking at the stage and start looking at the plumbing. AI isn’t here to be our replacement; it’s here to be our nervous system. And if that sounds less exciting than a robot overlord, it’s only because we haven’t realised how broken our “human” systems actually are.

The most transformative AI products today are the ones you’ll never see. They are “quiet” by design. Operators are now building tools that refuse to show off. Agastya Samat, CEO of Zeya Health, an AI-powered operational system that automates healthcare administration to give clinicians more time for chronic disease care, intentionally keeps AI out of the patient’s line of sight because the people closest to patients are often the least empowered by the tools meant to help them.

We don’t need an AI to pretend it has empathy; we need it to handle the “invisible work” — the scheduling, the follow-ups, and the administrative rot that forces doctors to treat patients like line items on a spreadsheet. 

Kritika Seth, CEO of Idle and Tetra, a management ecosystem that streamlines the fragmented “rhythms” of service-based businesses like salons and cafes, echoes this restraint, arguing that a service worker doesn’t want to feel “cognitive overload” or be managed by an algorithm within a messy workday.

If the work requires more of a human touch, the admin shouldn’t be involved. Still, this restraint has a cost: fewer flashy demos and less immediate market hype. But we have to decide: Do we want AI that makes for a great pitch deck, or AI that gives a business owner their evenings back?

Sovereignty over statistics

The truth we find hard to swallow is this: you cannot optimise your way through a human relationship. The tech industry is obsessed with “optimisation” — the idea that there is a single, mathematically perfect way to run a clinic, a salon, or an underwater inspection. But, to nobody’s surprise, humans are not homogeneous “lots” of data.

Take it from Grace Chia, CEO of BeeX, an autonomous underwater robotics company building small drones to inspect and protect critical sea-based infrastructure, who notes that “optimisation will not achieve good outcomes as fundamentally; we believe that humans always want a choice”. 

In the beauty industry, over-automation risks turning a local salon into a “conveyor belt”, stripping away the very “pauses” and personal touches where care actually lives. We see this failure in healthcare, too, where, according to Agastya, “averages don’t exist at the individual level”.

If an AI system flags a staff member for a performance dip based purely on revenue, as Kritika observed during early testing, it is technically correct but humanly wrong. It misses the “emotional labour” and the “life events” that a metric can’t capture.

The line between assistance and substitution is where the real battle for the future is being fought. These founders are drawing that line at accountability and judgment. Alan Lai of ProfilePrint, a digital fingerprinting platform that uses molecular signatures to provide rapid quality insights for food and materials, notes that while a thermometer provides accurate fever readings, the parent — not the tool — decides whether to call the doctor.

AI provides the “molecular signature, but the human makes the decision based on context that the machine will never have”.

Reclaiming the human loop

To move forward, we must stop asking whether AI will take our jobs and start asking whether we are brave enough to let AI take our boredom. A good example is what Chia discovered in her work at BeeX.

Currently, underwater inspections are stagnant, reliant on 100-year-old methods and gatekept by five to 10 years of required experience. AI can break that gate, allowing more people to do the work with 10x less strain on the planet’s resources.

And if AI succeeds, the world won’t (and shouldn’t) look like a sci-fi movie. Rather, it will look like a world where the “chaos” has stopped. It will resemble a subsea drone, with the human operator always deciding how much autonomy to use. It will look like a doctor who actually looks you in the eye because they aren’t fighting a database. 

After all, the hardest human boundary isn’t technical; it’s our own willingness to stay in the loop. AI shouldn’t be the star of the show. It should be the quiet, invisible force that allows us to be more human, more decisive, and — frankly — less exhausted. Anything else isn’t progress; it’s just a louder version of the same old mess.