He built solutions to his own problems. Then he learned to ask whose problem it was

Water utilities, energy transmission, mining, government. The organisations Biplab Das Choudhury works with are asset-intensive — their business is pipes, plant, substations and the maintenance schedules that keep them running. Most of them run IBM Maximo to manage it.

Valueztech, his Melbourne company, is an IBM Silver Business Partner that implements and upgrades those systems. But the part he is most interested in sits on top of them.

Enterprise asset management is not a market of startups. The competition for the work Valueztech goes after is large, established and well resourced, and it responds to the same tenders.

Biplab does not describe AI as a product feature. He describes it as a mechanical advantage — the thing that lets a small team do work that used to need a large one.

“I see AI as a new lever in the software and intelligence area. And any good engineer would know how a good lever can move things around.”

The effect shows up first in the unglamorous work. Marketing, bids, tenders — the volume of material a small firm has to produce simply to be considered.

“I am a small company in Melbourne, Australia, and competing against large organizations. So the amount of work that AI is doing around marketing, bids, tenders is humongous, so I can do it with a handful of people.”

The clearest number is the tender itself. A response that used to take one to two months now takes one to two weeks, and less than that on the smaller ones.

The first product came out of watching engineers do something repetitive and invisible. When a Maximo system misbehaves, somebody has to go into the logs and find out why — and that somebody is usually a software engineer, reading, for hours.

The Maximo Logging & Monitoring Platform is Valueztech's answer: Maximo-aware observability rather than a generic IT tool, built to show what the estate is doing before a user raises a ticket.

The second product addresses a problem every engineering team recognises. The information exists — it is inside a scanned drawing, an O&M manual, a photograph — and finding it means a person opening files until they hit the right one.

Engineering Document Intelligence lets an engineering team ask in plain language and get the value back out of the document. It is in working beta, with pilots available, and the site puts recall at roughly 90% on complex documents, with the remainder flagged for a human to check rather than quietly guessed.

Both products started the same way: Biplab hit something that annoyed him, and built the fix. It is the most natural instinct an engineer-founder has, and it is the one the AI Pathfinder program interrupted.

“We used to think of a problem which we have faced and try to create a solution around it. But it is not always my problem is somebody else's problem.”

What replaced it was not a framework so much as a change of vantage point — looking at the same product from outside it.

The second change followed from the first. If the problem belongs to a customer, the customer can tell you whether you have understood it — and they can tell you early, before the expensive part.

“It also helped us to redefine what we are building, put it in front of customer to get it validated quick and fast… Getting customer feedback is very crucial.”

He is also clear about what the program was like as a room rather than a curriculum. Founders tend to assume their problems are particular to them.

What is next is the piece Biplab thinks the current generation of models is missing. A model can be capable and still not know how a particular organisation actually works — which is the knowledge that lives with the people who have been there for years.

“The idea is that if the models are very smart, but they don't understand your business as much as your internal staff do.”

The team is researching an asset-management ontology: a structured layer that carries that context so the models can use it.