Commerce Online · Singapore Practical AI for the businesses that already run this economy Singapore · Est. 1988
Commerce Online
A distributor's floor with no wall down the middle: desks and monitors in the foreground, warehouse racking and a forklift down the aisle behind, an operations dashboard on the far wall.
Fig. 1A distributor with no wall down the middle — the desks and the racking in one frame, which is where our work sits. Generated image, not a photograph of a client site.
01Track record

The part you can check without taking our word for it

No logo wall. Most of the work sits under government contracts we cannot name. These four figures are verifiable.

1988Founded, and operating continuously since — through four platform generations.
24+Singapore government organisations we have delivered systems for.
1,000+Users supported on a single deployment, in production.
ISO15489The records management standard our practice is aligned to.

Registered Government Supplier · EPPU financial grade S8 — tender capacity to S$10M. It is a solvency and capability assessment, not a marketing badge.

Why a procurement grade is the evidence we lead with

The grade is what permits a public agency to award us work. It is assessed by someone other than us, and it is renewed against our accounts rather than our claims.

It is the plainest evidence we can offer that a system we build for you will still have somebody behind it in five years.

02Where this came from

We did not start as an AI company

We started in 1988 building B2B systems, and have followed our clients’ problems ever since: into document management, into the public sector’s digital transformation, and now into AI. Most of the recent decades were spent with Singapore government organisations.

That work sets habits. Requirements arrive as compliance obligations, not wish lists. QA is measured against a mandate somebody else wrote. A deployment serving a thousand users has to keep serving them after the project team has gone.

Which is why we can say this: AI that survives contact with a real business is not mainly a modelling problem. It is an integration, data and operations problem, and those we have been solving for a long time.

AI that survives contact with a real business is not mainly a modelling problem. It is an integration, data and operations problem.

What thirty-eight years taught us
03How we work

Four things we will say to your face on the first call

Every consultancy publishes values. These are the four that cost us work, which is the only reason to print them.

1
Sometimes the answer is that you do not need AI
Often the fix is a report nobody built, or a field nobody made mandatory. We have ended discovery calls saying so, and not invoiced.
2
Most of the work is not the model
Wiring a model takes about a week. The rest is data that is not where the documentation says, permissions nobody has audited, and four people who must change how they work.
3
One narrow thing beats a programme
The smallest change that produces a measurable result in production, with a rollback plan. Then widen. If nobody can tell whether it worked, we wrote it badly.
4
We stay after go-live, and expect things to break
Data drifts, businesses change, vendors deprecate things at a quarter’s notice. An AI system nobody owns eighteen months later is a liability with a login page.
04The difference

We don’t just advise. We build.

The gap between the roadmap and the build is where most of the money goes.

Advice, then a handover
A consultancy writes the roadmap. Someone else builds it months later, having been in none of the conversations. Assumptions are lost in the gap.
One team, discovery to support
The engineers who will build it sit in the discovery workshop. We cannot propose what we are not willing to be held to.
05Who you would be working with

The engineers who build it are the ones who sat in your discovery workshop

Three colleagues at one monitor in a cramped back office, all looking at the screen rather than the camera. Box files on steel shelving, a wall calendar and a desk fan behind them.
Fig. 2Two engineers and an operations lead at one screen, in the client’s own back office rather than a meeting room. Generated image, not a photograph of a client site.
06Fit

Who this works for, and who it does not

Printed in full because a wasted first call costs you more than it costs us.

You will get value quickly

  • 15–200 staff in Singapore, with a system of record you already pay for — ERP, CRM, a job or contract system.
  • A process that hurts weekly and that somebody could describe in ten minutes.
  • One person who can say yes without a committee.
  • Distribution, professional services, manufacturing, logistics, retail. The processes that hurt are much the same.

We would be selling you the wrong thing

  • Records living mainly in spreadsheets on individual machines. Nothing for an assistant to read.
  • A brief that starts with the technology rather than the process.
  • A board that needs the case proven before anything is built. Reasonable, but not what a short engagement does.
  • Anything where a wrong answer cannot be caught by a person first.

Not sure which side you are on? That is the first ten minutes of the call, free either way.

Talk to us

Come and test the claim

Everything on this page is an assertion. The only way to check one is to give us a real process and a fixed period.

Apply for a test bed Talk to us first

Questions we are happy to be asked

  • What have you built closest to my business? We will name it, or say we have not.
  • Who exactly would be on this? You meet them before you commit.
  • What has gone wrong on a project like mine? There is an answer. Ask for it.
  • What happens if we stop? You keep the code, the data and the documentation.