Commerce Online · Singapore Practical AI for the businesses that already run this economy Singapore · Est. 1988
Commerce Online
01Why a test bed

Nobody should buy AI on the strength of a demo

Every demo works. Demos are built on clean data, on a happy path, by the person who wrote them. Your business is none of those things.

  • It is small on purpose. One process, one measurable outcome. A wrong answer costs you weeks, not a year’s budget.
  • It runs on your data. Your documents, your exceptions, your volumes. That is where AI projects fail, so that is where we start.
  • It ends with a decision. A measured before-and-after and a written recommendation — including, when it is right, not to proceed.
02How it works

Five steps, one fixed period

Dates and scope are agreed in writing before we start. Nothing expands halfway through.

  1. 01ApplyTell us the process and what it costs today. Ten minutes.
  2. 02Scoping workshopOne session with the people who own it. We agree the measure.
  3. 03BuildA working solution on your data, with weekly check-ins.
  4. 04Run and measureYour team uses it for real, beside the current process.
  5. 05DecideWritten result and recommendation. Scale, park, or stop.

↑ The step a pilot without a decision never reaches

03The bargain

What we ask, and what you get back

Printed as a pair because they are one bargain. Neither column works without the other.

Four things from you

  • A named owner for the process who can decide how it should work. Not just an IT contact.
  • Access to real data, under whatever protection agreement you need. Synthetic data proves nothing.
  • A few hours a week from the people who do the work.
  • Honest feedback, especially when what we built is annoying to use.

Four things back

  • A working solution on your own data. Not a prototype switched off at the end.
  • A measured before-and-after against week one, in a form you can put to a board.
  • A written recommendation: what to do next, what it costs, and what we would not do.
  • Everything we built, documented, whether or not you continue with us.
What is not settled on this page yet

To confirm before launch. The commercial terms of the test bed — duration, price or whether places are free, how many places are available, and the eligibility criteria — still need to be settled and stated plainly here.

Prospects will look for them immediately, and vagueness costs more applications than an unwelcome number would.

04Week one

We measure the process as it runs today, before we change anything

An administrator keying a thick stack of paper dockets into an old desktop system, a colleague at the next chair with an open notebook. Filing shelves, a wall calendar and a window fan behind them.
Fig. 2The process as it runs today: paper in, keyed by hand, one docket at a time. Week one is spent measuring this, not replacing it. Generated image, not a photograph of a client site.
05Apply

Tell us about the process

No obligation, no mailing list. If a test bed is not right for you, we will say so and explain why.

The one that wastes the most time, or goes wrong most often.

Hours a week, error rates, missed orders, overtime. An estimate is fine.

Before you start

About ten minutes. The more specific the process, the more useful our reply.

If you would rather talk it through first, that is fine too — the call comes to the same place.

Talk to us instead