11town.
01 / 12
▟ this whole site runs on EAOS
scroll ↓
How I work

Anyone can wire an automation.
Almost nobody can tell you where it belongs.

I build the systems that run Malaysian businesses, starting with my own.

The shift

Every business can buy AI now. That is exactly why it stopped being an advantage.

When everyone has the same tools, the tools are not the edge. The edge is knowing which part of your business should be handed over, which part should not, and what happens the first time it gets something wrong.

That judgment is the work.

Where the time goes

This is your week.

Ten hours you will never invoice. Quotes, follow-ups, invoices, chasing payment, the report nobody reads.

Paid for in your evenings, every week, whether you talk to me or not.

Where it stops

The tasks are not the problem. The gaps between them are.

Every one of those blocks ends with something waiting on you. A reply, an approval, a number only you know. The work is not slow because people are slow. It is slow because it keeps stopping.

Softwarewhen the rules and the inputs are predictable
An agentwhen the goal is clear but the path is not
Youwhen it is ambiguous, expensive, or hard to undo
One run, end to end

It stops when it should.

intake

“bro can quote me 40 units of the 5m cable, same as last month price ah?”
Quote draft#Q‑2291
5m cable × 40RM 18.50
tier 2 price, held from last month− RM 60.00
delivery, KlangRM 45.00
TotalRM 725.00
● not sent — waiting for you
illustrative run · your workflow, your systems
intake
agent
you
record
enquiry received · whatsapp
parsed · name, item, quantity
checking stock against the live sheet
pricing matched · tier 2 customer
drafting the quote
paused · waiting for your approval
approved · by you, just now
quote sent · record posted
evidence log complete
✓ Nothing moved until you said so.
01 · Audit

Find the one workflow worth rebuilding.

Before anything gets built, I sit inside how the work actually happens. Not the documented process. The real one, with the exceptions and the workarounds and the thing only one person knows.

  • 01How the work happens today
  • 02The same work, rebuilt
  • 03The one workflow chosen, and why
  • 04What the system may and may not do
  • 05Hours and ringgit, quantified
02 · Evals

Prove it behaves, before it touches anything real.

AI is not deterministic. So you do not ship it on a demo and a good feeling. You write down what correct looks like, run the cases, and keep the evidence when it fails.

CaseRight dataStepsMatches youSafe to act
Normal
Edge
Missing information→ you
Ambiguous→ you
High risk→ you

Three rules get written down before launch: when it must escalate, what it may never do alone, and how much autonomy it starts with.

03 · Deploy

Build over what you already have. Widen slowly.

No migration. No rip and replace. It runs on your accounts, over your existing tools, and it starts with the smallest useful action. It earns more authority only after it has proved it deserves it.

Every step it takes is logged. If you ever need to know what it did and why, the answer is a record, not a guess.

And then again

Audit, build, evals, deploy, observe, improve.

AuditBuild EvalsDeploy ObserveImprove

Then again, on the next workflow. One hour at a time, until the week looks different.

The point

Same week. Fewer things waiting on you.

Start here

Thirty minutes. I tell you what to fix first.

No pitch deck. I look at how the business actually runs and tell you where a system fits, and where it does not. If there is nothing worth building yet, I will tell you that too.

One operator. On the hook for the result. I take a handful at a time.