Service · Kuala Lumpur
AI automation with a checking step.
I take one manual process that keeps breaking, such as lead intake, reporting or document checks, and rebuild it as an AI agent workflow. Every run is tested against a definition of correct output, failures go to a person, and the cost of each run is capped.
Processes worth automating first
Work that repeats every day, follows rules a person can write down, and costs you when it is missed.
Lead intake and triage
Enquiries from your forms and inbox are read, sorted by what the person needs, answered with the basics and passed to the right person the same day.
Reporting
Numbers pulled from the tools you already use and written up the same way every week, with the figures checked against their source.
Document checks
Quotes, invoices and method statements checked against your own list before they leave the building, with anything missing flagged to a person.
Research and first drafts
Background research and first drafts with their sources attached, ready for a person to edit rather than write from a blank page.
How a project runs
Brief, build, check, ship: the same loop every project of mine goes through.
Map it
We walk through the process as it runs today: where it fails, why, and who signs it off. The map and the definition of a correct result are written down and agreed before anything is built.
Build and check
Agents do the work and a separate step tests every output against that definition. Anything that fails is retried or routed to a person, never sent on silently.
Run and log
Every run is logged, so you can see what happened and why. The cost per run has a cap agreed in advance, and the workflow stops rather than overspends.
Questions about AI automation
What does AI automation cost?
It depends on the process and the tools it touches, so I quote per project. Describe what happens today and what goes wrong, and I reply within one working day with how I would approach it.
Which processes are worth automating first?
The ones that repeat every day, follow rules a person can write down, and hurt when they are missed: lead intake, weekly reporting, and checking documents before they go out are the usual starting points.
What happens when the AI gets something wrong?
A separate checking step tests every output against a written definition of correct. Anything that fails is retried or sent to a person, and the failure is logged so its cause can be fixed.
Do I have to change the tools my team uses?
No. The workflow is built around the tools you already use, such as email, web forms and spreadsheets, rather than asking your team to switch.
Who builds it?
I do, through Virin Digital, my studio in Kuala Lumpur. I work with businesses across Malaysia; most of the work happens online.
Why should I trust AI with a real process?
You should not trust it blindly. AI output often looks right when it is wrong, so every workflow I build has a separate check on each result, sends anything that fails to a person, and logs every run so a mistake can be traced to the step that let it through.
Have a process that keeps breaking?
Tell me what happens today and what goes wrong. I reply within one working day.