95% of billing automated in Malaysia. Flag-only in Erie. Who's right?
A Kuala Lumpur founder, an Erie CEO and a Melbourne CEO draw three different lines around money decisions.
One founder says "agents now automate 95% of billing tasks" in a published example.
Another uses automation to flag unusual activity, then hands the decision to a person.
A third has watched founders try to automate pricing and hiring, and run straight into walls.
We asked founders which parts of a job AI has actually taken, and which still need a person. When money is on the line, the answers stop lining up. That matters now because every company is deciding, role by role, whether to hire, deploy an agent, or do both.
Write the rules down, then hand over the invoices.
Komy A. (Mahmoud Ashraf), Founder of Genta AI Solutions (Kuala Lumpur, Malaysia)
“Work that is high-volume and can be written down as rules goes to an AI agent, with a person approving anything irreversible.”
“A published example from our work: agents now automate 95% of billing tasks for C&G Energy Services.”
The 95% is the headline. The approval step is the design. Agents get the volume, and a person keeps the veto. He splits the work by decision, not by job title.
Let the software find it. Let a person decide.
Kyle Barnholt, CEO & Co-founder of TrewUp (Erie, Colorado, United States)
“We can use automation to find inconsistencies, organize documents, and flag unusual activity.”
“We believe AI works best when it supports accountability rather than replacing it. It gives human reviewers clearer information, stronger evidence, and more time to make thoughtful decisions.”
Look at the verbs the software gets: find, organize, flag. The verdict stays with a human reviewer, and he treats that as the point, not a limitation.
The wall shows up exactly where judgment starts.
Shammika Munugoda, CEO of Apptimistic Pty Ltd (Melbourne, Australia)
“The real question isn't person versus agent - it's which decisions still need human judgment, and which don't.”
“I've seen founders try to automate hiring decisions, pricing strategy, or client relationship judgment and hit walls.”
He is not anti-agent. Follow-ups, scheduling, invoice tracking and basic triage already run on agents in his world, he says. The trouble starts when founders aim automation at judgment.
Where they split
Where they agree: judgment, accountability and relationships stay with a person.
Where they split: how much of the work around a decision an agent should finish before a person looks.
My take
I'd bet the 95% camp and the flag-only camp are closer than they look. Both keep a person at the moment money or trust changes hands. The real argument is how much work gets done before that moment, and who checks it. If that holds, the useful question for a hiring plan is who owns the moment money or trust changes hands.
Your turn
Where does your company draw the line between work an agent can finish and a decision a person has to own?
Next: a VC-funded startup that automated its operations at the expense of customers, and an Estonian founder who warns that under-automating wastes founder time just as surely.
