A VC-funded startup automated its operations and neglected its customers
A founder in Wilmington, an advisor in Melbourne and a founder in Tallinn say the costly mistakes sit on both sides of automation.
One VC-funded startup spent so much time automating operations that it lost sight of what actually mattered.
Very efficient. Slightly beside the point.
Meanwhile, a founder in Tallinn names the opposite mistake: "wasting founder time on repetitive tasks".
We asked which parts of a job AI has actually taken, and which still need a person. Three leaders in the US, Australia and Estonia answered with a practical test: what can you check, and who answers when it goes wrong?
Two ways to lose money. One of them looks responsible.
Nicolas Falourd, Founder of Cyber Techwear (Tallinn, Estonia)
“Most expensive mistakes are automating without oversight (bad decisions amplified) and under-automating (wasting founder time on repetitive tasks).”
“When choosing hire vs agent I look at error cost and feedback loops: low-risk repeatable work → agent; high-stakes or ambiguous work with learning value → hire; when both, use agents to scale capacity and humans to audit and coach.”
Price the error first, then pick the worker. Learning value is what tips a task toward a hire.
The startup that automated past its own customers
Gary Smart, Advisor and Author (Melbourne, Australia)
“I worked in a VC funded startup that spent so much time automating its operations that too little effort went into understanding real customer needs and delivering its product.”
“When review and documentation lag behind generation, teams can inherit systems they can change quickly but struggle to diagnose when something goes wrong.”
Gary's point isn't that speed is bad. It's that speed you can't diagnose becomes a bill that arrives later, and his fix is to name the outcome, the quality check and the owner up front.
Can AI do it? Ask a better question first.
Morteza Moezodini, Founder of KAVOMAZ (Wilmington, United States)
“Humans should retain responsibility for decisions involving commercial commitments, financial risk, sensitive information, customer relationships, and situations where context or ethical judgment matters.”
“The most important question is not simply whether AI can perform a task, but whether the task can be verified reliably and whether mistakes can be detected before they cause harm.”
Morteza gives agents the structured, repetitive work: research, first drafts, pattern spotting. The line he draws is verification.
Where they split
Where they split is emphasis, not direction.
Nicolas worries about the founder doing by hand what an agent could do, and counts that as a real cost.
Gary worries about the team that automates fast, then can't explain what it built or why customers stopped being the point.
Morteza sets the gate: if mistakes can't be caught before harm, a person keeps the task, and commercial commitments and ethical judgment stay human by default.
All three keep a human on the hook for the outcome. The disagreement is about where the hook sits.
My take
AI takes the pieces you can verify. People keep the pieces where someone has to answer for the result. Before any hire or any agent, ask who catches the mistake, and how fast. If the answer is nobody, that piece stays with a person.
Your turn
Which task have you handed to AI that you could actually verify, and which one do you now wish you had kept?
Next: a South African agency cut admin by more than half yet watched conversion quality drop, while legal and consulting leaders say lost trust costs more than any accuracy gain.
