Cancel an engineering hire when the work you planned to pay for has become cheap to test and the real bottleneck is still unanswered demand. Keep the role when a human engineer would create product judgment, reliability, or customer learning that tools cannot supply.
Dami, a composite Lagos founder who kept a notebook full of half-crossed-out product ideas, was staring at a take-home submission on Friday afternoon. His candidate had built the workflow Dami requested: a small tool that cleaned customer messages, classified the request, and drafted a response for review.
The candidate had also included the prompt history.
Dami had expected that task to tell him whether he could finally hand off the growing backlog. Instead, it showed a person directing an AI coding tool through an ordinary product problem in a few hours, checking the output, fixing the rough parts, then explaining the decisions. The candidate was capable. That made the decision harder.
The offer letter sat in another browser tab. By Monday, Dami needed to decide whether to send it or preserve the runway he had set aside for the role.
If he hired too early, he would add a monthly commitment before he knew whether customers wanted the workflow. If he passed, he risked staying stuck with a product held together by late nights and temporary fixes while a competitor moved first. The candidate had made the prototype look easier. The business decision had not become easier at all.
The task revealed a different bottleneck
The original hiring brief sounded familiar: build faster, clear the backlog, add features customers had requested. But the take-home task forced Dami to separate those requests.
A customer had asked for automatic replies. Another wanted better handoffs. A third had described a problem that sounded like a reporting need, though nobody had watched them use a report. Dami had grouped all of it under “engineering work” because that was the box available to him.
The candidate’s submission exposed the gap. A working demo could now be produced quickly enough to test with a small set of people. What Dami lacked was a decision about which problem deserved a durable system.
He cancelled the role before Monday. He did not cancel engineering. He cancelled a hiring plan built around a backlog that had not earned its own existence.
That distinction matters. AI tools can reduce the cost of turning an idea into something a customer can react to. They do not tell you which customer request contains a real budget, which workflow will fail under pressure, or whether the person asking for an automation will trust it once it sends the wrong message.
A similar tension sits inside Will an Engineering Hire Help You Learn Faster Than It Raises Monthly Costs?. The cost of a hire is never only salary. It is the roadmap you commit to while trying to justify the hire.
A prototype is evidence, not a product plan
By Saturday morning, Dami had changed the assignment for himself. He picked one customer workflow, kept human review in the loop, and set a narrow test: could the tool remove enough repetitive work that someone would use it again next week?
The first version still had awkward edges. One classification was too broad. The draft replies occasionally sounded like they had been written by someone who had never spoken to the customer. That was useful. A full-time engineer could have made those rough edges less visible, but Dami needed to see them.
This is where hiring rituals become expensive. A founder writes a role description that asks for a builder, a product thinker, an AI specialist, a reliable operator, and someone comfortable with uncertainty. Then the company spends weeks evaluating candidates against a task that an AI-assisted builder can complete before lunch.
The ritual may continue because the company wants reassurance. A new hire feels like progress. A finished demo feels like proof. Neither answers the question beneath both: what has to be true for this product to deserve more permanent capacity?
Dami’s candidate had shown something more valuable than speed. She had documented where the tool guessed, where she intervened, and what she would test with a real user. That was the signal Dami needed to preserve. He wrote back, explained that the role was paused, and kept the conversation open for a smaller, defined piece of work after the customer test.
Hire for responsibility after the learning is clear
The engineering job had not disappeared. Its shape had changed.
A strong engineer becomes more valuable when the founder can name the system that must keep working, the customer behaviour that has been observed, and the tradeoff that needs an owner. That could be a workflow handling sensitive customer information, an integration that has to survive real usage, or a product surface where reliability will determine whether a pilot becomes revenue.
At that point, AI-assisted prototyping does not replace engineering judgment. It makes the judgment more visible. The engineer is no longer being hired to convert every untested request into code. They are being hired to make the right things dependable, maintainable, and easier to extend.
Dami did not treat the cancelled offer as a victory for tools. He treated it as a warning about his own vague brief. On Monday, he had a smaller backlog, a clearer customer test, and the candidate’s notes beside his notebook. The crossed-out ideas were still there. This time, one of them had a reason to survive.
Comments
No comments yet.