Signs You Hired the Wrong AI Engineer (Or Scoped the Wrong Role)

Is It the Person, or Is It the Role You Wrote?

Ninety days in, nothing is in production, and the demos are getting less convincing. The instinct at this point is a personnel decision, and it is usually premature.

Here is the reason it matters. A badly scoped role and a badly filled one produce the same visible symptoms: no shipped work, shifting explanations, a growing gap between what was promised and what exists. From the outside they are indistinguishable. And the two have opposite remedies. Replacing a person against an unchanged brief reproduces the same outcome one quarter later, having spent twice as much to get there.

So the diagnosis comes first. This is not a generous framing offered to be fair to the engineer. It is the cheaper order of operations.

The Four Scope Checks

Run these before forming a view about the person. Each one describes a condition that makes the work impossible regardless of who is doing it.

Was the problem defined as a process, or as a goal? "Use AI to improve customer support" is a goal. "Classify inbound tickets into these six categories so the queue routes automatically" is a process. Against a goal, any competent engineer builds something plausible, and it gets judged against expectations that were never written down. If the brief was a goal, the outcome was set before the person started.

Was the data reachable on day one? If the information the system needs sat in a vendor platform with no export, or behind a team that would not prioritise access, the engineer spent the quarter negotiating rather than building. Check the calendar rather than asking the question. If the first eight weeks contain access meetings, you have your answer.

Did anyone on the business side own the process? Automation changes how work happens, so somebody has to agree to the change and answer questions during it. With no named owner, the technical work completes and adoption does not. The system exists and nobody uses it, which reads as failure and is not the engineer's failure.

Did the volume justify the work? If a person was handling the task comfortably, automating it was a hobby. The engineer may have delivered exactly what was asked for and produced nothing that mattered, because the thing being fixed was not constraining anything.

If any of the four is a no, stop. You have found the cause, and it is not a hiring cause. The wider pattern of AI initiatives failing for scoping and talent reasons rather than technical ones is set out in why AI projects fail on the talent gap.

The test that resolves ambiguity, when more than one thing looks wrong: would a different competent person have shipped against this same brief, with these same conditions? If the honest answer is no, replacement changes nothing.

Run the four checks with someone who was not involved in writing the brief. The person who scoped the role is the worst-placed person to assess whether the scoping was sound, and that is not a criticism of anyone in particular; it is the same reason surgeons do not operate on their own families. If the brief came from you, hand the four questions to a colleague and let them ask you rather than asking yourself. The answers change more often than people expect.

What Actually Indicates the Person

These signals count only once the four scope conditions have been met. Against a bad brief every one of them appears anyway, which is exactly why the order matters.

Demos that never narrow. A working prototype in week three is good. The same prototype in week ten, broader and still not in production, is a pattern. Strong practitioners narrow scope as they learn; they cut things. Someone who keeps adding surface area is avoiding the part where real inputs break the assumptions.

No answer to how quality is measured. Ask how they know the output is any good. If there is no evaluation set, no held-out examples, no sampling method, only impressions, that is a capability gap rather than a scoping one. This one is close to disqualifying on its own, because AI systems degrade silently and impressions do not detect it.

Explanations that get vaguer under questioning. Follow-up questions should produce more specificity, not less. When "the model needs more tuning" survives three rounds of asking what specifically is wrong, the engineer either does not know or does not want to say. Both matter.

Production standards absent from the work. No error handling, no logging, no monitoring, nothing that would tell anyone the system had stopped working. This is the difference between someone who has shipped and someone who has built. It is also the most common genuine mismatch, and often not the engineer's fault: a research-leaning person hired into a production role is misplaced rather than incompetent, and the mismatch was created at scoping.

Work that only they can run. If nothing is documented and no one else can operate any of it after three months, you have a dependency rather than a capability. This is visible without technical skill: ask someone else to run it.

What to Do When the Scope Was the Problem

Fix the brief before touching the personnel question, because until the brief is fixed you cannot tell whether the person can do the job.

Rewrite the scope around one named process, with a written description of what correct output looks like. One process, not a theme. If you cannot write down what a correct result is, you are not ready to restart, and that is the finding.

Secure data access before the restart date rather than after it. Get the export confirmed working, not promised.

Name a business-side owner who will answer questions during the transition and who has agreed to the change. Not a sponsor. Someone whose actual work is affected.

Record the baseline: volume handled, time per item, error rate, backlog. Before the restart, not after.

Then, separately, decide whether the current person works against the corrected scope. That is a different question from whether they failed the old one, and it deserves a clean answer rather than momentum in either direction. Plenty of engineers who look failed against a goal perform well against a process. Some do not, and now you will be able to tell.

Where the reset needs a structure, how to onboard a remote AI engineer in the first 30 days covers the sequence, and it works as well for a restart as for a first start.

What to Do When It Is the Person

Move quickly, and be clear with yourself about why.

The cost of a wrong AI hire is mostly not the salary. It is the elapsed time: the quarter spent finding out, and the second quarter usually spent hoping. Companies routinely spend more on the hoping than the salary was worth.

The US anchor gives a sense of the fixed part. No OEWS occupation covers AI engineering. The closest published proxy is Software Developers, SOC 15-1252, at a median annual wage of $135,980 (BLS OEWS, May 2025), loading to roughly $193,975 at 1.4265 per BLS Employer Costs for Employee Compensation (ECEC, Dec 2025) as an estimate of annual employer cost. Where the role sits closer to process work, Management Analysts, SOC 13-1111, at a $101,860 median (BLS OEWS, May 2025) loads to roughly $145,303. Both are proxies and may overstate or understate the real role, and the loaded figures are estimates from a national benefits average rather than a measured cost at your company. Two lost quarters against either figure is the real number, and it is larger than any recruiting fee you were trying to avoid.

This is where the hiring route starts to matter, because routes differ mainly in how expensive it is to be wrong. Direct employment makes a correction slow and personal. A managed arrangement moves the employment relationship, and the replacement, off your desk.

Through F5 Hiring Solutions, replacement is 7-14 days at zero cost, at any point, with no replacement fee and no minimum engagement period. Shortlists arrive in 7-14 business days from a network of 85,500+ pre-vetted professionals; that timeline is F5's own published commitment rather than an independently measured benchmark, and any competing timeline should be read the same way. AI roles start at $600 per week, all-inclusive, inside the $375-$1,200 band, with AI Solution Architects from $800 per week. All-inclusive covers salary, HR administration, payroll, equipment, compliance, and management, with no setup fee, recruiting fee, or termination cost. F5 Hiring Solutions has served 250+ US companies with a 95% client retention rate, measured as clients continuing beyond the first three months. Placements cover AI engineering, LLM engineering, AI/ML engineering, and AI solution architecture; a request outside those roles is a no rather than a stretch.

Whether the hire should be remade at all, rather than remade faster, is the question in whether an AI engineer is worth the cost at all.

The Bottom Line

Diagnose before you decide. Four checks: was the problem a process, was the data reachable, did someone own it, did the volume matter. If any is a no, the brief failed and no replacement fixes it.

If all four hold and the work still did not ship, the signals that indicate the person are specific: scope that widens rather than narrows, no way to measure output quality, explanations that lose detail under questioning, and nothing anyone else can operate.

Then move fast, because the expensive part of a wrong hire is the time spent not deciding.

To make that correction cheap rather than slow, hire remote AI and ML engineers from India.

Schedule a 15-minute call: https://calendly.com/joel-f5hiringsolutions/f5