When an AI says you lack a skill, be careful
It didn't assess you. It assessed two pages of text.
· 5 min read · Interview Preparation
Paste your CV and a job description into any AI tool and you'll get a gap analysis back. It will be well written, confidently phrased, and specific: "Lacks demonstrated stakeholder management experience."
It reads like a finding. Most of the time it's something much weaker, and the difference matters — because people act on these. They rewrite CVs around them. Sometimes they decide not to apply.
Here's how to read one properly, including ours.
It didn't assess you. It assessed a document.
This is the whole thing, and almost every other problem follows from it.
The model compared two texts. When it says you lack something, what it observed is that evidence for it wasn't present in the two pages you uploaded. Whether you can actually do the thing is not a question it has any access to.
Absence of evidence becomes a stated absence of skill, phrased in the grammar of a conclusion. You ran stakeholder management for six years; you just never used that phrase, because at your company it was called "the Thursday steering call."
Four things "you lack X" can mean
When a gap gets flagged, it's one of these — and only the last is about you.
1. You did it and didn't write it down. By far the most common. Everyone under-documents their own work, particularly the parts that felt routine. The fix is a CV edit, not a skills problem.
2. You call it something else. Your "delivery oversight" is their "program management." Your "incident command" is their "production support leadership." The model is matching vocabulary, and industries have dialects.
3. The job description asked for everything. Job ads are wish lists assembled by committee, routinely listing tools the team touches once a quarter and skills nobody on it has. A gap against a wish-list item isn't a gap against the job. Analysis tools rarely distinguish the two, because the text doesn't.
4. You've genuinely never done it. This one is real and worth knowing. It's also, in my experience, a minority of what gets flagged.
Fluency is not confidence
There's a specific trap in how these outputs sound.
A language model writes every sentence in the same assured register, whether the underlying judgment is solid or a coin flip. "Lacks demonstrated experience in X" and "may benefit from additional depth in Y" are produced by the same process with the same amount of evidence behind them — which is to say, whatever was in your two pages.
Humans read fluency as competence. We've been trained by a lifetime of reading confident prose written by people who had reason to be confident. That instinct is wrong here. The confidence is a property of the generator, not of the finding.
The part that actually worries me
A machine telling you that you lack something lands differently from a person telling you.
A recruiter's opinion comes with visible context — you can see they skimmed, or misunderstood, or have their own agenda. A tool's output feels like a measurement. Neutral. Objective. As though it ran your career through a process and returned a result.
So people internalize it in a way they don't internalize human feedback, and the ones most affected are the ones already inclined to doubt themselves — often the same people who were under-selling their experience in the first place, which is precisely why the tool flagged them. The feedback loop runs the wrong way.
If a gap analysis has ever made you quietly decide not to apply for something, it's worth knowing it was a text comparison, not a verdict.
How to triage a gap list
Take each flagged item and ask, in this order:
- Have I done this? If yes, it's a documentation gap. Add it, with a number. Done.
- Have I done something they'd call this? If yes, it's a vocabulary gap. Use their word for it.
- Is this actually central to the role? Look at where it sits in the ad, whether it's marked required, and whether the responsibilities section mentions it at all. If it's a wish-list item, let it go.
- Have I genuinely never done it, and does the job clearly turn on it? Now you have a real gap.
Most lists shrink dramatically. A twelve-item gap analysis typically contains two genuine gaps, and often only one of those is central.
What to do with a real one
Real gaps don't mean don't apply. Every hire is made with gaps; the question a hiring manager is actually answering is whether this gap is cheaper to close than the alternative candidate's gaps.
So: name it, put your nearest adjacent experience next to it, and be straightforward about it. "I haven't run this at your scale. I've run the same thing at a tenth of it, and here's what I'd expect to break first." That answer gets people hired regularly. Silence on the gap, followed by a fumbled question about it in the interview, does not.
This applies to us too
labor.quest produces gap analysis, so everything above is a description of our output as much as anyone's.
It's why we report which specific requirements your CV doesn't currently evidence, rather than asserting what you can't do — the first is checkable, the second isn't ours to claim. But the distinction is easy to lose when you're reading quickly and the sentence sounds authoritative.
So treat any gap list, ours included, as a set of prompts rather than findings. The useful question was never is this true about me. It's is this missing from my CV, and does this job actually require it — and you're the only one who can answer the first half.
All posts