← All writing

Field Notes / Interview Prep

AI Mock Interviews: The Honest Version

I hired solution engineers for a decade. Then I became the candidate. Here is what these tools fix, what they miss, and how to use one without fooling yourself.

By Nicolai Imset7 min read
interview prepaicareers
The pitch, audited

What the page promises

  • Real-time vocal tone analysis
  • Confidence scoring
  • Eye-contact coaching
  • "87% of companies screen with AI"

What a text tool delivers

  • Text the model never hears
  • A structure score on written answers
  • No camera, no eyes, no contact
  • A statistic with no source

I have sat on both sides of the table

I spent thirty years in enterprise tech. For the last ten I ran presales teams. I hired solution engineers across eighteen countries. I built the scorecards. I sat on the panels. I rejected strong resumes for weak answers.

Now I am the candidate. Oracle made my role redundant. At fifty-four I write cover letters again. I book calls again. I rehearse answers again.

So I built an AI interview tool. Then I read how the rest of the market describes these products. Most of it is fiction. This is the honest version.

The fiction sells well

Open ten AI interview pages. You meet the same promises every time. Real-time vocal tone analysis. Confidence scoring. Eye-contact coaching. A headline statistic about how many companies screen with AI, printed without a source.

I read one product FAQ last week. It claimed the tool coached eye contact. The tool reads typed text. It owns no camera. That is not a feature. That is a sentence a language model wrote at three in the morning. Buyers notice the gap on day one. They ask for a refund on day two.

The hype hurts the category. It trains candidates to expect a flight deck of sensors. The real value sits somewhere narrower. It also sits somewhere more useful.

What an AI mock interview fixes

Strip the fiction. A genuine product remains. An AI mock interview built on text fixes the part of your answer you control most directly. The words.

  • Structure. The tool reads your answer. It shows where the logic breaks. You see the moment you buried the result under three sentences of backstory.
  • STAR discipline. It flags a missing Result. It flags a Situation that runs too long. Behavioral answers live or die on the result. Most candidates skip it.
  • Over-explaining. It flags the answer that runs long and says little, or leans on jargon to sound senior. Tighten it. The same claim lands harder short.
  • Volume. You run twenty reps before a human coach answers your email. Reps build the pattern. The pattern lowers the panic.
  • Resume alignment. It checks that your spoken story matches the claim on your resume. A recruiter probes that gap. The tool finds it first.

None of this needs a microphone. None of it needs a camera. It needs a model that reads structure. That part works today.

What an AI mock interview misses

The limits matter more than the features. Read them before you trust the score.

  • Chemistry. A panel decides in part on whether they want you in the next meeting. No model grades that.
  • Fit. Whether your answer suits this company is a judgment call. The tool grades the answer in the abstract.
  • The follow-up that goes sideways. A sharp interviewer hunts your weak point. The drift of that exchange decides real interviews. Scripted follow-ups approximate it. They fall short of it.
  • The room. Tone. Pace. The pause before a hard question. A human reads all of it. A text tool reads none of it. Train the words with the model. Rehearse the room with a person.

A tool that admits these limits earns more trust than one that paints over them. I would rather lose a sale than ship a claim I would reject in a candidate.

How to use one in a real search

Random practice wastes the reps. Run the sequence instead.

  1. Align the resume to the role first. Match your resume to the target job description. Your practice library grows from that overlap.
  2. Generate a focused question set. Pull questions from the skills the role names. Master a tight set. Skip the generic hundred.
  3. Run a cold baseline. Answer once with no prep. Find where your delivery falters before you drill.
  4. Fix structure before delivery. Repair the STAR gaps first. Cut the filler second. Worry about polish last.
  5. Close with a human. Let the tool carry the first eighty percent. Book a person for the last twenty. That is where the room gets rehearsed.

Where PrepEdge fits

I built PrepEdge to do the narrow thing well. It runs the mock interview itself — the AI plays the recruiter, the hiring manager, the technical round, and the panel, grounded in your resume and the target role. You answer in writing. It coaches each answer as you go, scores your answers for STAR structure, and hands you a scorecard at the end: answer quality, communication, role fit, and the gaps to work on.

It does not analyze your voice. It does not watch your face. I left those features out on purpose. I will not print a capability I would fail a candidate for inventing.

Use the tool for the eighty percent it owns. Run the rounds. Fix your structure. Align your story to the role. Then sit with a person and rehearse the part no model reaches. Candidates who treat AI as the whole answer walk in with a polished script. They meet a cold room. Candidates who treat it as the first eighty percent walk in ready.