Job search and interview prep

Use it when you are applying for a job and want to use an assistant to tailor your applications and rehearse for interviews without inventing anything.

Not when you plan to put things on your CV that you could not talk about, or to use AI help during the interview itself. Follow the employer's rules on AI in interviews and tests.

You need an assistant, your current CV, a plain text file for your experience, the job ad and about 2 hours per application, plus 75 minutes once for the file. Leave out your ID numbers and other people's private details: What not to paste. The steps and times are our suggestion.

Steps

Step 1 happens once, and steps 2 to 8 repeat for each application.

For a quick application, do steps 2, 3 and 4 only (about 36 minutes). For an interview you already have, do steps 2, 5, 6 and 7 (about 73 minutes).

Repeat these three lines for each role or project in your file:

Role or project: <ROLE>, <EMPLOYER>, <DATES>
What I did and the result: <ACTION>, which led to <RESULT_WITH_NUMBER>
How I could show it: <EVIDENCE>

Replace the following:

• <ROLE>, <EMPLOYER> and <DATES>: such as Support team member, a regional retailer, 2023 to 2025 (illustrative).
• <ACTION> and <RESULT_WITH_NUMBER>: what you did and the outcome you can verify, such as trained three new hires and all three handled tickets alone within a month (illustrative).
• <EVIDENCE>: a document, a manager who would confirm it, or a story you can tell in two minutes.

The prompts

For step 2:

Here is a job ad. List, under four headings:
1. Must-haves.
2. Nice-to-haves.
3. The problems this role is hired to solve.
4. Likely interview topics.
For every item, quote the line of the ad it comes from. If the ad does not say, write "not stated". Do not add requirements the ad does not contain.
Job ad:
<JD>

Replace the following:

• <JD>: the full text of the ad.

Read the list, delete anything you disagree with, and keep your edited version for step 3.

For step 3:

Here is a job ad analysis and my experience file. Make a table with the columns requirement, my strongest matching evidence (quote it from my file), and strength (strong, partial or gap).
Use only what is in my experience file. If nothing matches, write "gap". Do not stretch or invent.
Then list the three gaps most likely to come up in an interview and how I could address each honestly.
Job ad analysis:
<JD_ANALYSIS>
My experience file:
<EXPERIENCE_FILE>

Replace the following:

• <JD_ANALYSIS>: the must-haves and nice-to-haves from step 2, as you edited them.
• <EXPERIENCE_FILE>: your file from step 1.

For step 6:

Act as an interviewer for <ROLE> at <COMPANY_TYPE>. Ask me one question at a time: four behavioural questions and four about the role, based on this job ad: <JD>.
After each answer, give a score out of 5, what was strong, what was missing (situation, action, result, numbers) and a one-line suggestion. Then ask whether I want to retry before we move on.
Be direct and do not praise weak answers.

Replace the following:

• <ROLE> and <COMPANY_TYPE>: such as customer support lead and a mid-size software company.
• <JD>: the ad text.

A worked example (illustrative)

Lena is applying for a customer support lead role. The details are invented. Her file has an entry on training three new hires in 2024 and another on writing help articles. The map shows both as strong matches and "run a weekly report" as a gap, because her file only shows a monthly sheet. The model suggests she say so plainly, and describe the monthly sheet and how she would build the weekly one. In the mock interview her answer on escalations scores 3 out of 5 for lacking a result. On the retry she adds the number from her file and scores 4.

How to tell it worked

• Every line of the CV traces to your file, and you can speak to each for two minutes.
• You did a mock interview and retried the weakest answer.
• It has failed if a line in your CV came from the model and not from your file. Go back to step 4.

Why no invented lines and no hidden text

Employers look for it. In a November 2025 vendor survey, 65% of US hiring managers said they had caught applicants using AI deceptively (AI-written scripts 32%, hidden prompts 22%, deepfakes 18%) (Greenhouse, vendor survey, self-reported).

Sources

1. Greenhouse, AI in Hiring Report release, November 2025. Vendor survey of 4,136 people in the US, UK, Ireland and Germany. Link

Changes

1. v1.010 Oct 2026first version