The resume sounds impressive. What did the candidate actually do?

Read this sentence before you decide

Example resume claim

“Led a complex customer migration, aligning stakeholders and improving adoption.”

There is plenty to like in that sentence. There is also very little you can assess. What moved? What did this person decide? Did adoption improve, or did the project simply finish?

AI can make an application easier to read. It can also make an ordinary contribution sound larger than it was. A reviewer still has to work out what happened. The same problem existed before AI: people compressed projects into achievement bullets and teams filled in the missing detail.

Take Mara, an invented candidate for an implementation coordinator opening. Her sentence above is relevant to the role. It deserves a question. It does not yet establish technical ownership or an adoption result.

Three questions that get past the wording

• What was being migrated, and what made the handover difficult?
• Which decisions were yours? Who owned the technical work and approved the launch?
• How did the team check completion? What supports the adoption claim?

Keep the original sentence beside your notes. Otherwise “aligned stakeholders” can quietly become “managed the steering group”, and “migration” can become evidence that the candidate wrote the scripts. Those are additions made by the reviewer.

Give Mara room to explain the work without sharing customer records. An anonymised account of the sequence, responsibilities and checks can answer these questions. Formal verification, where needed, belongs in the employer’s established process.

What Mara’s answer changes

Mara says she coordinated the timetable, collected sign-offs and escalated a missing field mapping. An engineer wrote the migration script. Her manager approved the launch. The team checked completion through reconciliation and customer confirmation. She cannot explain the broader adoption claim.

Scroll sideways to read the full table.

That is a useful result. The coordination example remains relevant. Technical execution belongs to somebody else. Adoption needs clarification. A reviewer can now choose the next question without defending their first impression.

Ask Mara what she meant by adoption and let her correct the wording. A loose phrase may reflect compression, misunderstanding or an inaccurate claim. Record a contradiction and the response if one remains; do not choose an explanation before hearing it.

A detector score does not answer these questions

OpenAI withdrew its text classifier in July 2023 because of low accuracy. Its published limitations included confidently mislabelling human text and difficulty with short passages. That finding concerns the discontinued classifier; it is not a test of every detector sold today.[openai-classifier]

A suspicion about AI use should not enter the evidence column as a fact about honesty. Repeated sentence patterns and polished business language do not establish who did the project. A candidate’s admission that they used an editor does not establish it either.

NIST describes confidently false generated content as confabulation. That risk is a reason to check facts, not a reason to conclude that Mara’s sentence came from AI.[nist-profile]

Set application rules before candidates apply. If writing without assistance matters to the job, assess that skill under clear conditions. A finished resume cannot show you the drafting process.

For this opening, the next conversation is about a coordination decision. Write down what answer would change your assessment. That will help more than another guess about how the sentence was written.