Use AI as a career coach

01What is AI actually good at in career coaching?#

Use it to organize your history, compare options against criteria you define, surface unanswered questions, and design small experiments. Do not ask it to identify your one true career, diagnose your personality, or make a high-consequence decision from a résumé and a few chat messages.

02What context should you give the system?#

Create a concise career brief with verified facts: roles, projects, strengths others have named, constraints, values, financial needs, and open questions. Include contradictions rather than polishing them away. Remove personal identifiers and sensitive employer information that are not necessary for the coaching task.

• A timeline of roles and major transitions
• Projects that gave or drained energy
• Feedback you have received more than once
• Skills you can demonstrate with evidence
• Non-negotiable constraints and current obligations
• Three questions you are genuinely trying to answer

03How should a useful coaching session work?#

Use one question at a time and require the system to separate observation, inference, and suggestion. Ask it to challenge your assumptions with evidence from the brief, then end with one small action and one question for a real person. Save the decision record, not every conversational detour.

1. Frame

   Name the decision, time horizon, and what would make the session useful.

2. Explore

   Generate several options without ranking them too early.

3. Challenge

   List assumptions, missing evidence, and the strongest case against your favorite option.

4. Test

   Choose a reversible action that produces new information.

5. Debrief

   Record what you learned and what still requires human conversation.

04How do you turn advice into a career experiment?#

Test interest and fit before making an identity-level commitment. Interview someone doing the work, complete a small project, shadow a meeting, teach the topic, or take a bounded freelance assignment. Define in advance what evidence would make you continue, revise, or stop.

05When should you bring in a person?#

Bring in a person when the decision affects health, identity, money, family, legal obligations, or a relationship you cannot model in a prompt. A manager, mentor, peer, therapist, lawyer, or financial professional contributes accountability and lived context that a language model does not have.

If a conversation is making you feel more certain without producing new evidence, pause. Confidence generated by fluent language is not the same as a better decision.

Questions people ask#

Should I keep one long career chat?

A project can preserve useful context, but keep a separate concise brief and decision log that you can review. Long conversation history can accumulate outdated assumptions and make them feel authoritative.

Can AI assess my strengths from my résumé?

It can identify patterns in the document, not validate your actual strengths. Pair the analysis with work samples and feedback from people who have seen you operate.

Can I upload performance reviews?

Only if the account, employer policy, and information involved permit it. A safer default is a de-identified summary containing only the feedback relevant to the question.

How do I avoid endless reflection?

End every session with a dated experiment or human conversation. If the next step produces no new information, it is probably more rumination rather than coaching.

Sources#

1. 01
   Projects in ChatGPT

   OpenAI

   Official guidance on persistent project context, files, instructions, and shared project visibility.

2. 02
   Data Controls FAQ

   OpenAI

   Current personal-account controls for training, retention, exports, and temporary conversations.

3. 03
   Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile

   NIST

   A primary framework for defining supported tasks, testing outputs, documenting limits, and retaining human oversight.