Career reinvention: map transferable evidence before chasing titles

A career change often begins with the wrong question:

What new job should I do?

That asks for a title before you have mapped the evidence. It encourages an AI assistant to produce an attractive list based on a short biography: project manager, consultant, product owner, trainer. The suggestions sound plausible because the titles are broad. They do not show whether your experience actually transfers or whether an employer would recognize it.

A stronger process starts with episodes from real work. What situation did you face? What did you do? Under which constraints? What changed? Who could verify it?

Then you can form a career hypothesis and test it cheaply.

AI can help organize evidence, translate terminology, and prepare research. It cannot predict who will hire you, determine whether a field will suit you, or turn unsupported claims into experience.

Keep private history private

Before opening an AI tool, separate career evidence from employer data.

Do not paste:

• customer names or records;
• confidential project details;
• unpublished financial or performance data;
• colleague evaluations;
• internal incidents;
• contracts or documents you do not own;
• a complete employment history when a smaller extract will do.

Replace names with roles, exact values with defensible ranges, and proprietary methods with a description of the capability. Use a work-approved tool for employer information. For personal accounts, work from a minimized summary.

Workplace data hygiene covers the underlying boundary in more detail.

Step 1: collect evidence moments

Write down eight to twelve moments from the last several years. Include paid work, volunteering, community activity, caring responsibilities, study, and substantial personal projects where relevant.

Look for moments when you:

• resolved an unusual problem;
• made a process more reliable;
• learned something difficult;
• coordinated people with different priorities;
• explained complexity clearly;
• recovered a failing piece of work;
• handled a customer or stakeholder concern;
• made a decision with incomplete information;
• built, repaired, organized, or delivered something;
• helped another person become capable.

Describe what happened before assigning a skill label.

Weak:

Strong communicator and strategic thinker.

Evidence:

Two suppliers interpreted the specification differently. I ran a joint review, turned twelve disputed points into an agreed acceptance checklist, and the next delivery passed without rework.

The second statement gives you material to examine. The first gives the model flattering words to repeat.

Step 2: build the capability-evidence matrix

Record one entry per evidence moment, with six fields each:

Suppliers disagreed on specification

• Action you personally took: facilitated review and wrote acceptance checklist
• Capability demonstrated: clarifying requirements; facilitation
• Observable result: next delivery accepted without rework
• Conditions: two suppliers; fixed deadline
• Evidence or verifier: checklist; project owner

New colleague struggled with monthly close

• Action you personally took: broke process into checks and coached two cycles
• Capability demonstrated: process design; instruction
• Observable result: colleague completed third cycle independently
• Conditions: existing finance system
• Evidence or verifier: runbook; colleague

Volunteer event lost its venue

• Action you personally took: compared alternatives and renegotiated schedule
• Capability demonstrated: contingency planning; negotiation
• Observable result: event ran on original date
• Conditions: small budget; 48 hours
• Evidence or verifier: revised plan; organizer

The conditions field prevents overclaiming. Coordinating five people for a community event is evidence of coordination; it is not proof that you can lead a global department.

Ask AI to extract cautiously:

From each evidence entry, propose:
1. one narrow capability directly supported by the action;
2. one broader capability that might be supported but needs more evidence;
3. the limitation or context that should remain visible.

Do not add achievements, scale, responsibility, or outcomes.
Flag vague claims and ask me for evidence.

Reject labels you cannot explain through the entry.

Step 3: create a capability inventory

Group repeated capabilities into four types:

1. Domain knowledge — what you know about a sector, customer, process, or regulation.
2. Methods — analysis, facilitation, research, quality control, planning, writing, selling.
3. Tools — software, equipment, languages, systems, and technical practices.
4. Operating qualities — work under uncertainty, attention to detail, conflict handling, persistence, judgement.

Do not discard domain knowledge merely because you want a new field. It may be the strongest bridge. A logistics specialist moving into software implementation may contribute more through process and customer knowledge than through beginner coding skill.

The EU’s Europass career-planning guidance recommends recording skills from work, learning, volunteering, and other achievements. Its purpose is communication and reflection, not automatic recognition. For standardized occupation and skill vocabulary, the European Commission’s ESCO classification can help you understand how occupations and skills are described across Europe.

Both resources were rechecked on 29 July 2026. Use their terminology as a translation aid, not as proof that you possess a skill.

Step 4: choose two target-role hypotheses

A target role is a hypothesis, not a new identity.

Define it with:

Target role:
Type of organization:
Problem the role is hired to solve:
Capabilities apparently required:
Evidence I already have:
Evidence I lack:
Questions that current job descriptions do not answer:

Choose two directions, not twenty. One should be an adjacent move where much of your evidence transfers. The other may be a larger change that interests you but requires a clearer test.

Research through:

• several current job descriptions from actual employers;
• professional or trade bodies;
• people doing the work;
• official descriptions for regulated occupations;
• training providers, checked against employer requirements;
• ESCO or another maintained occupational framework.

Do not infer demand from an AI answer. Do not treat a training provider’s sales page as neutral labour-market evidence.

For regulated professions, verify qualification and recognition requirements through the relevant authority. Europass itself notes that a profile can communicate qualifications but does not grant automatic recognition.

Step 5: run the gap test

For each target role, classify apparent requirements:

Separate three gaps:

• knowledge gap: you need to learn something;
• evidence gap: you may be capable but cannot yet demonstrate it;
• access gap: you need a person, environment, or opportunity to test the work.

Courses help some knowledge gaps. They do not automatically solve evidence or access gaps.

Step 6: choose two transition experiments

A good experiment produces information even if you decide against the field.

Options include:

• interview three people who currently do the work;
• complete a realistic, non-confidential work sample;
• shadow the work for half a day;
• volunteer for one bounded project;
• take one module and test the skill before enrolling in a full programme;
• ask a knowledgeable practitioner to review your evidence matrix;
• apply to a small number of roles and record where screening stops;
• test an adjacent responsibility in your current organization.

Write the decision rule before starting:

Hypothesis:
Experiment:
Evidence I expect:
Cost and deadline:
What would support continuing:
What would make me stop or revise:
Person who can challenge my interpretation:

Do not make “felt excited” the only success criterion. Interest matters, but so do aptitude evidence, working conditions, access, compensation, and the actual daily tasks.

Use AI as an interviewer, not an oracle

Ask the model to find weak evidence:

Act as a skeptical career-evidence editor.

For this target role and capability matrix:
- distinguish direct, adjacent, and missing evidence;
- quote the evidence entry supporting each conclusion;
- flag claims that depend on an assumption;
- propose three questions for a person currently doing the role;
- propose one low-cost work-sample test.

Do not predict employability or recommend a career.
Do not invent labour-market demand.

Then validate its interpretation with a person who knows the work.

The honest limit

Past evidence does not guarantee future fit. Hiring depends on location, language, networks, timing, credentials, discrimination, employer preferences, and circumstances no matrix can model.

The process also cannot choose what you value. Higher pay, stability, autonomy, status, craft, social contribution, and time outside work may point in different directions.

What it can do is replace an identity-level guess with a testable transition:

• real episodes instead of adjectives;
• bounded capabilities instead of inflated titles;
• two hypotheses instead of an endless possibility list;
• small experiments instead of an expensive declaration.

Use the career transition evidence matrix to fill in your own evidence rows, capability inventory, target-role hypotheses, and gap test in one place, rather than scattering them across separate documents and a chat history you may not keep.

Use AI job-hunting guidance after you have chosen a direction. Before that, build the evidence and let reality—not a fluent recommendation—narrow the path.

If the disruption behind this transition was involuntary — a redundancy, a company closing, a contract ending — after redundancy or mid-life disruption covers the administrative and emotional groundwork worth handling first, before this evidence-mapping process is the right next step.