Career Memory: the profile every JobMentis tool reads before it writes
Career Memory is the profile every AI feature in JobMentis reads before it writes anything. 16 compact blocks of your real career data, sorted into four tiers by where each fact came from, refreshed the moment the underlying data changes rather than on a timer.
Why a general chatbot cannot do this
A chatbot starts every conversation from zero, so you re-explain yourself every time and it re-invents the parts you leave out. Career Memory is the opposite arrangement: your material is stored once, in a form built for a model to read, and every tool opens on it. A cover letter, a mock interview and a career report all begin from the same picture of you, and none of them begins from a blank window.
The practical difference
Write a STAR story once and it shows up in your next CV rewrite, your mock interviewer already knows it, and the fit engine counts it as evidence. Nothing you do here is a dead end, which is the part a stack of separate tools cannot reproduce however good each tool is.
What is actually in it
16 blocks. Each one is a short, dense, machine-readable summary of one part of your material, and a tool asks for only the blocks it needs. Empty blocks are left out entirely, so an unfinished profile costs you nothing but reach.
Attested
10Your own data, serialized deterministically. No model sits between the fact and the block.
- Workfive DNA
- Your work-style profile across 30 measured facets.
- Career history
- Roles, employers, dates and what you actually did.
- Career archetype
- The archetype derived from your CV, plus the facet shape behind it.
- 360 feedback
- What colleagues said when you collected feedback, including the role they would put you in.
- LinkedIn positioning
- Your headline, About and skills, kept separate from anything we generated for you.
- Story Bank index
- Every STAR story you have written, indexed by the competency it evidences.
- Identity
- Your values, vision, mission and purpose, in your own words.
- Outcomes
- Your active goals and key results, with progress.
- Content themes
- The subjects you have decided to post about.
- Confirmed facts
- Claims you attested as true when a check could not find them in your CV.
Coach-set
1Set by a human who knows you, so it outranks anything a model inferred. Present only if you work with a coach.
- Coach plan
- The programme, tracks and open tasks your coach assigned. Client-visible material only.
Reviewed draft
1Drafted by AI, then yours to correct. Each entry is marked as edited by you or still untouched.
- Application answers
- The free-text answers you have given employer application forms.
AI-derived
4Produced by another feature reading your data, never raw input. Each one carries a visible provenance stamp so downstream tools weight it below anything attested.
- Interview insights
- Recurring strengths and concerns across your real and mock interviews, plus the questions you were actually asked.
- Upskill gaps
- Skill themes that keep recurring across the jobs you track.
- Market worth
- Your benchmarked compensation range.
- Disruption risk
- How exposed your role is to automation.
Every fact carries where it came from
The four tiers are a ranking, not a filing system. When two blocks disagree, the higher tier wins, and the prompt reading them is told so explicitly rather than left to guess.
- 1Coach-setA human decided this. Where the coach's plan conflicts with anything inferred from other blocks, the instruction is to follow the coach.
- 2AttestedYou supplied it or measured it. The block is generated by code, not by a model, so it cannot drift on the way in.
- 3Reviewed draftWe drafted it, you may have corrected it. An edited entry is your voice, an untouched one is still ours, and the block says which.
- 4AI-derivedA model read your data and concluded something. Stamped with its source and date, and deliberately ranked last.
That last line is the one most systems skip. A conclusion a model drew about you six weeks ago is not the same class of fact as your employment dates, and treating them as equals is how confident nonsense gets written into a cover letter.
It refreshes on change, not on a timer
Every write that changes your material drops the blocks it affected, so the next tool that asks rebuilds them. There is no refresh interval to sit through and no window where a tool quotes something you already replaced.
Edit it, and it is already gone
Saving an Identity answer, finishing an assessment or confirming a fact clears the affected block on the way out. The next read rebuilds from what you just wrote.
Almost nothing is cached
Only the Story Bank index is stored between reads, because building it means one database pass plus a read per story. The rest is recomputed every time it is asked for.
Expiries are a backstop, never the plan
The one cached summary carries a 30-day expiry. That exists so a missed invalidation self-corrects, not as the mechanism that keeps you current.
The Master Brief: the capstone
On top of the 16 blocks sits one short narrative, three to five lines answering who this person is. It is composed from the blocks rather than written from scratch, cached so repeated use is cheap, and dropped the moment any source block changes.
- It is what a live voice mock interviewer is given as opening context, which is why the interviewer already knows your history before the first question.
- It is what your coach reads at the top of your client page, with the date it was built shown and a button to rebuild it.
- It is written in English even when your interface is not, because its readers are an English-only coach console and a system prompt whose speaking language is set separately.
Where it is deliberately not used
A memory that leaks into everything is a liability. Four places refuse it on purpose, and each refusal buys something specific.
Your CV audit score
The audit never reads Career Memory. A score has to be reproducible from the CV alone, so re-auditing an unchanged file returns the same number and a small edit shows up as a real change rather than a random walk.
Coach surveys
Survey answers never enter anyone's profile. A survey is about the programme, not the person, and on an anonymous one, folding answers back into an individual would undo the promise that nobody can trace them.
Your coach's private notes
Only material your coach shares with you reaches the coach block. Their private scratchpad stays private and never becomes context a tool writes from.
Voice rehearsal partners
A networking rehearsal is briefed with an anonymised generic role, never the real name or company of the person you are about to meet.
What unlocks as it fills
The memory is not a form to complete. Each addition is worth something on its own, and the gates below open as a side effect of doing the work.
- Upload your CVCareer archetype, CV audit, and your first optimized assets
- Add and rate 3 jobsYour Light career report generates itself
- Complete WorkfiveMeasured work-style DNA and Pathfinder eligibility
- Write 2 or more STAR storiesFull career report and evidence-backed rewrites
- Collect 360 feedbackAdvanced report with blind-spot analysis
Career Memory against a general assistant
Both will write you a cover letter. The difference is what each one knows and what happens when it is wrong.
| Dimension | General AI assistant | JobMentis |
|---|---|---|
| Starting point | An empty window. You paste your context again, or it works without it. | 16 blocks of your material, loaded before the first word is generated. |
| What it retains | Chat-level recall of things you happened to mention, with no structure. | Structured blocks per domain, each one asked for by name by the tools that need it. |
| Where facts rank | Everything said in the thread carries the same weight. | Four tiers. A coach instruction beats your data beats a model's conclusion. |
| When it invents something | Nothing checks it. The invention reads as well as the truth. | A groundedness review flags claims your real CV cannot support, and mutated numbers are reverted automatically. |
| Staying current | You correct it again in the next conversation. | Correct it once. The affected block is dropped and every tool sees the new version. |
Career Memory by the numbers
One profile, read through one entry point
16
blocks of your own material, each requested by name by the tools that need it
4
provenance tiers, so a model's conclusion never outranks a fact you attested
0
guessed refresh timers. A block is dropped the moment its source changes
It is your data, and it leaves when you do
Career Memory holds nothing you did not give us. Every store behind it is registered in our deletion manifest, which means a deletion request removes it rather than orphaning it, and the removal is verified afterwards instead of assumed.
- Your material is scoped to your account, and an organisation client can be placed in a separate database and a separate identity provider entirely.
- Text generation runs in an EU region.
- Deleting your account deletes the blocks, the cached brief and the material underneath them.
Career Memory questions
What is Career Memory?
Career Memory is the persistent profile every AI feature in JobMentis reads before it generates anything. It holds 16 compact blocks built from your own material, from your CV and career history to your work-style assessment, STAR stories, goals and interview feedback, and it is refreshed whenever the underlying data changes.
Is this the same as ChatGPT's memory?
No. A chat assistant's memory is a loose set of things you happened to mention, held at conversation level and weighted equally. Career Memory is structured per domain, requested block by block by the tool that needs it, and ranked in four tiers so a conclusion a model drew never outranks a fact you supplied or an instruction from your coach.
Does every tool see everything?
No. A tool asks for the blocks it needs and gets only those. A cover letter does not need your compensation benchmark, and a salary estimate does not need your content themes. The only surface given the whole picture is the short Master Brief used where the question genuinely is who this person is.
Can I see and change what is in it?
Yes, because every block is built from material you can edit in the product. Change a role in your career history, rewrite an Identity answer or correct an application answer, and the affected block is dropped immediately so the next tool rebuilds it from the new version.
What happens when the AI gets something wrong about me?
Two mechanisms. A groundedness review reads generated CVs and cover letters against your actual source material and flags claims it cannot trace, and any number the model changed is reverted automatically before you see it. If the flagged claim is true but simply not on your CV, you confirm it once and it becomes an attested fact that every later tool treats as verified.
Does my coach see my Career Memory?
A coach you have connected to sees the Master Brief summary on your client page, with the date it was built. The reverse is not symmetrical: what your coach assigns you becomes part of your memory, but their private notes about you never do.
Where is it stored, and can I delete it?
It is stored against your account, an organisation client can be placed in entirely separate infrastructure, and text generation runs in an EU region. Every store behind Career Memory is listed in our deletion manifest, so deleting your account removes the blocks, the cached summary and the source material, and the deletion is verified after it runs.
Your memory starts with one upload
Upload a CV and three blocks fill immediately: your career history, your archetype and your first audit. Everything after that compounds.
Start freeCurious how it compares to running prompts by hand? Our career prompts and Claude skills are free and open source