# Career Internal Visibility > Use when building standing inside a current employer rather than in the external market — setting goals that survive the review, capturing evidence through the year instead of reconstructing it in December, writing the mid-year and year-end self-assessment, maintaining an internal talent or skills profile in systems like Eightfold that now decide which opportunities surface to you, using Confluence documentation and internal blog or news posts as deliberate visibility, assembling a promotion packet from evidence already recorded, choosing 360 reviewers for rubric coverage and writing feedback about others, and surfacing real scope when the HR record has said the same title for years. - Skill: `joogy06/career-internal-visibility` (Agent Skill) - Install (CLI): `npx skillmds@latest add joogy06/career-internal-visibility` - Raw SKILL.md: https://api.skillmd.com/api/skills/joogy06/career-internal-visibility/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Docs & Writing - Author: joogy06 (https://skillmd.com/u/joogy06) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/joogy06/career-internal-visibility --- # Career: internal visibility **Verified 2026-07-29.** Volatile figures live in `career-coach/references/market-snapshot-2026-06.md`, per family convention — this file carries the qualitative claim. ## 1. The asymmetry that governs everything here **Your external brand is discovered. Your internal standing is decided in rooms you are not in.** Ratings and promotions are settled in calibration, by people who mostly have not watched you work. What reaches that room is a manager's summary, a rating history, and whatever a sponsor can recall under time pressure — see `career-leadership` on sponsorship, which is the other half of this. **Two consequences, and both are practical:** 1. **Your evidence has to survive being retold by someone else.** A contribution nobody can state in one sentence with a number attached does not travel. 2. **The person writing the summary is working from your text.** In practice, most managers draft from the self-assessment. **Whatever you fail to write down, they will not invent** — that makes §4 the highest-leverage writing you do all year. **This is not self-promotion, and framing it as such is how good engineers lose promotions.** It is making real work legible to people who did not see it. ## 2. The annual rhythm | When | The work | Failure if skipped | |---|---|---| | **Cycle start** | Set goals that are checkable (§3) | You are measured against goals someone else remembers differently | | **Continuously** | Capture evidence as it happens (§4) | December reconstruction — the biggest and most common loss | | **Mid-year** | Course-correct while it still counts | Discovering a gap when nothing can be done about it | | **Year-end** | Write the summary from the record | Writing from memory, which systematically undersells | | **When ready** | Assemble the promotion packet (§7) | A case built in three weeks against a bar built over years | **The cycle is compressing.** Most large organisations now run mid-year or quarterly conversations alongside — or instead of — a single annual appraisal. **More checkpoints means more chances to correct, and more occasions where an unwritten contribution silently fails to count.** ## 3. Goals that survive the review **Write a goal so that whether you met it is not a matter of opinion.** The test: could a reasonable person who dislikes you agree it was achieved? **The distinction worth internalising: measure what CHANGED, not what got SHIPPED.** "Delivered the migration" is an activity. "Cut month-end close from 6 days to 2" is a result. Activity goals reward motion and are impossible to argue with in your favour, because the achievement is invisible. - **Name the baseline at the start.** A goal that improves something has to say what it was before, and nobody will accept a baseline reconstructed after the fact. - **Write the evidence line while writing the goal** — *how* it will be shown, and where the number will come from. A goal whose evidence source does not exist yet is a goal you cannot prove. - **Include one goal that is visibly bigger than your grade.** Promotion cases are built on operating at the next level *before* the title, and that has to be visible in the written goals or it reads as scope creep rather than readiness. - **Get goals written down in the system**, not agreed in a corridor. A manager change mid-year is common, and the new one inherits the record, not the conversation. ## 4. Evidence capture — the brag document **Keep a running record, updated when things happen.** Not weekly if that fails; but at every completion, incident and piece of praise. ``` ## — Problem: what was wrong, and who it hurt Did: MY specific contribution, distinct from the team's Result: the number, and the baseline it moved from Evidence: link — dashboard, PR, doc, ticket, email, meeting note Who saw it: names — who can corroborate this in a calibration room Level: which behaviour at which grade this demonstrates ``` **Why reconstruction fails, concretely:** you forget the mid-year work entirely by December — recency bias is brutal and it is *your own* memory it distorts first. Dashboards get rebuilt, tickets get archived, the colleague who could corroborate it leaves. **The evidence decays faster than the memory, and the memory decays fast.** - **Capture praise verbatim, with the sender and date.** A quoted line from a stakeholder outweighs a paragraph of your own framing. - **Record what broke and what you did about it.** Incidents you handled are among the strongest evidence of seniority, and the only people who never cite them are the ones who did not keep notes. - **`Who saw it` is the field people skip and the one that converts.** Evidence nobody can confirm is an assertion. **Keep it somewhere you own and can take with you** — the same record feeds the CV (`career-application-writer`) and the interview stories (`career-storytelling`). Respect confidentiality: figures, client names and internal metrics do not leave with you. ## 5. Writing the self-assessment **Specific, behaviour-based, forward-looking.** Name what happened, why it mattered, and what comes next. - **Numbers, not adjectives.** "Significantly improved reliability" is unquotable in a calibration room. "Cut P1 incidents from 14 to 3 across two quarters" survives being repeated by someone else. - **Lead each item with impact, not chronology.** The reader is skimming; the first clause is the one that lands. - **Attribute honestly and claim precisely.** "I led" where you led, "I contributed to" where you contributed. Over-claiming on something a peer also reports is the fastest way to lose a manager's trust, and it is checkable. - **Answer the growth question with a real gap and what you did about it.** A non-answer reads as either no self-awareness or no ambition; a real gap with a documented response reads as senior. - **Mid-year is for course-correction, not summary.** State where you are against each goal, what has changed, and what you will do with the remaining time — and **surface a slipping goal at mid-year rather than at year-end**, when it is still recoverable and raising it still reads as ownership. **Write it from the brag document, in one sitting, and cut.** Assembling from a record takes an hour; assembling from memory takes a day and produces less. ## 6. The internal skills profile — a system, not a formality Large employers increasingly run **internal talent marketplaces** — Eightfold and its equivalents — where your profile is matched against internal roles, gig projects, mentoring and learning paths. **This changed what the profile is for.** It used to be an HR record. It is now an **input to a matching model** that decides which opportunities are surfaced to you at all. How it behaves, and what follows: | Behaviour | What to do about it | |---|---| | Matching is **semantic**, scoring adjacent skills and career-progression patterns rather than keywords | Describe capability in normal language; keyword-stuffing does not help and can hurt | | It infers **potential**, not just history | Record skills you have used, even outside your job description — side projects, incident response, mentoring | | **Incompleteness reads as absence** | An empty profile is not neutral. It is a profile with no skills | | It surfaces **gig and project work**, not only vacancies | Short internal projects are the cheapest way to acquire evidence for a role you do not yet hold | | Aspiration fields feed the matching | State the roles you want. Opportunities you never say you want are opportunities you are not shown | **Update it on the same rhythm as the brag document**, not when you are already looking — a profile edited the week a role opens is visible as such, and by then the matching has already run. **One thing to know rather than to act on:** these platforms also model attrition risk from engagement, progression and compensation patterns. That is how they are sold. It is not a reason to game the profile — it is a reason to understand that the system is inferential, and to make sure what it infers about your skills is accurate. ## 7. Internal writing as deliberate visibility **A Confluence page or an internal blog post is the only artefact that keeps working while you are not in the room.** It is read by people who will never attend your demo, and it is quotable months later. **The document that gets cited beats the document that is thorough.** Optimise for being *linkable*: - **Write the thing people keep asking about.** The runbook everyone reconstructs from Slack, the decision record for the argument that recurs quarterly, the onboarding page that saves a week. The test is whether someone links to it without you. - **Put your name and the date on it, and keep it current.** An owned, maintained page is a reputation asset; an abandoned one is evidence you do not finish things. - **Decision records travel furthest.** *What we chose, what we rejected, and why* is the document people return to — and it demonstrates judgement, which is exactly what promotion bars ask about and what task lists cannot show. - **Internal blog and news posts reach sideways and upwards**, across the organisation, which is where sponsors are. A post-incident write-up, a "what we learned shipping X", a plain-language explanation of something your area does that others depend on. - **Write for the reader two levels up and one department across.** Jargon-dense internal writing is invisible to precisely the audience that matters for progression. **Then cite your own writing in the self-assessment.** A link is verifiable evidence in a way a claim is not — this is the cheapest conversion of work into recorded impact available. Use `confluence-content-creator` for the mechanics of the page, `human-voice-writing` for keeping the prose yours. **Getting a specific thing in front of a specific person** — a proposal, a demo, a deck — is `career-advocacy`: pre-wiring the decision, the internal channels, and how to make the ask. ## 8. The promotion packet **Assemble, do not compose.** If §4 has been running, this is an editing job. 1. **State the case in one sentence** — the level, and the claim that you already operate there. 2. **Map evidence to the published criteria**, in their language and their order. The committee reads against a rubric; making them translate is making them work against you. 3. **Three to five deep examples beat fifteen shallow ones.** Depth is what survives challenge. 4. **Show scope and duration** — sustained operation at the level, not one visible quarter. 5. **Name corroborators** and give them notice, with the specifics they can confirm. 6. **Address the obvious objection yourself.** Every case has one; leaving it for the room means someone else frames it. **The bar is "already operating at that level", not "ready to".** Cases fail on evidence of scope far more often than on quality of work. ## 9. Targeting a senior AI or executive technical role For a Head of AI, senior AI consultant or CTO-class role, **what committees test is narrower and harder than the job description implies.** They ask you to walk through a real programme you owned end to end — *including the thing that broke, and what it cost* — and they push on governance, decisions you reversed, and how you funded the work against sceptical stakeholders. Three assets recur, and the strongest candidates have all three: - **Technical judgement** deep enough to earn engineers' trust — architecture and data, not vendor selection. - **Organisational design** — building and running the team, not only the system. - **Translation into revenue, risk and velocity** — the ability to state what the work was worth in the terms an executive committee uses. **The internal implication is the actionable part**: none of that can be assembled at application time. **Own something end to end, keep the record of what broke, and be able to state the commercial effect in one sentence.** Sitting in the AI-adjacent parts of a large programme produces a CV that reads senior and an interview that does not survive the second question. **Say the target out loud, internally.** Sponsors cannot advocate for an ambition they have not heard, and the internal profile (§6) will not surface AI-leadership opportunities to someone who never declared the interest. External positioning for the same target is `career-positioning`; public credibility is `career-online-presence`. ## 10. When the HR record undersells what you actually did **The situation:** you have built things, run a business, operated well above your grade — and the official record says the same title for years. Every system downstream reads the title. **Understand what kind of problem this is.** It is **not** that people underestimate you. It is that **the title is the only machine-readable fact about you**, and semantic matchers, recruiter filters, internal talent systems and calibration rooms all anchor to it. Arguing harder does not help. **You change it by attaching evidence to the record, and by changing the record itself.** ### The rule that keeps this honest **Never state a title you did not hold. Always state the scope you did hold.** Claiming an unheld title is checkable — background and reference checks verify title and dates routinely — and being caught ends the process and the relationship. Describing real scope accurately is not inflation; it is the correction. Everything below stays on the right side of that line. ### Make scope legible independently of title **Numbers do the work a title will not.** A reader cannot calibrate "Senior Engineer", but they can calibrate the scope: - **Budget, headcount, systems, users, revenue, geography, regulatory exposure.** "Owned the £Xm reconciliation platform, 40k daily users, 3 jurisdictions" is unambiguous at any title. - **Decision rights.** What you signed off, what could not ship without you, who escalated to you. Decision rights are the truest signal of level and almost nobody writes them down. - **End-to-end ownership of something named.** "Led from proposal to production" beats six years of "senior engineer" because it has a boundary, a start and an outcome. - **What broke and what it cost.** Owning an incident is seniority evidence a title cannot convey — and it is exactly what senior hiring probes (§9). **Use the dual-line format** the family already applies to grade/title mismatches: `Official Title — functional scope` (e.g. *"Vice President — Head of AI Engineering"*). Both are true, the official one is verifiable, and the functional one is what gets matched. ### Show progression inside a static title **A long unchanged title reads as stagnation unless you show the scope curve.** Break the tenure into dated phases rather than one undifferentiated block: what you owned in year one, year three, year six. **The growth was real; if it is not on the page, the reader assumes there was none.** ### The entrepreneurial work **List it as a real role, with real dates, real scope and real outcomes** — not as an interest. A company you built is the strongest available evidence of operating at a level your employed title does not reflect. **Two honest cautions before you publish it:** - **Check your employment contract** for moonlighting, outside-business-interest and IP-assignment clauses, and any regulated-employee declaration duty. In regulated sectors an undeclared outside interest is a compliance matter, not a preference. - **Publishing it is disclosure.** Deliberate, not accidental — the same decision as the two open-to-work modes. **State outcomes plainly, including the modest ones.** A small business run properly for three years is credible; an implied unicorn is not, and the first specific question exposes it. ### Corroboration converts assertion into fact **Line up people who will confirm the scope** — the executive who escalated to you, the client who signed the work, the peer who reported into your programme. Give them the specifics you want confirmed. An assertion that a named senior person will back is a different object from a claim on a page (§4, `Who saw it`). ### Fix the record itself — this is the highest-value move Everything above is compensation for a record that is wrong. **Where possible, correct it:** - **Request a title or levelling change with evidence.** Most job architectures have a defined route, and most people never use it. **One title change outperforms years of explaining.** - **Get acting or interim responsibilities documented** at the time, in writing, from whoever assigned them. Undocumented acting-up is invisible six months later and impossible to evidence at exit. - **Put the real scope in the internal skills profile** (§6), which is not constrained by the HR title and is what internal matching actually reads. - **Use a secondment or internal move** to reset the title where the current line will not — a lateral move with a correct title is often faster than a promotion into a wrong one (`career-transition` §4). **Start this while you are still in the role.** Every one of these is far easier to obtain from people who currently work with you than to reconstruct after you have left. ## 11. The 360 — choosing reviewers, and writing feedback for others ### Choosing who to ask is a coverage problem, not a popularity one **Pick reviewers so that, between them, they can evidence every competency you are rated against.** Most people pick the colleagues they like, get warm and useless feedback, and cover half the rubric. Work backwards from the criteria: | Include | Because | |---|---| | **Someone senior who saw you operate at the next level** | The scope evidence a peer cannot give | | **A peer from another team or function** | Cross-boundary influence — usually the thinnest evidence in a packet | | **A stakeholder you delivered *to***, not with | Outcome and reliability, from the receiving end | | **Someone who saw you handle something difficult** | Incidents, disagreement, pressure — where seniority actually shows | | **A junior you mentored**, where the rubric includes it | Direct evidence for a criterion self-assessments cannot cover | **An all-friendly panel is visible as one and it reads as weak.** Uniformly glowing feedback from people who all sit next to you tells a calibration room nothing, and experienced readers discount it. **Include someone who will say something true and mildly uncomfortable** — a genuine development point in an otherwise strong 360 makes the whole set credible. - **Cover the *period*, not just the present.** If your strongest work was in the first half of the year, include someone who was there for it. - **Watch the reciprocity read.** Mutually-exchanged glowing 360s between two people are noticed. - **More reviewers is not better.** A focused set that covers the rubric beats a long list of thin responses. ### Brief them — this is the step almost everyone skips **Ask for something specific.** A generic request gets a generic answer, which is worth nothing. > "Would you cover the migration programme — particularly the point where we lost the vendor and > had to re-plan in two weeks? That is the example I am using for handling ambiguity." That is not putting words in their mouth; it is telling them **which of the many things you did together is the one you need described.** They cannot know that, and left to guess they will write about whatever is most recent. **Give them your dates, the artefacts and enough notice** — a reviewer writing in a rush produces vague text that damages you more than silence would. ### Writing feedback about others — and what it says about you **What you write about colleagues is read by their manager, and often by people well above you. It is a visibility surface, and almost nobody treats it as one.** Sloppy, generic feedback marks you as someone who does not observe carefully. Precise, fair, well-judged feedback marks you as someone with judgement about people — which is exactly what promotion into leadership tests. - **Behaviour and evidence, not personality.** "Rewrote the rollback plan after the first failed deploy, which is why the second attempt was uneventful" — not "is proactive". - **Be specific about impact on *you* or the work.** First-hand observation is the thing only you can contribute. - **Give a real development point, kindly and usefully.** Blank or "nothing comes to mind" is a wasted opportunity for them and reads as disengagement from you. - **Never raise something you have not already said to their face.** A 360 is not the venue for a first disclosure, and being the person who does that is remembered. - **Write about what they did, not what they are.** Attributed traits are unactionable and, where they track demographics, are where bias enters 360 processes. - **If you genuinely cannot help, decline early** rather than filing something thin. **Do it properly and it compounds**: people ask reviewers who write well, which puts you in more rooms, in front of more managers, saying useful things about their people. ## 12. Anti-patterns - **Reconstructing the year in December.** The single largest, most common loss. - **Assuming good work is self-evident** to people who did not see it. - **Activity goals** — "delivered X" with no baseline and no changed number. - **Agreeing goals verbally** and never getting them into the system. - **Waiting until year-end to surface a slipping goal**, when it is no longer recoverable. - **Adjectives instead of numbers** in a self-assessment. - **Over-claiming shared work** — checkable, and it costs more than it gains. - **An empty or stale internal skills profile**, which reads as absence rather than neutrality. - **Updating the profile only when you are already looking.** - **Never stating the role you want**, then not being shown it. - **Writing internal documentation nobody links to**, or abandoning what you wrote. - **Not citing your own writing** in the review, leaving verifiable evidence unclaimed. - **Building the promotion case in three weeks** against a bar measured in years. - **Claiming an AI-leadership target** without owning a programme end to end. - **Stating a title you did not hold** — checkable, and it ends the process. - **Arguing that a title undersells you** instead of attaching scope evidence to the record. - **Six years shown as one undifferentiated block**, so real growth reads as stagnation. - **Leaving acting-up undocumented** until you need to evidence it. - **Publishing outside business interests** without checking the contract and declaration duty. - **Never requesting a levelling review**, which most architectures provide and few people use. - **Choosing 360 reviewers by warmth** rather than by rubric coverage. - **An all-friendly 360 panel**, which experienced readers discount. - **Sending a generic 360 request** and getting a generic answer. - **Raising something in a 360** you never said to the person directly. - **Writing thin feedback about others**, which is read by people senior to you.