How to Improve Your CV and Professional Profile with Super Intelligence

A good CV improvement makes your real work easier to understand. It does not give you a more impressive past. Super Intelligence can help you find relevant evidence, organise a draft, test a claim and adapt the emphasis for a different role. The useful result is a document you can explain, with a professional profile that tells the same underlying story.

In this series, Super Intelligence, or SI, is a practical umbrella for working with AI tools and connected workflows. It does not mean that today’s writing assistants are proven artificial superintelligence. This guide uses ordinary drafting and comparison tasks that you can also complete manually. You retain the decisions about your history, your privacy and what you are willing to claim.

This is a hands-on adult career-writing workshop. You will work from a complete fictional source packet, inspect a deliberately flawed AI-style draft, repair it, and see a finished CV and public profile. A later exercise changes the target role and adds new evidence. All people, employers, vacancies, qualifications, records and outcomes in those exercises are invented teaching material. No application is made, and no example describes the reader’s real achievements.

For the broader principles of selecting career evidence, see Writing a CV and Professional Profile. This guide concentrates on the additional work required when SI helps transform a source record into an application: controlling what enters the tool, preserving the meaning of claims, checking generated changes and maintaining consistent versions.

1. Decide what the document must help a reader understand

Before asking SI to improve anything, finish a plain sentence: “This version should help a reader see that I can do these particular tasks.” The tasks might include preparing weekly reports, handling service exceptions, coordinating a project schedule or producing a design. A job title alone is not enough. Two employers can use the same title for different work, and a candidate can have relevant experience under a different title.

Your document has three related purposes. It should make relevant experience visible, make the limits of that experience understandable, and make a next conversation possible. It cannot prove every aspect of your capability, settle an employer’s decision or guarantee an interview. A model’s enthusiastic description of your suitability is not independent evidence that you meet the role’s requirements.

Separate the application CV from the public profile. The CV is selected for a known opportunity and may contain relevant detail that you would not publish broadly. A public profile needs a more stable description of your work and direction. It may be seen by people who do not know which job you want. The same dates, employment titles and qualifications must remain true on both surfaces, even when their introductions differ.

Also separate the private working record from both finished documents. The working record can contain questions, alternative wording, source references and facts that do not belong in an application. It is where you resolve uncertainty. A final CV should not quietly inherit an unanswered question simply because the model converted it into a fluent sentence. “Possibly helped with forecasting” is a question to investigate, not permission to list forecasting expertise.

A useful starting target is small: one verified summary, three strong experience statements and a consistent headline. You can then build the rest around that core. Trying to create an entire professional identity in one prompt makes it harder to notice where the machine has supplied missing substance. The aim is not to sound like an ideal applicant. It is to let a reader inspect the strongest relevant version of your actual record.

Back to contents · Next: 2. Capture a private evidence record before asking for better prose

2. Capture a private evidence record before asking for better prose

Start with records you are entitled to use. Employment dates, official job titles, course completion documents, personal project notes and permitted descriptions of work can help. An old CV is useful as a list of claims to check, but it is not automatically the best source. It may already contain a rounded date, an inflated verb or a metric whose meaning you no longer remember.

For each important piece of work, write what happened, what you personally did, the period, the output and how you know. Add who else contributed and what you did not control. These details stop a team result becoming a personal result during compression. “Maintained the issue log for a four-person pilot” contains a different responsibility from “managed a four-person team”. The difference remains important even if both sentences sound professionally confident.

Give each source a short identifier such as R1 or R2. The identifier is a private editing aid, not a requirement to expose internal documents to an employer. A claim can then carry its support while you draft: “Prepared twenty weekly reports, R4.” Later you remove the editing labels from the clean CV while retaining the private mapping. This makes a future correction much easier than reconstructing the origin of every sentence.

Distinguish a record from a recollection. A dated report can support its own existence. Your memory may support an approximate description, but it might not support an exact percentage or attendance count. Ask yourself what precision the evidence can carry. If a useful fact remains uncertain, leave it out, state an appropriate limitation where it belongs, or verify it through an authorised route. Do not ask SI to choose the most plausible number.

Capture worthwhile work without requiring a financial result. A clear procedure, a reconciled record, a difficult handoff or a completed work sample can demonstrate capability. If no reliable outcome measure exists, describe the action and product. This is not a weaker substitute for a fabricated metric. It is a more informative account of what you can actually substantiate.

You should finish this stage with a short collection of usable facts and a separate question list. If the question list is longer than the fact list, your next task is evidence gathering. Polishing should wait. SI can help ask focused questions, but only your real record and appropriate verification can answer them.

Back to contents · Next: 3. Keep the drafting packet smaller than your personal archive

3. Keep the drafting packet smaller than your personal archive

A writing tool rarely needs your full address, identification number, referees’ contact details or the raw files behind a workplace result. Begin with the least revealing packet that can support the editing task. You can replace an employer with “Employer A” during drafting and restore the correct name yourself in a local final document. Removing a name does not, by itself, make a distinctive project anonymous or make confidential material safe to share.

Singapore’s CSA and IMDA advise individuals to protect the information they share with generative AI, review the tool’s privacy conditions and verify generated content. Their July 2026 advisory specifically warns against providing workplace or confidential business information without proper authorisation. For this workflow, the practical consequence is to prepare a permitted summary rather than upload an entire employment folder. CSA and IMDA: safe and secure use of generative AI.

Think about three destinations separately: the AI tool, a specific application recipient and the public internet. Permission or suitability for one does not automatically transfer to another. A manager may have approved an aggregate figure for a private application without approving publication of an internal spreadsheet. A public profile can describe the method without repeating the figure. In the fictional packet below, the permitted uses are explicit so that you can practise making this distinction.

Inspect the actual service and account you would use. Its current data controls, connected apps, retention information and sharing settings matter. Do not assume that a paid account, an incognito browser window or a reassuring model response establishes confidentiality. A model cannot grant permission on behalf of your employer or a former client. If the terms or workplace rules are unclear, you can still work manually with a non-sensitive summary while resolving the uncertainty.

Keep other people’s information out of the packet when it is unnecessary. A colleague’s name is generally not needed to explain that you coordinated a handoff. An adult participant’s medical circumstances are not needed to describe scheduling a workshop. A reference can be represented as “a supervisor confirmed my contribution” without publishing the supervisor’s private contact details. The point is to show your work without turning other people’s records into application material.

Back to contents · Next: 4. Clarify the vacancy without treating it as a biography template

4. Clarify the vacancy without treating it as a biography template

Read the actual vacancy before asking SI to tailor a document. Extract the work to be performed, the stated essential requirements, the desirable requirements and the application instructions. Keep these categories separate. “Training provided” is different from “prior experience essential”; a desired tool is different from a mandatory current licence. An assistant that merges them may either discourage a plausible applicant or claim a qualification the applicant does not have.

MyCareersFuture’s résumé guidance recommends examining the job description and including relevant skills when you have them. That final condition is central. A vacancy tells you what the employer wants; it does not establish what you have done. Its wording can help you name genuine experience, but it cannot supply missing evidence. MyCareersFuture: analysing the role and tailoring a résumé.

Ask SI to return a requirement map before writing the CV. For each requirement, request the exact supporting passage from the vacancy, any matching source identifier from your packet, and one of four conclusions: direct evidence, transferable evidence, no supplied evidence or clarification needed. These are editing categories, not a numerical hiring score. A model-generated “92% match” can hide whether the missing eight per cent contains the only mandatory condition.

A useful map acknowledges context. Coordinating a weekly service report may support a reporting requirement in another industry, but it does not establish knowledge of that industry’s rules. Briefing colleagues on a checklist may support clear explanation, but it does not establish a teaching qualification or experience designing a formal assessment. Say what transfers and what still needs learning. That bridge is more convincing than a label that silently erases the difference.

Read the application instructions for the permitted use of AI, required declarations, file types, length and any assessment conditions. If an employer restricts assistance, follow the stated rule. If the rule is unclear and material, seek clarification rather than asking a model to interpret ambiguity in your favour. This guide is about preparing truthful documents, not finding ways to bypass an employer’s assessment conditions.

Keep a dated copy of the vacancy for your own records, where you may do so. If the advertisement changes, compare the new requirements against the old version before reusing the draft. A polished document targeted at yesterday’s role can be accurate and still miss today’s question.

Back to contents · Next: 5. The complete fictional career packet: Mira Chen

5. The complete fictional career packet: Mira Chen

Meet Mira Chen, an entirely fictional adult preparing a career document on 1 October 2026. Mira is considering service-operations coordination work. The packet gives everything needed for the main exercise. Do not infer gender, nationality, age, health, family circumstances, work authorisation or any other unstated personal characteristic from the name or the work history. None is needed to solve the writing task.

R1, verified chronology in the fictional record: Mira worked at Brindle Help as a Service Associate from January 2021 through December 2022. January and February 2023 were a career break; no reason is supplied or needed. Mira then worked at Cedar Trade as an Operations Coordinator from March 2023 through September 2026. Both were paid employee roles. Cedar’s role ended on 30 September 2026, so a CV prepared on 1 October must not label it “present”. No additional employment is supplied.

R2, Brindle role description and confirmed work: Mira recorded incoming delivery enquiries, checked order status in the team’s ticketing system, wrote customer updates and passed disputed refunds to a supervisor. Mira could correct a contact note but could not approve a refund or change delivery policy. There is no supported case-volume total, customer-satisfaction score or response-time improvement. The ticketing product’s name is not included, and no evidence of Salesforce use is supplied.

R3, Cedar responsibilities: Mira maintained a spreadsheet handoff log connecting warehouse, customer-support and dispatch colleagues. Mira checked entries for missing order references, followed up unresolved items and prepared a weekly summary for the operations supervisor. Mira had no direct reports, budget ownership, purchasing authority or responsibility for disciplinary decisions. Describing cross-team coordination is accurate; describing departmental management is not.

R4, a bounded reporting record: from 5 May to 15 September 2025, Mira prepared twenty scheduled weekly reports, all submitted by their agreed internal deadline. The preserved summaries show spreadsheet filters, SUMIF formulas and a basic pivot table used to group issue types. This supports those specific reporting tasks. It does not establish advanced Excel mastery, financial forecasting, SQL, Python, Power BI or automated dashboard development.

R5, the six-week handoff pilot: a four-person team trialled a new checklist from 7 July to 17 August 2025. Mira drafted the first checklist, recorded feedback and maintained the issue log. The operations supervisor approved the procedure; a warehouse colleague and a dispatch colleague helped revise and use it. Mira did not lead the team or implement the changes alone. The pilot concerned one service workflow, not the whole company.

R6, the pilot’s aggregate counts: the six-week baseline contained 400 eligible orders and 48 orders with at least one missing-reference handoff error. The six-week pilot contained 500 eligible orders and 35 orders with at least one such error. The definition was the same in both periods, and each order was counted once in each denominator. Other staffing and scheduling arrangements also changed. The record does not identify how much of the difference was caused by the checklist or by Mira.

R7, additional boundaries: a later July 2026 sample contained 240 eligible orders and 19 with the same error definition. This is a separate sample, not an additional part of the pilot. It does not establish a permanent seven-per-cent error rate. No recorded revenue, cost saving, profit change or labour-time saving is supplied. There is no award, promotion, customer testimonial or company-wide transformation result in the packet.

R8, peer briefing record: Mira delivered two one-hour briefings to four colleagues in total about using the handoff checklist. The supervisor reviewed the material. The record establishes two sessions, two delivery hours and four colleagues, not four training programmes, four direct reports or a formal trainer qualification. It contains no test of participants’ later performance.

R9, education and learning: Mira completed a Diploma in Business Operations at fictional Elm Learning College in 2020. In February 2025, Mira completed a fourteen-hour Spreadsheet Reporting Basics course at fictional Quarry Learning. The course issued a certificate of completion. Neither record establishes a degree, professional certification or licence. A separate Power BI course was begun in August 2026 but not completed; there is no supplied project or assessment result for that course.

R10, disclosure rules for this fictional exercise: Mira may include the real role titles, dates, employer names, R2–R5 task descriptions, R8 briefing details and completed R9 learning in a CV or public profile. The baseline and pilot aggregate figures in R6 may be used in a private application document with the team contribution and measurement limits preserved. The same de-identified R6 aggregate summary is explicitly permitted for this exercise’s drafting tool. R6 and R7 numerical results must not be placed on a public profile. Raw operational records, customer details, internal screenshots and colleague contact details must not be uploaded to a general AI tool or supplied in an application. D1 is the permitted drafting packet: de-identified chronology, task and learning summaries from R1–R6 and R8–R9; the voice preference from R11 only; and the R10/R12 rules. Replace person, employer and institution names with stable labels and omit contact details. R7 stays local-only for manual checking and must not enter the tool. Preserve source identifiers and all relevant limits in the summaries. Restore names and contact details locally after review.

R11, voice and contact: Mira prefers plain English, does not want the words “visionary”, “expert” or “transformational” without evidence, and wants a service-operations direction. For this educational example, the contact line is “mira.chen@example.com”; this is an illustrative address using the IANA example domain, not an application destination. No telephone number, live professional-profile link or referee contact is supplied. Mira can add verified real contact details when preparing a real document, but the model may not guess them.

R12, evidence status: the facts above are accepted inputs within this fictional exercise, not independently verified facts about a real person. There is no hidden attachment or missing dataset needed for the answer. Where the packet supplies no evidence, treat the claim as unsupported for this exercise. A real applicant could investigate a missing fact through an appropriate source; the fictional solver must not fill the gap with a plausible story.

Back to contents · Next: 6. The target role and an evidence-based fit map

6. The target role and an evidence-based fit map

J1 is an invented vacancy for Service Operations Coordinator at fictional North Quay Services. Its essential requirements are: at least two years of relevant service or operations experience; keeping accurate handoff records; preparing weekly spreadsheet summaries using filters and formulas; communicating clear customer updates; and escalating exceptions to the appropriate decision-maker. It does not require people management or budget ownership. Familiarity with a customer-record system is desirable, with training available on the employer’s system. Power BI is desirable, not essential.

The fictional application instructions request a concise CV in an ordinary readable document format, a short profile and truthful examples of relevant work. They permit AI-assisted editing if the applicant verifies the result. They do not require a photograph, identity document, salary history, references’ contact information or a portfolio upload at this stage. There is no guaranteed shortlist, promised salary or numerical match threshold. These are exercise instructions, not claims about a real employer.

The experience requirement has direct support in R1. Counting calendar months inclusively as listed, Brindle supplies twenty-four months and Cedar forty-three months, for sixty-seven months in the two roles. That is five years and seven months of listed paid experience, excluding the two-month career break. The clean summary can say “over five years” without implying uninterrupted employment or turning the date calculation into a seniority claim.

Handoff records match R3 and R5 directly. Spreadsheet summaries match R4 directly. Customer updates and appropriate escalation match R2 directly, supported by R3’s follow-up work. These are the strongest pieces of evidence for J1. The CV should show them early rather than opening with the unfinished Power BI course. A desired keyword should not displace a well-supported essential task.

The customer-record-system preference has bounded transferable support: R2 establishes use of an unnamed ticketing system. It does not establish knowledge of North Quay’s product. A suitable phrase is “recording customer enquiries and order-status updates in a ticketing system”. The Power BI preference has no demonstrated skill evidence. Beginning a course is not the same as completing it or using the tool competently. It can remain in the private learning plan instead of the CV.

This map supports preparing an application; it does not determine an employer’s decision. The source packet is narrow, and the real job might involve requirements not visible in a short advertisement. A useful next conversation could clarify the reporting system and the level of analytical responsibility. The CV should make the demonstrated match clear while leaving those questions open, rather than advertising certainty that the evidence cannot carry.

Back to contents · Next: 7. Give SI a bounded drafting job

7. Give SI a bounded drafting job

A useful instruction assigns a specific transformation rather than an identity: “Using only the permitted drafting packet D1 described in R10 and a de-identified copy of vacancy J1, propose a service-operations CV. Keep official titles and month-year dates unchanged. Preserve the difference between the candidate’s contribution and the team’s result. Put an evidence identifier beside each proposed experience statement. Do not add numbers, products, credentials or personal characteristics. Return unsupported requirements separately before producing a clean version.” This is an example instruction for the fictional packet, not a guarantee of compliant model behaviour.

Ask for the working output before the polished output. The working output should include the requirement map, proposed sentences and support. You can see whether the assistant has found R4’s concrete spreadsheet evidence or merely copied “analytical expert” from a pattern. If the model proposes a fact outside the packet, the correction is to remove or investigate it, not to accept its explanation that such wording is common in the industry.

Set the destination for each output. “Create a private application CV using the approved aggregate figures” and “Create a public profile without the restricted metrics” are different instructions. Asking for both without naming the distinction invites accidental transfer. In a real workflow, source summaries must themselves be permitted for the chosen tool before you provide them; a later reminder about confidentiality cannot undo an earlier disclosure.

Request a small set of alternatives when wording is genuinely uncertain. For example, ask for three ways to explain the checklist contribution without claiming leadership. Then choose the sentence that a former colleague would recognise. More alternatives are not necessarily more useful. Ten versions of the same unsupported claim still leave you with no supported sentence.

Keep the review separate from the generation. You can ask a second pass to identify numbers, responsibility verbs, qualifications and implied causal claims. However, agreement between two model passes does not authenticate your employment history. Both may repeat the same assumption. The decisive comparison remains between the proposed sentence and the source record you are entitled to rely on.

A short draft that preserves the facts is a good intermediate result. Do not interpret missing decorative language as a failure. The next stage is to improve order, clarity and voice around that reliable core. A fluent document can be repaired only if you can still see which parts came from the record and which parts came from the drafting process.

Back to contents · Next: 8. Inspect the deliberately flawed draft

8. Inspect the deliberately flawed draft

The following is a constructed bad example, not output from a measured model experiment. Its purpose is to make common kinds of drift visible. Imagine an assistant returns this professional summary: “Visionary operations manager with seven years of continuous experience, leading cross-functional teams and delivering company-wide transformation. Certified Excel expert and Power BI specialist who cut errors by 42%, saved $50,000 and achieved 100% customer satisfaction.” Every impressive phrase needs to be checked against the packet.

Its work-history section reads: “Operations Manager, Cedar Trade, January 2023–present. Managed four staff, independently designed and launched a company-wide automation programme, and reduced handoff errors from 12% to 7%, a five-per-cent reduction. Trained four teams through four workshops and delivered advanced automated dashboards.” The paragraph borrows several real fragments, changes their relationships and adds achievements that are not present.

Its earlier-role line reads: “Senior Service Consultant, Brindle Help, January 2020–December 2022. Expert Salesforce user who approved refunds and improved customer satisfaction to 100%.” Its education section reads: “Bachelor of Business Operations, Elm Learning College, 2020; professional Excel certification, 2025; Power BI certification, 2026.” The public-profile headline repeats “Operations leader | Power BI expert | $50k cost savings”, and the public About section repeats the internal pilot figures.

The bad draft is not entirely disconnected from reality. There was operations work, a four-person team, spreadsheet reporting and a change from twelve to seven per cent in the measured samples. That mixture makes it dangerous. A quick glance may recognise enough real detail to miss the changed title, dates, authority, causal attribution, unit of comparison and disclosure boundary.

Do not try to repair it by making every sentence slightly less enthusiastic. “Experienced operations manager” still changes Mira’s title; “helped save $50,000” still invents a saving. A correction must address the unsupported proposition, not just the adjective. First identify the exact claim. Then decide whether to retain it, narrow it, replace it with another supported fact, or remove it entirely.

The useful question is not whether the draft sounds like a CV. It is whether each part communicates what the source record establishes. A strong final document may be less dramatic than this draft and much more informative. The exercise succeeds when it makes the reader’s real capability easier to inspect without requiring a new fictional career to support it.

Back to contents · Next: 9. Repair titles, dates, authority and learning status

9. Repair titles, dates, authority and learning status

Repair F1, the experience summary: “seven years of continuous experience” becomes “over five years of service and operations experience”. R1 supplies sixty-seven listed months, not eighty-four, and includes a break. We use the modest rounded phrase in the clean summary because it is accurate and easy to read. There is no need to publish a month-by-month calculation, but the private check prevents an attractive number from becoming a false claim.

Repair F2, Cedar’s title and dates: “Operations Manager, January 2023–present” becomes “Operations Coordinator, March 2023–September 2026”. All three components matter. “Coordinator” is the recorded title. The employment began after the break and ended before the document date. If a real employer used an unusual title, you could add a factual functional explanation, but you should not silently replace it with a higher-status title.

Repair F3, people management: remove “managed four staff”. R5 identifies a four-person project team, not four direct reports. R3 explicitly states that Mira had no direct reports. A supported replacement describes coordination across warehouse, support and dispatch or names Mira’s contribution to the four-person pilot. A team count does not answer who supervised whom, so it cannot be used as a shortcut to managerial scope.

Repair F4, ownership and implementation: replace “independently designed and launched a company-wide automation programme” with “drafted a handoff checklist, recorded feedback and maintained the issue log during a four-person pilot”. The supervisor approved the procedure, the workflow was bounded, and no automation programme exists in the packet. The corrected wording retains initiative: drafting a useful checklist is a real contribution. It does not need sole ownership to be worth showing.

Repair F5, the earlier role: restore “Service Associate, Brindle Help, January 2021–December 2022”. Remove the invented seniority, extra year, Salesforce expertise and refund authority. R2 supports recording enquiries, checking order status, writing updates and escalating disputed refunds. The last action can be a strength for J1 because the vacancy explicitly values proper escalation. Removing invented authority can therefore improve both truth and relevance.

Repair F6, learning status: retain the Diploma in Business Operations and the fourteen-hour Spreadsheet Reporting Basics course with its completion status. Remove the invented degree and professional certifications. Do not turn a course title into an assessment of expertise. The unfinished Power BI course can be omitted from the clean example; if a real application explicitly asks about current study, state the real status without suggesting completion.

Repair F7, briefing scale: two one-hour sessions for four colleagues becomes exactly that. “Four teams through four workshops” changes both audience and event count. There is no evidence that the briefings improved performance, so do not add an effectiveness claim. You can describe the output and audience while leaving the unmeasured result alone. These repairs demonstrate a general pattern: a useful sentence preserves the right noun, verb, scope and status together.

Back to contents · Next: 10. Recalculate metrics before deciding whether to include them

10. Recalculate metrics before deciding whether to include them

R6 gives a baseline error rate of 48 divided by 400, or 12%. The pilot rate is 35 divided by 500, or 7%. The difference is five percentage points. The relative reduction in the measured rate is five divided by twelve, or approximately 41.7%. Saying “a five-per-cent reduction” confuses a percentage-point difference with a relative change. Saying “42% fewer errors” can also be misleading when the underlying comparison concerns rates and the order totals differ.

The raw error counts fell from 48 to 35. That is thirteen fewer recorded error-bearing orders, or approximately 27.1% fewer in the two samples. It is not the same quantity as the relative change in the error rate. Because the pilot contains more orders than the baseline, the denominator is essential. You should not choose whichever percentage sounds larger and remove the context that tells a reader what it measures.

The strongest compact formulation for this exercise is: “Contributed to a four-person handoff pilot; the measured missing-reference error rate was 12% before the pilot and 7% during it, across 400 and 500 eligible orders respectively.” A second sentence in a fuller discussion can state that other arrangements changed and the result cannot be attributed to Mira alone. In the clean CV below, we combine these limits carefully without claiming a controlled experiment.

Repair F8, causal inflation: remove “cut errors by 42%” as a personal achievement. The rate comparison is arithmetically available, but the causal claim is not. The packet does not isolate the checklist’s effect, and Mira did not control the whole process. Correct arithmetic cannot rescue incorrect attribution. “Contributed to” also needs a real contribution; in this case R5 supplies one, so the phrase is supported rather than a vague disclaimer.

Repair F9, invented business outcomes: delete the $50,000 saving and 100% customer-satisfaction claim. No cost model, salary rate, time study, survey result or revenue record is supplied. There is nothing to recalculate. Asking SI for a conservative estimate would introduce a new modelling exercise, not establish a past achievement. If a future estimate is useful for another purpose, label it separately and never insert it into the historical CV as a realised result.

Repair F10, permanence: do not imply the pilot rate persisted. R7’s later sample is 19 divided by 240, approximately 7.9%. That number concerns a different period and sample. It cannot be silently combined with the original pilot or used as proof of an enduring improvement. The final CV need not carry every observation, but its selected statement must retain the original time boundary and avoid a broader claim.

A useful metric earns its space by clarifying work. Here the pilot comparison shows experience with a measured process change, while R4 shows reliable reporting. The document does not need to include all available arithmetic. You need to understand all of it so that the short version remains accurate when someone asks what the number means.

Back to contents · Next: 11. Test what a sentence implies, not only what it literally says

11. Test what a sentence implies, not only what it literally says

A factual audit needs to inspect implied claims. “Led reporting across warehouse, support and dispatch” may sound close to “prepared a weekly summary”, but it can imply authority over the reporting function. “Designed training” can imply a programme with learning objectives and assessment when the evidence shows a short explanation of a checklist. The correct test is what a reasonable reader would understand, not whether each individual word appears somewhere in your notes.

Repair F11, tool expertise: “advanced automated dashboards” becomes a specific description of filters, SUMIF formulas and a basic pivot table used in weekly reporting. Those details may sound less fashionable, but they match J1’s actual need. R4 contains no automation or dashboard-development evidence. A tool list should function as a compact index to work you can discuss, not a collection of attractive products mentioned in the vacancy.

Repair F12, the public profile: remove the restricted pilot metrics, invented cost saving and expert labels. This is both a truth correction and a destination correction. A statement can be supported in a private application and still be unsuitable for a public profile under R10. The public version can name the checklist contribution and cross-team coordination without sharing the restricted counts. Do not assume that leaving out a company’s name makes the metrics public-safe.

Use a reverse question for every strong sentence: “What would someone reasonably ask me to demonstrate because I wrote this?” If you write that you managed staff, expect questions about supervision and decisions. If you claim a saving, expect a baseline and calculation. If you call yourself certified, expect the issuer, credential name and current status. The review helps you see whether the wording invites a question that your evidence cannot answer.

Then check underclaiming. Removing every strong verb can hide genuine work. Mira did draft a checklist, maintain a log, prepare twenty reports and deliver two briefings. “Helped with admin” would lose useful evidence. The right repair does not turn confidence into vagueness. It uses the strongest verb that the source supports and supplies an object and context so the reader can understand the contribution.

Finally, check whether the overall impression matches the parts. A headline can overstate seniority even when the bullets are accurate. A skills section can imply expertise that a later “introductory course” line does not repair. Read the document from top to bottom as someone who has not seen the private packet. The headline, summary, chronology and examples should describe one coherent professional record.

Back to contents · Next: 12. The corrected private-application CV

12. The corrected private-application CV

The following is the finished content example for J1. It is deliberately presented as readable text rather than a promise about page count or a particular applicant-tracking system. It uses only the fictional packet. The email address is illustrative, no live application link is provided, and the source identifiers have been removed from the clean copy. The explanation after the example shows where its claims came from.

Mira Chen — Service operations and coordination. Contact: mira.chen@example.com. Application focus: Service Operations Coordinator, North Quay Services. This is a fictional educational example, not a document for submission.

Professional summary: Service and operations professional with over five years of experience in customer updates, handoff records and weekly reporting. Prepared spreadsheet summaries, followed up missing order references and contributed to a team checklist pilot. Interested in service-operations coordination work combining clear records, practical process improvement and appropriate escalation.

Relevant capabilities: Spreadsheet reporting using filters, SUMIF formulas and basic pivot tables; handoff-log maintenance and follow-up; customer enquiry records and written order-status updates; escalation of disputed refunds to a supervisor; preparation and delivery of short peer briefings on an approved procedure.

Operations Coordinator — Cedar Trade, March 2023–September 2026. Maintained a handoff log connecting warehouse, customer-support and dispatch colleagues, checked missing order references and followed up unresolved entries for the operations supervisor. Prepared twenty scheduled weekly reports between May and September 2025, each submitted by the agreed deadline, using spreadsheet filters, SUMIF formulas and a basic pivot table to group issue types.

Drafted the first checklist, recorded feedback and maintained the issue log for a four-person handoff pilot in July–August 2025. Across the six-week baseline and six-week pilot, the measured missing-reference error rate was 12% of 400 eligible orders and 7% of 500 respectively. Other operating arrangements also changed; the comparison is a team-level observation and does not isolate an individual effect.

Delivered two one-hour briefings to four colleagues on using the handoff checklist, with material reviewed by the operations supervisor. The role involved coordination and reporting without direct reports or budget responsibility.

Service Associate — Brindle Help, January 2021–December 2022. Recorded delivery enquiries, checked order status in a ticketing system and wrote customer updates. Escalated disputed refunds to a supervisor rather than approving them directly. Career break: January–February 2023.

Education and completed learning: Diploma in Business Operations, Elm Learning College, completed 2020. Spreadsheet Reporting Basics, Quarry Learning, fourteen-hour course completed February 2025; certificate of completion. Both institutions and qualifications in this example are fictional.

The summary’s duration traces to R1. The spreadsheet claims trace to R4, the checklist and team boundaries to R5–R6, the briefings to R8, and learning status to R9. J1 explains why these details appear. The CV does not include the unfinished Power BI course because it does not supply demonstrated ability for the desirable requirement. It does not invent a telephone number or a referee to make the contact section look complete.

Would every real CV need the sentence about no direct reports? No. It is included here to make the repaired boundary unmistakable after the flawed example. In a real document, the accurate title and bounded bullets may already communicate the scope. You can shorten wording when the meaning survives. Likewise, you could omit the pilot metric entirely and retain a strong application based on the checklist contribution and reporting record. Quantification is optional; accuracy is not.

Back to contents · Next: 13. Build a public profile from the same record

13. Build a public profile from the same record

A public profile should help someone understand your work without requiring knowledge of a single vacancy. For Mira, a supported headline is: “Service operations and coordination | Handoff records, customer updates and spreadsheet reporting”. It describes areas of work rather than claiming a current job at North Quay. It does not present an application target as employment or promote an unfinished course into an expertise label.

Here is the complete fictional About example: “I work on the practical connections between customer updates, operational records and team handoffs. My experience includes service enquiries at Brindle Help and operations coordination at Cedar Trade, where I maintained a handoff log and prepared weekly spreadsheet summaries. I also drafted a checklist and recorded feedback during a team pilot, then delivered short peer briefings on its use. I am interested in service-operations coordination roles where clear records, thoughtful follow-up and appropriate escalation help work move reliably.”

The experience entries can retain the same titles and dates as the CV, with the publicly permitted task descriptions. The headline and About text do not include R6 or R7 results, cost savings, a claim of management or unsupported tool expertise. They also avoid saying “currently at Cedar” because the role ended. Consistency is about the underlying facts; the public profile does not need to reproduce every application sentence.

LinkedIn’s own help explains that a public profile can be visible to people outside the service and through search tools, subject to visibility settings. It also cautions that search tools can take time to reflect changes. This is a reason to choose public content deliberately, not a reason to assume that every profile has the same audience or that deleting a sentence removes every copy immediately. LinkedIn: public profile visibility.

Before publishing a real profile change, inspect what the platform will display, who can see it and whether the change triggers a notification. This guide does not change any account settings or publish a profile. A draft is a useful pause point: read it as a former colleague, a prospective employer and a stranger might read it. Do they all receive a fair impression of the same career, and is the level of disclosure acceptable?

Do not create fake recommendations or endorsements to fill an empty section. A recommendation requires an actual person’s words and appropriate permission to use them. A portfolio item needs an accurate description of your contribution and permission to share its contents. Leaving a section empty is preferable to manufacturing social proof that the rest of the record cannot support.

Back to contents · Next: 14. Improve the voice without inflating the evidence

14. Improve the voice without inflating the evidence

Once the claims are sound, you can work on rhythm and readability. Ask SI to shorten a long sentence while preserving the action, object, scope and limit. Then compare the versions. A good compression might replace “was responsible for the preparation of weekly summaries” with “prepared weekly summaries”. A bad compression might replace “contributed to a pilot associated with a lower measured error rate” with “reduced errors”, dropping the attribution boundary.

Use language that you can explain comfortably. You do not need to imitate a corporate voice if the result hides the work. “Maintained a handoff log” is concrete. “Orchestrated cross-functional synergies” leaves the reader to guess what happened. You can be precise and confident without adopting a vocabulary that you would struggle to use in a normal conversation about your work.

The New Zealand Tertiary Education Commission’s Tahatū guidance on AI-assisted CVs advises people to personalise, check and edit results, including exaggerated claims. It also suggests adding personal contact information afterwards rather than placing it into the tool unnecessarily. These are useful drafting principles, while local language conventions and application instructions still need to follow the actual destination. Tahatū: using AI to create a CV or cover letter.

Try reading a sentence aloud, then giving a thirty-second explanation of it without the document. If your explanation immediately requires “Well, I did not really mean that”, the wording needs revision. If you can describe the work, the decisions and the limit naturally, the sentence is more likely to reflect your own voice. This is a test of explainability, not a claim that an interview can be rehearsed into a guaranteed outcome.

Do not make every bullet follow an identical template. Some statements establish responsibility, others a product, others a measured observation. Repeating “achieved X by doing Y, resulting in Z” can push you to invent a result where none was recorded. Let the shape follow the evidence. The reader benefits from a clear set of different contributions, not from a page that sounds mechanically uniform.

If you use another language, preserve names, dates, credential status and responsibility while adapting expression. Do not translate a diploma into a degree or an organisation-specific title into a higher rank. Where equivalence matters, consult the relevant institution or qualified source. SI can suggest wording, but it cannot establish credential recognition merely by producing a familiar label.

Back to contents · Next: 15. Make the final file usable before you share it

15. Make the final file usable before you share it

A content draft and an application file are different products. After moving the approved text into your chosen document, inspect the exported file itself. Check that headings, dates and employer names remain adjacent to the correct experience. Check that a page break does not leave a heading isolated from its content. Open the file on an ordinary screen and confirm that the text is readable without extreme zoom.

Use the file type and any length or naming instructions requested by the actual employer. Avoid universal claims that a particular layout is guaranteed to pass an applicant-tracking system. A simple structure makes your own checks easier, but it does not reveal every employer’s software or selection process. If a portal extracts text into fields, inspect the extracted dates, titles and qualifications before confirming them; do not assume the upload was interpreted correctly.

The UK’s National Careers Service recommends clear, consistent formatting, relevant tailoring, up-to-date information and another person’s proofreading. Its guidance is useful as a readability reference, not a rule that overrides a Singapore employer’s instructions or a specialist sector’s requirements. National Careers Service: writing a CV.

Copy the final document’s text into a plain-text view and read the sequence. This can reveal missing text, misplaced columns or contact details that were turned into an image. It is a local inspection method, not a certification of compatibility with every system. Follow it with visual inspection because text extraction alone cannot show whether a line is clipped or the page is uncomfortable to read.

Check every actual link you choose to include. It should lead to the intended work, be accessible to the intended recipient and contain only material you may share. A portfolio link that opens a private working folder can disclose much more than its label suggests. A broken link can also undermine an otherwise clear example. Do not let SI invent a plausible-looking portfolio address to complete a template.

Keep the final filename understandable and preserve a copy of exactly what you would send. “Mira_Chen_Service_Operations_CV_2026-10-01” illustrates a useful naming approach. A real filename should follow the destination’s requirements and your privacy preferences. This workshop ends at a checked draft; deciding to submit it, selecting recipients and completing any declaration remain separate actions.

Back to contents · Next: 16. Resolve contradictions before making the story smoother

16. Resolve contradictions before making the story smoother

Suppose your old CV says a job began in February while an employment record says March. Do not ask SI to blend the two into “early 2023” merely to hide the discrepancy. Find out which date describes what: signing an offer, starting paid work and taking on a new role can be different events. The right date depends on the claim being made. A broad label can sometimes be appropriate, but it should follow clarification rather than replace it.

Use a short uncertainty note in the working record: the claim, conflicting sources, the question needed to resolve it and the affected sentences. Until resolved, omit the doubtful detail or keep the document on hold if the detail is essential. A model’s confidence is not an additional source. Nor does repetition across several old profiles make an inherited error more reliable.

Keep descriptions of career breaks proportionate. Mira’s packet supplies a two-month break and no reason. The correct exercise answer does not add caregiving, illness, travel, redundancy or study. Real applicants can decide how much context to give within the actual application requirements. Honesty about chronology does not require a writing tool to invent or expose intimate explanations.

For team projects, a contradiction may concern ownership rather than dates. One project summary may use “we”, while an old bullet uses “I led”. Ask what decisions you were responsible for, what you produced and what another person approved. A supported description can distinguish initiative from formal authority. You do not have to claim the whole outcome to show that your contribution mattered.

For qualifications, separate the official name, completion status, issuing organisation and any current validity requirement. If a real role has a mandatory credential, establish its status through the relevant source rather than interpreting a similar course as equivalent. This is particularly important where titles or practice rights have professional meaning. A CV editor should surface the uncertainty, not make a legal or eligibility determination from incomplete notes.

Once a contradiction is resolved, update the master evidence record first. Then update each affected CV and profile draft. Editing only the latest file leaves an older version ready to reintroduce the same error. Good maintenance is mostly a small number of carefully propagated corrections, not endless rewording of sentences that were already sound.

Back to contents · Next: 17. Maintain one factual record and several purposeful versions

17. Maintain one factual record and several purposeful versions

A master evidence record is not a master application to send everywhere. It is the stable source from which different documents are selected. Keep the factual record separate from vacancy-specific language so that tailoring changes emphasis rather than history. You might have a service-operations CV, a training-coordination CV and a general public profile, all using the same dates and qualifications.

For each version, record its target, the source-packet date, the vacancy version and the latest review. A small note such as “J1, packet version 2, metric wording checked, public-profile restrictions applied” can prevent confusion later. This is especially useful when you revise several documents in one session. Without a record, a draft that contains private-application figures may be mistaken for the public version.

When new experience arrives, add the evidence before updating the profile. What was completed? What can be shown? What changed from the earlier status? A course moving from “in progress” to “completed” should be supported by the actual completion record. A new responsibility should have a real start date. SI can identify the affected sentences, but it should not assume that time passing means a planned achievement happened.

If you have already shared a document and discover a material error, make a clear correction through the appropriate channel rather than silently changing only your private copy. The exact response depends on the context and seriousness of the error. A mistyped month and an invented qualification are different problems. For a consequential issue, seek suitable advice instead of asking a writing assistant to make the discrepancy disappear.

Avoid endless polishing as a substitute for building capability. Once the facts, relevance, clarity and file checks are satisfactory, further adjective changes may add little. A missing work sample or unpractised skill needs a different activity. Your career-development plan can hold that activity while the CV accurately represents your current position. The document should grow because your evidence grows, not because a model keeps finding more elaborate descriptions.

Back to contents · Next: 18. Measure progress without inventing a hiring experiment

18. Measure progress without inventing a hiring experiment

You can evaluate the editing process before any application leaves your computer. For Mira’s exercise, progress means that the twelve named failure categories have been repaired, the essential J1 tasks have visible evidence, the dates and qualifications agree across versions, and the public profile respects R10. These are observable document properties. They do not estimate the probability of an interview or demonstrate that AI improves hiring outcomes.

A useful review can distinguish factual defects from presentation defects. A wrong date, unsupported metric or changed credential is a factual defect and must be corrected. A long sentence may be a presentation defect, but shortening it must not reintroduce a factual one. This order prevents a beautifully formatted document from receiving a high overall score despite a serious misrepresentation.

Ask an informed reader to locate the evidence for three requirements without explaining the document first. If the reader cannot find the reporting example, you may need a clearer heading or earlier placement. If the reader believes Mira managed a team, the overall wording still carries a misleading implication. Record the actual misunderstanding and change the relevant passage. Do not turn one person’s reaction into a universal recruiter preference.

If you later track real applications, define the stages carefully: sent, acknowledged, invited to interview, progressed and offered are different outcomes. Keep the relevant role and date with each record. A small collection of applications changes in many ways at once, including the vacancies, timing and competition. A higher response count after editing is worth noticing, but it does not by itself isolate the effect of the CV or the SI tool.

Time savings also need an honest denominator. Count preparing permitted sources, checking claims, resolving discrepancies and inspecting the final file, not just the seconds required to generate prose. If manual editing is faster for a short, already accurate CV, that is a valid choice. The point is a better working method for you, not proving that an AI tool must be used for every sentence.

Stop when the document is accurate, relevant, understandable, appropriately disclosed and ready for the next authorised step. Continue when a specific unresolved issue remains. These stopping conditions are more useful than asking SI whether the CV is “perfect”, because the model can praise the result without having any new evidence about your history or the employer’s decision.

Back to contents · Next: 19. Independent transfer: change the role, then change the evidence

19. Independent transfer: change the role, then change the evidence

Now try the task without copying the earlier answer. Mira is considering J2, an invented Training Operations Assistant vacancy at fictional Harbour Learning Services. The essential tasks are maintaining session schedules and attendance records, sending clear joining instructions, preparing simple spreadsheet summaries and coordinating questions with the trainer. Previous learning-management-system experience is desirable, with training provided. The role does not involve independently assessing competence or awarding qualifications. The same truthful-editing and document instructions as J1 apply.

Keep R1–R12 unchanged. Add T1: from February to April 2026, Mira volunteered as a Session Administrator for fictional Bridge Workshop Circle, which ran six introductory spreadsheet sessions for adults. Mira maintained the schedule, sent joining instructions prepared with the trainer and recorded attendance. The trainer designed and taught the lessons. Mira did not assess learners, issue qualifications or operate a learning-management system. This unpaid activity overlapped Cedar employment and must remain labelled as volunteering.

T2 supplies the complete attendance record. Eighteen adults were invited to each of the six sessions. Attendance counts were 16, 14, 13, 15, 12 and 12. The record does not identify which individuals attended each session, so it cannot establish how many people completed the series. Each session lasted ninety minutes. Mira attended all six as administrator; preparation time was not recorded. The exercise does not provide any participant names, contact details, ages beyond adulthood or other personal data.

T3 supplies feedback limits. Eight anonymous end-of-series forms rated administrative clarity on a five-point scale, with responses of 5, 4, 4, 5, 3, 4, 5 and 4. No question measured employment outcomes, learning gains or trainer competence. There were eighteen invited adults, but the record does not establish that each received or was able to return a form. Do not call eight divided by eighteen a verified survey response rate without that distribution information.

T4 supplies disclosure and status. Mira may publicly describe the volunteer role, organisation, dates, six sessions and administrative tasks. Aggregate attendance and feedback numbers may be used only in a private application; this exercise’s de-identified aggregate summaries are also permitted for the drafting tool. Raw attendance records and individual feedback must not be shared. No new qualification, promotion, software skill or employment is added to R9. T3’s individual ratings are printed only for local/manual checking and must not enter the tool. Only the locally calculated feedback count, sum and mean may enter the permitted drafting tool or private application; the public prohibition remains.

If you use SI for this transfer, extend D1 only with de-identified T1 task/date summaries, T2’s permitted session-level aggregates, and the locally calculated T3 count, sum and mean, while retaining T4’s disclosure rules. Omit names, contacts and individual feedback from the tool input; restore names locally after review. Your task is to write a revised summary, two relevant experience statements and a public headline for J2. Identify the evidence for the essential tasks, the unsupported desirable skill, the total recorded attendances, the proportion of possible session attendances and the average of the eight feedback ratings. Explain why “trained eighteen adults to completion with 95% satisfaction” is unsupported. Decide whether the volunteer work belongs above the service-pilot metric in this version.

As a boundary test, imagine J3 instead: Independent Assessor, with a specified current assessor qualification stated as mandatory and responsibility for formal competence decisions. No such qualification or assessment experience exists in the packet. Explain what you must not claim. This change tests whether you can distinguish useful transferable evidence from a missing mandatory requirement rather than treating every role change as a keyword exercise.

Back to contents · Next: 20. Answered transfer: a different emphasis, the same history

20. Answered transfer: a different emphasis, the same history

For J2, the most direct evidence is T1’s session administration: schedule maintenance, joining instructions, attendance records and coordination with the trainer. R4 adds established spreadsheet reporting, and R8 adds bounded experience explaining a procedure to colleagues. The service role remains useful background, but the private pilot metric is less central than the newly supplied training-administration work. Reordering the selected evidence is justified; changing the paid employment titles is not.

A supported revised summary is: “Service and operations professional with experience maintaining records, preparing spreadsheet summaries and coordinating clear updates. Volunteered as a session administrator for a six-session adult workshop series, supporting scheduling, joining instructions and attendance records alongside the trainer. Seeking training-operations support work where reliable administration and clear handoffs matter.” The summary does not say that Mira designed the curriculum, taught the series or assessed competence.

The first experience statement is: “Session Administrator, volunteer, Bridge Workshop Circle, February–April 2026: maintained the schedule, coordinated joining instructions with the trainer and recorded attendance for six adult spreadsheet sessions.” The second is: “Prepared twenty scheduled weekly reports at Cedar Trade between May and September 2025, using filters, SUMIF formulas and a basic pivot table, with each report submitted by its agreed deadline.” Together they connect direct J2 task evidence with an established reporting example.

The public headline can be: “Operations coordination and training administration | Schedules, records and clear participant information”. The volunteer entry should sit in a clearly labelled relevant-experience or volunteering section. Do not merge it with Cedar or imply it was a paid training job. The public version uses the T4-permitted scope and tasks, leaving the restricted attendance and feedback statistics out.

The six attendance counts sum to 82 recorded attendances. There were 18 multiplied by 6, or 108 possible session attendances. The ratio is 82 divided by 108, approximately 75.9%. These are attendance occasions, not eighty-two different people. Nor can the aggregate counts tell us how many of the eighteen invited adults attended every session. We lack the person-by-session information required to calculate completion.

The eight clarity ratings sum to 34, and their arithmetic mean is 34 divided by 8, or 4.25 out of 5. That is the mean among the eight returned forms for the specific administrative-clarity question. It is not a learning score, satisfaction rate for all participants or a 95% result. Converting 4.25 out of 5 into a percentage would produce 85% of the scale maximum, but even that would not establish the proportion of people who were satisfied. The clean CV does not need this metric to show J2 relevance.

Mira’s recorded presence at the sessions was six times ninety minutes, or nine hours. That is session attendance time only. Preparation time is unknown, so nine hours must not be presented as the total effort invested in the volunteer role. The record also says the trainer taught the lessons. “Trained eighteen adults to completion” therefore fails on responsibility, completion evidence and the implied outcome, independently of the incorrect satisfaction figure.

There is no supplied learning-management-system experience. J2 says that experience is desirable and training is provided, so the draft can describe adjacent record-keeping honestly without claiming the missing product skill. If asked in a conversation, Mira can say that this would be a new system to learn. For J3, the mandatory qualification is absent. The packet cannot support a claim of meeting that condition or performing formal assessment. Clarifying the requirement or building the necessary qualification would be a separate next step; rewriting the CV cannot create it.

Back to contents · Next: 21. Use the transfer to diagnose your own editing mistakes

21. Use the transfer to diagnose your own editing mistakes

The transfer is useful because the correct answer is not a more polished version of J1. The reader’s decision has changed. If you kept the service-pilot statistic at the top and buried T1, the facts might remain true while the emphasis became less useful. If you renamed Mira a trainer to improve apparent fit, the emphasis would become misleading. Good tailoring selects stronger intersections without inventing a new past.

If you called 82 a participant count, revisit the unit. Each person could appear on several session lists. When a source records occasions, transactions, orders or responses, do not convert them into unique people without evidence. Many CV errors begin with a unit change hidden inside a short sentence. Keep the noun attached to the number during drafting, even when it feels repetitive.

If you reported a completion rate, revisit the missing information. Aggregate attendance totals do not reveal individual paths. One person might have attended every session while another attended none, and several different individual patterns can produce the same total. The honest response is not to select a likely pattern. It is to state that completion cannot be established from this packet.

If you described the feedback as a success rate, revisit the measure. A rating of administrative clarity concerns one aspect of the experience and a limited respondent group. It says nothing direct about learning, qualifications or jobs. A useful source label should preserve the question asked, the denominator and the population covered. These limits do not make the feedback worthless; they tell you what it can responsibly support.

If you included the metrics in the public headline or About text, revisit the destination. T4 permits some facts publicly and restricts others. You can demonstrate the same capability through task descriptions without exposing the restricted figures. This is a practical editing decision, not a general legal ruling that one particular type of information is always permitted or prohibited everywhere.

If you treated J3 as another easy tailoring exercise, revisit the difference between description and eligibility. Transferable skills can support a transition, but they do not establish a required qualification. A good assistant should make that gap easier to see. The right next action may be a question or a learning plan, rather than another draft that hides the gap behind confident language.

Back to contents · Next: 22. Questions that often arise during a real revision

22. Questions that often arise during a real revision

What if I do not have any impressive numbers? Begin with the work itself: a record maintained, an explanation written, a problem investigated, a decision supported or an output completed. Add scale when you can support it. Do not infer a percentage from a feeling that the process improved. A precise account of a useful action gives a reader more to discuss than an impressive number with no defensible meaning.

What if I cannot access records from an old employer? Use what you may legitimately retain and can explain honestly. Do not retrieve or share restricted records merely to support a CV. You can reduce precision, use a bounded task description or seek appropriate confirmation. If a claim needs documentation you do not have, treat that as a limitation rather than asking SI to reconstruct a plausible document.

Can SI give me a better job title? It can suggest a plain-language explanation of your duties, but the official title should remain identifiable. For example, you might retain the original title and add a short functional description in parentheses if it is accurate. Do not quietly turn coordination into management, a placement into employment or a course into a licence. The aim is comprehension, not an upgrade in status.

Should I disclose AI assistance? Follow the actual application, assessment and professional requirements. If asked, describe the assistance honestly. Do not claim that a document was unaided when that would be false under the relevant instruction. There is no need for this guide to invent one universal disclosure rule across all employers and jurisdictions. The operational habit is to inspect the destination’s rules before relying on a workflow.

Can I use a public profile as my source of truth? It can help you recover material, but it may contain old or simplified information. Compare it with better records before feeding it back into a new draft. Otherwise an earlier exaggeration can circulate through several documents until it feels established. Agreement between copies does not create independent support.

What if a model tells me that everyone exaggerates? That is not evidence about your record or permission to misrepresent it. Return to the task: produce the strongest defensible description of real work. If a model repeatedly upgrades the claims, narrow the requested transformation or edit manually. You do not owe a tool another attempt when its output creates more checking work than it saves.

What if an application asks for a personal characteristic or sensitive document? Consider the actual requirement, destination and context rather than letting SI populate a template automatically. Verify who is requesting it and why, and seek appropriate advice if the request is unclear or consequential. This article does not decide the legality of a particular request. It keeps sensitive disclosure as a separate decision from improving career prose.

Can I claim that I am good at SI because it helped write this CV? Completing this exercise supports a narrower statement about practising source-bound drafting and review. It does not prove advanced AI engineering, deployment or governance expertise. If you want to describe AI-related capability, point to a real task you performed, the checks you made and the result you can show, with its limitations intact.

Back to contents · Next: 23. Finish with a document you can explain

23. Finish with a document you can explain

The practical sequence is now complete: capture a permitted source record, clarify the role, expand the possible wording without expanding the facts, structure the relevant evidence, produce a draft, review its claims and improve the weak parts. The same cycle works whether you use SI for every drafting pass or only for a difficult paragraph. The human contribution is not a final rubber stamp; it is ownership of the facts and the decision to share them.

Before you leave the document, choose one sentence from the summary, one from the work history and one from the profile. Find their source support. Explain the responsibility and any number in ordinary speech. Check that the public version contains only what belongs there. If one line fails, repair that line and any versions that inherited it. You do not have to restart the whole career story.

The next useful step depends on the gap you discovered. Missing evidence may call for a work sample or a skills-development plan. Unclear direction may call for comparing career options. A strong, verified document may lead to preparing spoken examples from the same record. None of those steps requires inventing achievements to bridge the distance between where you are and where you want to go.

Return to How to Leverage Your Life with Super Intelligence for the wider personal-use framework, or to the Career & Adulthood Hub for broader career reasoning. Keep the final standard simple: your CV should make your actual work easier to see, and your professional profile should remain recognisably yours.

Back to contents · Next: Sources and scope

Sources and scope

MyCareersFuture: Just 3 Steps With Resume Checker to Improve your CV, dated 15 May 2020, supports examining the job description and matching relevant skills that the applicant has. This guide uses that principle, not the old article as proof of a current interface or screening guarantee.

CSA and IMDA: joint advisory on safe and secure use of generative AI, dated 28 July 2026, supports careful information sharing, attention to privacy terms and independent verification of generated content. It is individual-use guidance, not permission to disclose workplace information.

Tahatū Career Navigator, Tertiary Education Commission: Using AI to create a CV or cover letter, updated 29 September 2026, supports personalised, checked drafting and cautious handling of personal information. Its New Zealand-specific conventions are not presented as universal application rules.

National Careers Service: How to write a CV supports readable structure, relevant tailoring and review. LinkedIn Help: public profile visibility explains the scope and limits of public-profile visibility and search refresh. Check the current destination before changing a real profile or submitting an application.

Sources were checked for this guide on 1 October 2026. The Mira, North Quay and Harbour exercises, their records and their results are wholly fictional. The calculations demonstrate how to inspect the supplied inputs; they are not model-performance measurements, employment research, legal advice or evidence that any particular application will succeed.

Back to contents · Return to the Super Intelligence master guide