Use AI to turn your work history into a clearer CV

“Responsible for administration” can hide years of useful experience. Perhaps you coordinated deliveries, spotted mismatched invoices or trained colleagues to use a booking system. AI can help turn those details into clear CV entries, but the strongest improvement starts before the rewrite: recover what you actually did, how you did it and what changed.

This is a proposed text-based workflow with an original fictional work-history exercise. It is designed to produce a CV you can check and edit; we have not tested whether these prompts improve hiring outcomes. Begin with the employer’s application instructions. Oxford’s careers guidance notes that organisations differ over permitted AI use, including whether assistance is allowed during applications or assessments.[4]

Recover the material before polishing it

Keep a master record of jobs, dates, qualifications, tools and examples. The National Careers Service recommends considering skills gained through employment, education and daily life; transferable skills can include communication and problem solving. A career change may alter which experience is relevant without making the earlier work disappear.[2]

Paste selected work notes into an assistant, with the vacancy separately labelled. Leave out client names, other people’s contact details and confidential business information that the exercise does not require. Your working notes can use placeholders, with your real contact details restored in the final document. Oxford specifically advises care when supplying personal or confidential data to AI tools.[4]

Imagine a fictional stock administrator who writes: “I handled deliveries and helped with errors.” Follow-up questions reveal that they matched delivery notes to purchase orders, recorded discrepancies in an existing spreadsheet and sent a daily exception list to a supervisor. They trained two starters on that checking process. They did not design the spreadsheet or manage the supplier contracts.

Those details are useful even without a percentage improvement. “Improved operations” provides little evidence; describing the checking process explains a real skill. Oxford’s CV guidance recommends showing contribution, achievement and scale in straightforward language. A count or result should clarify the work you can substantiate.[3]

Write an accurate achievement

Suppose the same person also maintained a log of unresolved delivery differences. In our invented example, the log contained 18 open items on 1 May and six on 31 May. The supervisor handled supplier negotiations while the administrator checked records and followed up missing paperwork. That gives a measurable team outcome and a bounded personal contribution.

A defensible draft might read: “Checked records and followed up missing delivery paperwork during a team effort that reduced the open-items log from 18 to six in May.” The reduction is 12 items, or about 67%. Those are calculations from the fictional inputs. Neither figure establishes that this employee alone caused the change, or that the same rate continued afterwards.

Ask for two versions of a bullet: one emphasising the process and one emphasising the result. Read each back against the notes. If the shorter version removes who did the negotiating, it may create an inflated impression even when every remaining number is correct. Choose the wording that makes your contribution easiest to understand.

Use the vacancy to choose what deserves space

The National Careers Service recommends tailoring a CV to the job, keeping qualifications and work history current, and listing the most recent experience first. An experienced applicant can place work history before education. Give the relevant work enough detail and compress older material that adds little to this application.[1]

For the stock administrator, an operations role might justify leading with delivery checks. A training-support role might give the two new starters greater prominence. Both versions describe the same person. The selection changes because the reader’s question changes; a requirement for a qualification you do not hold must remain a gap.

What writing-assistance research actually tested

There is experimental evidence that clearer writing can help. A study published in Management Science in 2025 found an 8% relative increase in hiring after jobseekers received nongenerative algorithmic writing assistance. This was correction and editing support, not a trial in which a modern chatbot wrote an entire CV. It cannot establish the benefit of the prompts in this article.[5]

The original experiment ran in 2021 in an online freelance labour market. Its manuscript describes suggestions for spelling, grammar, punctuation and phrasing, with hiring examined during a new user’s first 28 days. The distinction matters: an 8% relative increase means multiplying a baseline probability by 1.08, rather than adding eight percentage points. The result concerns that intervention and setting.[6]

Make the final document readable after upload

Use the employer’s requested file format and a simple layout. Greenhouse’s own documentation names columns, text boxes, images and complex tables among possible causes of failed or partial CV parsing. Parsing means transferring information into candidate fields; a parsing problem is not, by itself, evidence that an AI rejected the application.[7]

After exporting, reopen the file and select or copy its text to check reading order. If an application form fills itself from the upload, inspect the job titles, employers and dates. Greenhouse says a failed parse can leave the document attached while the candidate details require manual entry. A readable PDF and correctly filled fields are separate things.[7]

Keep the final CV and the brief factual notes that support it. When someone asks about a line, you should be able to explain the work naturally: what arrived, what you checked, whom you worked with and what happened next. That is the point of the revision—the experience becomes easier for someone else to recognise.

Report an error or suggest a correction