Screening Resumes at Scale

The basic technique mirrors what you learned yesterday with LinkedIn profiles, but CVs offer richer data — detailed responsibilities, specific achievements, education details, and sometimes portfolios.

Here's your core prompt for individual CV screening:

"Here's a CV and the role requirements. Summarise this candidate in 5 bullet points: (1) Overall experience level and relevance. (2) Key skills match against requirements. (3) Notable achievements or metrics. (4) Gaps or concerns. (5) Interview recommendation with reasoning."

What makes this better than reading the CV yourself isn't just speed — it's extraction. AI pulls out the quantified achievements buried in paragraph three of their second job. It notices that their "marketing coordinator" role actually involved managing a budget and a team. It catches that they listed a certification that directly matches your requirements but buried it in the education section.

You'd find all of this too, if you had 10 minutes per CV. But you don't. AI gives you that 10-minute quality in 30 seconds.

Not all CV information is equally useful. Here's what AI should focus on:

Skills match — Does the candidate have the must-have skills? What about the nice-to-haves? AI should map each requirement to evidence from the CV.

Experience level — Is this person at the right level? Look beyond years of experience — what was their scope of responsibility? Did they manage people, budgets, or clients?

Career progression — Are they growing? Promotions, expanding scope, and increasing responsibility are strong positive signals. Lateral moves or downward moves might need context.

Quantified achievements — "Increased sales by 35%" is infinitely more useful than "responsible for sales." AI is excellent at finding and highlighting these metrics.

Recency — Skills from 8 years ago matter less than current capabilities. AI should weight recent experience more heavily.

Cultural indicators — Volunteer work, side projects, continuous learning, industry involvement — these suggest a candidate who goes beyond the minimum.

Use this prompt to get a signal-focused screening:

"Screen this CV against the attached role requirements. For each requirement, tell me whether the candidate meets it, partially meets it, or doesn't meet it — and cite the specific evidence from their CV. Then highlight any standout achievements or concerns."

This is where AI becomes transformational for high-volume roles. Instead of screening CVs one by one, you process them in batches and ask AI to rank them.

For smaller batches (5-10 CVs):

"Here are 8 CVs for a Marketing Manager role. Here are the role requirements. Rank these candidates from strongest to weakest fit. For each candidate, give a 2-line summary and a score out of 10. Explain your top 3 picks."

For larger volumes (20-50 CVs):

Process in batches of 10. Get a ranking for each batch. Then take the top 3-5 from each batch and run a final comparison:

"Here are my top candidates from each batch. Now compare them against each other and give me a final ranked shortlist of the top 5 to interview, with reasoning."

For very high volume (100+ CVs):

Use a two-pass approach. First pass with a quick screening prompt:

"Quickly classify this CV as Strong Match, Possible Match, or Not a Match for this role. One sentence explaining why."

Run through all CVs with this fast filter. Then take the Strong and Possible matches and do the detailed assessment you learned above.

Different roles require different screening lenses. Here are templates ready to use:

Technical role screening:

"Screen this CV for our [Role Title] position. Focus on: (1) Technical skills match — do they have our required stack? (2) System design experience — have they built or architected systems? (3) Scale — have they worked on systems with significant traffic or data volume? (4) Collaboration signals — do they show evidence of working cross-functionally? Rate them: Strong / Possible / Pass."

Sales role screening:

"Screen this CV for our [Role Title] position. Focus on: (1) Revenue numbers — have they hit or exceeded targets? (2) Deal size and sales cycle — does their experience match our market? (3) Progression — have they consistently moved up? (4) Industry relevance — do they understand our vertical? Rate them: Strong / Possible / Pass."

Executive screening:

"Screen this CV for our [Role Title] position. Focus on: (1) Scope — what size teams, budgets, and organisations have they led? (2) Strategic impact — have they driven business outcomes, not just managed operations? (3) Board or C-suite exposure — have they operated at this level before? (4) Transformation experience — have they led change or built something new? Rate them: Strong / Possible / Pass."

Save these templates. Adjust them for your specific roles and reuse them constantly.

This is critical. AI can help reduce screening bias, but only if you use it intentionally.

What AI does well: When you give AI a CV and a set of requirements, it evaluates skills against criteria. It doesn't know the candidate's age, ethnicity, or what neighbourhood they live in (unless you include that information). It focuses on what the person has done and can do.

What you must watch for: AI models are trained on historical data, which can contain biases. If you ask AI "does this person seem like a strong cultural fit?", it might make assumptions based on patterns that correlate with demographics rather than actual culture alignment. Stick to skills-based criteria.

Best practices for fair AI screening:

- Always define your criteria before screening, not during

- Use the same prompt and criteria for every candidate for the same role

- Ask AI to evaluate skills and experience, not "fit" or "potential" (which are subjective)

- If AI flags something as a concern, ask it to explain specifically why — vague concerns like "doesn't seem like the right fit" aren't actionable or fair

- Review AI's reasoning, not just its recommendation. If it's downranking a candidate, make sure the reason is related to job requirements

"When screening these CVs, evaluate only against the stated requirements. Do not make assumptions based on names, universities, locations, or employment gaps. Focus on demonstrated skills and achievements."

Add this to every screening prompt. It's a simple safeguard that keeps your process fair.

Here's how to integrate AI screening into your weekly workflow:

Monday morning: Check your ATS for new applications from the weekend. Do a quick first-pass classification (Strong / Possible / Pass) for all new CVs.

Throughout the week: As applications come in, do the quick classification immediately. This prevents the pile from growing.

Wednesday or Thursday: Take your Strong and Possible candidates from the week. Do detailed assessments. Build your ranked shortlist.

Friday: Send your shortlist to the hiring manager with structured notes. Each candidate has a summary, strengths, gaps, and your recommendation.

This system means you never face a mountain of unscreened CVs. You process continuously, and by the end of the week, your hiring manager gets a clean, reasoned shortlist — not a list of names with "looks good" next to them.

The total AI time per week? About 2-3 hours for a heavy pipeline. Compare that to the 15-20 hours you'd spend doing the same work manually.