Prepared for Handler · Executive search AI workflows

Candidate-to-role fit analysis, rebuilt as a repeatable AI workflow and tested for real

One of the use cases from your brief, worked end to end. Same search, same candidates, run the way most teams prompt today and then through a structured workflow your consultants can reuse on every assignment. Every AI output on this page is a real model run on a top-tier model. Wording is as it came back; I only trimmed length and swapped dashes for commas.

Sample role and candidates are fictionalNo real candidate data usedRuns made 2 October 2026
77real model runs behind this page, on two models
10 of 10identical scorecards on a top-tier model once scores were anchored
5unsupported claims in one single-prompt answer from a smaller model
8overstatements caught by the check step before a client would see them

Before and after: Candidate A

Left: the actual answer to one prompt, "Analyze this candidate's fit for the role below. Give a fit score out of 10 and a recommendation." Right: the same inputs through the workflow. Same top-tier model, same day.

Single prompt · actual output, run 1 of 5 (excerpt of 718 words)

Fit Score: 8.5 / 10

Recommendation: Advance to client shortlist.

Key Strengths

Rare combination for an exit mandate. This candidate can anticipate buyer findings and fix them years before the process begins.

Direct sector and model analog. There will be little ramp-up on payer dynamics, clinic-level P&Ls, or sponsor expectations.

Georgia CPA license. This is administratively convenient and may suggest prior ties to Georgia, which is worth asking about.

Gaps and Risks

No CFO title. Some sponsors strongly prefer someone who has already sat in the chair. Also assess retention risk if their current sponsor is near an exit, since timing and equity vesting may affect availability.

Bottom Line

If relocation and board-facing experience check out, this is a top-tier contender.

  • Good analysis, not a deliverable. 700 to 790 words, with a different shape every run (6 or 7 sections, between 0 and 7 interview questions). A consultant still has to rewrite it for a client.
  • Five runs, four wordings of the call: "Advance to client shortlist", "Advance to first-round interview (priority candidate)", "(strong priority)", "Advance to interview (strong shortlist candidate)".
  • 8 or 8.5 out of 10 with no criteria or weights behind it, so nobody can defend the number to a client.
  • Speculation sits next to evidence: Georgia ties guessed from the CPA licence, equity vesting, "little ramp-up". Run 2 also turned "led finance integration of 9 acquisitions" into "9 clinic acquisitions in 3 years" and estimated "$6 to 7M in cash" released.

Red = not in the candidate's record. Amber = opinion, prediction or hype presented alongside the facts.

Workflow · actual output, after the check step (trimmed)

Candidate A, VP Finance, PE-backed multi-state physical therapy network

Candidate A has been VP Finance for four years at a multi-state physical therapy network (~$140M revenue, PE backed). Their strongest matches to the brief are in revenue cycle, acquisition integration and sponsor reporting. Candidate A owned a revenue cycle turnaround that reduced days in A/R from 58 to 41. They led the finance integration of nine clinic acquisitions over three years. They also built a monthly board reporting package and a 13-week cash model for the sponsor. Their exit-readiness experience comes from the advisor side: six years as a Senior Manager in Big Four transaction advisory, working on sell-side QoE. Candidate A holds a CPA (Georgia). The record does not show a CFO title, board-level presenting or leadership of a company-side sale. Candidate A is based in Charlotte, NC. Willingness to relocate to Atlanta is not recorded.

Areas to explore

  • Has Candidate A led a company-side sale process or buyer diligence as the finance lead, beyond advisor-side sell-side QoE?
  • Is there a CFO or other senior finance leader above the VP Finance role, or is Candidate A the top finance lead?
  • Does Candidate A present to the board or sponsor directly, or do they prepare materials that someone else presents?
  • Is Candidate A willing to relocate from Charlotte to Atlanta, and on what timeline?

Recommendation: Advance. The weighted score is 4.25/5, with a measured revenue cycle result (days in A/R from 58 to 41), finance integration of nine clinic acquisitions, and a board reporting package and 13-week cash model built for the sponsor.

What the check step fixed in this draft (3 items):
  • "has not confirmed willingness to relocate" implied someone asked. The record is silent, so it now says "not recorded".
  • "measured results in revenue cycle, integration and sponsor reporting": only revenue cycle has a measured result. Reworded.
  • "as they could change the outcome": not supported by the scorecard. Removed.

Every claim traces to a line in the record. Gaps are asked, not filled in. Same shape and same wording rules every run.

How the workflow works

Most drift comes from asking one prompt to read, judge and write in a single pass. Splitting the jobs gives each step one task, a fixed format and clear rules. A final check, done by a separate prompt that did not write the draft, catches what slips through.

1

Extract the criteria

Turns the brief into weighted criteria with score anchors: what a 5, a 3 and a 1 look like. Agreed once per search, reused for every candidate.

2

Score against evidence

Picks the anchor that matches and quotes the exact line. No evidence means NOT EVIDENCED. The model never does the arithmetic.

3

Write in house style

Turns the scorecard into a client-ready summary with a fixed tone guide. Gaps become questions to probe.

✓

Check before it goes out

A separate checker compares the draft with the record. The template also checks each score against its anchor and does the weighting.

Criterion (Step 1)WeightScoreAnchor matched · evidence (Step 2, actual output)
PE exit readiness (QoE, data room, diligence)25%4 / 5Sell-side QoE from the advisor side. "Prior: Senior Manager, Big Four transaction advisory, 6 years (sell-side QoE)". Company-side sale: NOT EVIDENCED.
Sponsor / board-grade FP&A and reporting20%5 / 5Built and owns reporting used by a sponsor. "Built monthly board reporting package and 13-week cash model for sponsor"
Multi-site healthcare operations and revenue cycle20%5 / 5Multi-site with a measured result. "Owned revenue cycle turnaround: days in A/R from 58 to 41"
Add-on acquisition integration15%5 / 5Led 3 or more. "Led finance integration of 9 clinic acquisitions over 3 years"
Prior CFO / reports to a board10%2 / 5Prepares board materials that someone else presents. CFO title and presenting to the board: NOT EVIDENCED.
CPA5%5 / 5"CPA (Georgia)"
Atlanta based or will relocate5%1 / 5Lives elsewhere, relocation not confirmed. "Based in Charlotte, NC"
4.25 / 5 weighted by the template, not the modelAdvance to first interview

What testing showed

I didn't want to show you a workflow that only looks good once. So I ran each version several times on the same inputs, on a top-tier model and on a smaller, faster one, and kept every result. These are the numbers that came back.

Single prompt

Steady score, no structure

Candidate A fit score, 5 runs
8.58888
Candidate B, 5 runs
55555

A strong model gives sensible scores, but they rest on nothing you can show a client. Length, shape and wording change every run, and speculation sits beside the facts.

3 steps, first version

Close, but still drifting

Candidate A scorecard, 5 runs
samesame1 moved1 movedsame
Candidate B scorecard, 5 runs
samesamesamesamesame

Without written anchors, judgment calls wobble. In two runs one criterion moved a point (healthcare operations 5 to 4, relocation 1 to 2), enough to shift the weighted score from 4.25 to 4.05 or 4.30.

3 steps + anchors + check

Same answer every time

Candidate A scorecard, 5 runs
samesamesamesamesame
Candidate B scorecard, 5 runs
samesamesamesamesame

With anchors for every score and the arithmetic in the template, all 10 scorecards matched exactly. The check step then caught 8 overstatements across the two summaries and the post.

Same test on a smaller, faster model
Single prompt

5 unsupported claims in run 1, including "Led 9 clinic acquisitions" (it was the finance integration of them) and calling the candidate "he".

3 steps, first version

Got its own weighted total wrong in 9 of 10 runs, and the "Prior CFO" score swung between 2 and 4.

With anchors + check

Candidate A identical in 5 of 5 runs. The check caught three invented facts, such as "Director of Finance" and "available to start immediately".

The takeaway: on a smaller model, the workflow is what stops invented facts. On a top model, it is what makes the output consistent, checkable and ready for a client. Your team gets the same result whichever tool they open.

How this was run: two models through their APIs at default creativity settings, 2 October 2026, 77 runs in total. Everything quoted above the grey box is from the top-tier model. Single prompt and scoring were run 5 times per candidate on each model, plus the summaries, checks and the content post. Fictional inputs only. The prompts are plain text, written to work in ChatGPT, Claude or Copilot.

Content development from the same brief

The same discipline works for content. Here, a search announcement for LinkedIn, written from the brief with a fixed tone guide, then put through the same check.

Actual output, with the checker's fixes applied
We have been retained to appoint a Chief Financial Officer for a private equity-backed outpatient healthcare group in the Southeast, with around $180M in revenue. The business plans to grow through acquisition, with two to three add-ons planned each year. The business is targeting a sale in three to four years. The next CFO will lead finance through that exit, from quality of earnings readiness to the data room and buyer diligence. They will also build FP&A and monthly reporting to board and sponsor standard. This role suits someone who: - Has served as a CFO or senior finance leader reporting to a board - Knows multi-site healthcare operations, including payer mix and revenue cycle - Can integrate add-on acquisitions while preparing a business for sale A CPA is preferred. If this sounds like you, or someone you know, I would welcome a conversation. Message me directly or reply here. #CFO #HealthcareFinance
What the check caught
  • "The business is growing through acquisition": the brief only says add-ons are planned. Changed to "plans to grow".
  • "The sponsor is preparing for a sale": the brief says the business is targeting a sale. Corrected.
  • Flagged, left for the consultant: the post leaves out "Atlanta based or willing to relocate", a stated requirement. The checker suggested safe wording but didn't add it.

The client stays confidential (sector, size, region and ownership only), there are no hype words, and the post adds no requirements beyond the brief. A consultant still gives it a final read before posting, and that is built into the workflow.

Prompt library

The exact prompts behind this page, for the use cases in your brief. Paste into ChatGPT, Claude or Copilot and fill in the brackets. In a full engagement these get tuned to Handler's house style and saved as a shared library for the team.

You are a research analyst at a retained executive search firm.

TASK: Turn the role brief below into a scoring rubric. Do not evaluate any candidate.

RULES
- List 5 to 8 criteria. Each must be something a candidate's record can show or not show.
- Mark each as MUST-HAVE or NICE-TO-HAVE.
- Assign weights that add up to 100%. Must-haves carry most of the weight.
- For each criterion, write SCORE ANCHORS: one line each for what a 5, a 4, a 3, a 2 and a 1 look like, in terms of evidence a record could contain.
- Use only what is in the brief. If something important is unclear, list it under OPEN QUESTIONS FOR THE CLIENT instead of assuming.

OUTPUT FORMAT (table)
Criterion | Must/Nice | Weight | Score anchors (5 / 4 / 3 / 2 / 1)
Then: OPEN QUESTIONS FOR THE CLIENT (bullets)

ROLE BRIEF:
[paste brief]

Why this holds up: each prompt has one job, a fixed output format and an explicit "do not invent" rule. Scores are tied to anchors, the arithmetic lives in the template, and a separate check reads every draft before a client does.

What a full engagement looks like

Built around the deliverables in your brief: improved prompts, reusable templates, output-quality guidance, documentation and team training.

Week 1

Audit your current prompting

Review the prompts and outputs your team uses today across search launch, candidate profiles, content and fit analysis. Run them the way this page does, so we can see where drift and inaccuracy come from.

Week 2

Build the library

Workflows and templates for each use case, tuned to Handler's tone and formats, with score anchors, a scoring template that does the maths, an output-quality checklist and short documentation.

Week 3

Train the team

Working sessions with your consultants on live searches, then a refinement pass based on what they find in real use.