Calibrated talent decisions
Assess every person against the same defined bar, so the outcome does not depend on who ran the review.
— Talent & Leadership
Export public employer reviews for your company and two or three competitors you hire against — Glassdoor, Indeed, Kununu or similar — into one table with Company, Source, Date, Rating and the Review text (paste pros and cons into one cell).
2–3 weeks → ~40 minutes
For a full population, not a sample
No coding required
— USE CASES
Assess every person against the same defined bar, so the outcome does not depend on who ran the review.
Every rating traces back to the evidence behind it, so a board or committee can interrogate it.
Produce the same depth of output for the whole population, not only the most visible names.
— HOW IT BEHAVES
The mechanics behind this specific template — what it reads, what it calculates, and where a human stays in the loop.
Current external sources on the talent population are researched during the run rather than recalled from training data, and every source travels with the output.
Each row of your export is processed on the same basis, so no person is skipped however long the table is.
Movement across the talent population is visualised from the computed data, so the trend is legible at a glance.
Findings on the talent population are written up as a document that reads like professional output, with each claim tied back to a person.
Detail rows are summarised into the grouped view of the talent population without losing the underlying people.
Judgements about the talent population are usually made from evidence spread across assessments, reviews and conversations, with the bar shifting between assessors. Calibrated talent decisions is the typical trigger — assess every person against the same defined bar, so the outcome does not depend on who ran the review. Get it right and the conclusion holds up in the room; get it rushed and it gets picked apart. Either way it costs roughly 2–3 weeks of experienced attention.
Skillsize turns that work into a Skill: you supply the material, and what comes back is a calibrated view of each person with the rationale shown, with a population-level read a talent committee can act on. What sits between input and output is the codified method: thresholds, sequencing and the points where a human confirms a call — all of it visible and editable in the Skill. The practical effect: 2–3 weeks of manual work becomes a ~40 minutes run, held to an identical standard on the tenth engagement as on the first.
Employer brand & competitor benchmark exports as a structured SKILL.md file and is MCP-ready, so the same method runs in ChatGPT, Claude, Copilot or your own AI products. Adapt it to your methodology, and the intelligence stays yours — not locked to one vendor.
Absence & burnout early warning
Spot the teams running hot before people go off sick — outliers today, projected absence six months out.
Attrition risk radar
Learns from who has already left and who stayed to score retention risk across the workforce, showing what actually differs between the two groups and the retention plan that follows.
High-performer profile finder
Learns what separates your strongest people from the rest, ranks the factors that genuinely predict performance, and turns the result into a hiring profile and a development profile you can act on.