Admins and leadership · Thresholds and model versions
Control the cut-offs and the model, with a safety gate
Admins decide how strict the shortlist is and which model is in use. Every model is tested against plain rules before it can be switched on, and every change is kept in history.
The problem
AI tools change behaviour without warning, and nobody inside the company can set how cautious they are or undo an update that made things worse.
How it works
- 1Set the shortlist and review cut-offs, with a reason. Each change is a new version and applies to the next run.
- 2Compare trained models on test tasks they never saw: best person in the top five, best person first and confidence error.
- 3Switch to a model or roll back in one click. A model that ranks worse than the current one is refused unless you override it on purpose.
- 4Past runs keep the model and cut-offs they used.
What's included
Your own thresholds
Set the shortlist and review cut-offs for your company, with a reason. Every change is versioned and audited.
Model versions with a safety gate
Switch to a newer model, or roll back, in one click. A model that ranks worse than the current one is refused unless you override it on purpose.
Where teams use it
- For resource managers in it services and consultingStaff an open task with the right personRank everyone who can actually take a task, with a probability and a cited reason for each person, and keep the final call with the manager.
- For leadership, hr and complianceMake staffing decisions you can audit and defendKeep a full trail of every recommendation, test recommendation rates for fairness, and never use protected attributes.
Questions about thresholds and model versions
Can a worse model be switched on by accident?
No. A model that ranks worse than the current one on held-out tests is refused unless an admin overrides it on purpose.
Do threshold changes rewrite old results?
No. A change applies to the next run; past runs keep the cut-offs they used.
More of FitRank
- Health and fairness dashboardsSee how often managers agree with the shortlist, override rates by week and model, real placement outcomes, and recommendation rates tested for fairness.
- Roles, sign-in and auditAdmin, resource manager, HR and viewer roles, hashed passwords and short-lived tokens, row-level security per company, and an audit CSV of every decision.
- Plain-language task intakeType the role you need in plain words. FitRank turns it into skills, level, dates and location, and you confirm every field before matching runs.
Be one of the first teams to staff with FitRank
We're taking on a small number of pilot companies. Pre-register for early access, or talk to us about an annual licence.