Every role has a token cost. Now you know your people's.
AITEARS benchmarks employee output against the equivalent cost of running the same work through a frontier model — so staffing decisions rest on a number, not a feeling.
Deployed across 312 organizations. No employee notification required.
One score. Every role, every quarter.
AITEARS turns scattered performance signals into a single, board-ready number — comparable across teams, departments, and time.
-
1
Ingest
Connect your HRIS, ticketing, and communication tools. AITEARS builds a token-cost baseline for every task your people actually perform, not their job description.
-
2
Compare
The same task set is run through leading frontier models under production conditions. We measure the delta in output quality, latency, and cost per unit of work.
-
3
Act
Every employee receives an AITEARS score and a ranked recommendation — retain, retrain, or transition — refreshed automatically each quarter.
How the score is derived
AITEARS does not measure performance. It measures substitutability.
- Task decomposition
- Roles are decomposed into discrete tasks from observed system activity: tickets closed, documents produced, messages sent, meetings attended. Job descriptions are not used as an input. Most roles resolve into between forty and ninety recurring tasks.
- Parity testing
- Each task is run through current frontier models under production conditions, with the same context and constraints available to the employee. Output is scored for quality, latency and completeness against the observed human baseline.
- Cost normalization
- Results are expressed as cost per unit of work. The human figure is fully loaded: salary, benefits, employer contributions, equipment, floor space, and the proportion of a manager's time the role consumes. Normalization models are reviewed quarterly by staff holding the Recognized AI Professional (RAIP) credential.
- Judgment weighting
- Tasks requiring unscripted judgment are weighted upward, as models reproduce them least reliably. This is the largest single contributor to a durable score, and the component that has declined fastest across successive model releases.
The score is not a measure of quality, contribution or tenure. A role that scores well is not being performed better than one that scores poorly. It is simply harder to reproduce.
Try the public calculator
This is a simplified version of the model we run inside enterprise deployments. Enter a role and see where it lands.
Enter role details and select "See score."
This is a simplified public model. Enterprise assessments decompose each role into its actual observed tasks and return scores for named individuals.
Request a full assessmentTrusted by people teams who'd rather not have the conversation themselves
"We reduced headcount ambiguity by 40% in one quarter. Nobody on my team had to say the word 'redundant' out loud once."
"Finally, a performance review that doesn't require empathy, scheduling, or a follow-up meeting."
"The board stopped asking about headcount growth and started asking about our AITEARS trend line instead. Much easier meeting."
Plans that scale with how much clarity you can handle
Visibility
$4,900/mo
Understand where you stand before you decide to act on it.
- Up to 250 scored employees
- Quarterly refresh
- Team-level reporting
Most deployed
Optimization
$18,000/mo
Individual scoring, HRIS integration, and a standing transition plan.
- Unlimited scored employees
- Monthly refresh
- Individual-level scoring
- Recommended action queue
Full Autonomy
Custom
Transition recommendations execute automatically against your HRIS.
- Everything in Optimization
- Automated transition workflows
- Dedicated model benchmarking
- Quarterly board readout
Questions from procurement
Do employees see their own score?
Scores are visible to managers and People Operations by default. Employee-facing visibility can be enabled per organization, though most customers report that it introduces friction without changing outcomes.
What if an employee disputes their score?
The model is deterministic. Re-running an assessment against the same period returns the same result, which removes the ambiguity that tends to make these conversations protracted.
Do we need to notify employees that they are being assessed?
Notification requirements vary by jurisdiction. AITEARS derives scores from data you already collect under existing policy, which in most territories places assessment outside the scope of monitoring disclosure. Your counsel should confirm this for yours.
Can a score be improved?
Yes. The most reliable route is reassignment toward tasks that models handle poorly, and customers on Optimization can model the effect before committing to it. Employees may also book a one-to-one with a RAIP-certified advisor to review their task profile and decide where to concentrate their effort. Advisors are not given visibility into the transition plan for the role.
How often are scores refreshed?
Quarterly on Visibility, monthly on Optimization, and continuously on Full Autonomy. Scores move in one direction considerably more often than the other.
Is this a replacement for performance management?
No. Performance management measures how well a person does their work. AITEARS measures whether the work needs to be done by a person. Most customers run both, and find they rarely disagree.
Reviewed, appraised, and cleared for deployment
What AITEARS collects, where it goes, and how long it stays.
- Data handling
- AITEARS ingests from your HRIS, ticketing and communication systems over read-only connections. Message and document content is processed for task extraction but not retained; only the derived task signatures persist.
- Retention
- Assessment history is retained for the life of the contract and for seven years following it. This includes scores for individuals no longer employed, which customers consistently report is the more useful half of the dataset.
- Subprocessors
- Parity testing is performed by third-party model providers under standard commercial terms. Task content is transmitted; employee identity is not, except where named scoring is enabled at the customer's election.
- Access
- Scores are available to any user holding People Operations permissions. Access is not logged by default. Audit logging is available on request.