Evaluation transparency

How Talent Autopilot evaluates a match

Five short answers explain what the system does, what the score means and where people stay in control.

The score supports a decision. It is not the decision. A person decides what happens next.

Last reviewed against the live evaluation flow25 August 2026

The essentials

Five things to know.

Open any section for more detail.

What does a Talent Autopilot evaluation do?

An evaluation compares the requirements of a job or project with the information in each candidate profile. It helps recruiters decide which profiles to review first.

When there are many candidates, the system may select a smaller group for full evaluation. Candidates who are not selected have not been rejected or given a low score.

How candidates are evaluated

An evaluation can happen in two stages. When there are many candidates, Talent Autopilot may first select the profiles that appear most relevant. It then evaluates the selected profiles in more detail.

Select profiles for full evaluation
Talent Autopilot uses the available job and profile information to identify a smaller group that appears most relevant. This step helps recruiters work through a large candidate list.
A candidate who is not selected at this stage has not been rejected or given a low score. The candidate has not yet received a full evaluation.
Evaluate the selected profiles
Talent Autopilot compares each selected profile with the confirmed requirements and produces a score and explanation. Recruiters can also choose to run this full evaluation for every candidate.
A profile may still receive no score if its information cannot be evaluated or a technical error occurs. Recruiters can review the information behind a completed result, investigate anything unclear and run the evaluation again.

What information does it use?

It uses job requirements confirmed by the recruiter and relevant information from the candidate profile, such as skills and work experience. Talent Autopilot does not independently verify the profile information.

The system can only assess the information recorded in the profile. Missing information does not mean that a candidate lacks a skill or experience.

Exactly what information is used

The full evaluation compares confirmed job or project requirements with information stored in specific candidate-profile fields. Other written descriptions can help the system select profiles or check whether an evaluation can proceed, but they do not directly determine the final score.

From the job or project

Recruiter-confirmed skill requirements
Confirmed minimum-experience and leadership requirements, when provided
Must-have and preferred labels on confirmed skills
An explicit, confirmed education requirement when one is relevant
How much weight the recruiter chooses to give skills compared with experience

From the candidate profile

Recorded skills and technologies
Technologies, dates, length of experience and level of seniority recorded in work and project entries
Recorded education, but only when the job or project has a relevant education requirement

How masking changes what is seen

When masking is active, the system hides or reduces identifying information before it assesses the profile. This can limit exposure to information that may reveal a candidate’s identity, but it cannot guarantee an unbiased process. Access to the hidden information remains controlled and recorded.

How is the score calculated?

The system assesses the candidate’s skills and experience separately, then checks them against the requirements confirmed by the recruiter. Education affects the score only when the recruiter has confirmed that the job or project requires it.

See the scoring details

The system first scores skills and experience separately. It scores education only when the recruiter has confirmed an education requirement. It then combines the applicable results into one score from 0 to 100.

Step 1

Score each relevant part

70%Evidence assessment

30%Requirements check

70% evidence assessment + 30% confirmed-requirements check

The evidence assessment considers the relevant information in the candidate profile. The requirements check measures how many of the recruiter-confirmed requirements that information covers.

If there is not enough information to check the confirmed requirements, the system uses only the evidence assessment. It does not make up a value for the missing check.

Step 2

Combine the parts

No applicable education requirement

Skills and experience share the full score in the ratio selected by the recruiter.

Equal-balance example

Skills: 50%
Experience: 50%
Education: 0%

Applicable education requirement

Education gets a fixed 20%. Skills and experience share the remaining 80%.

Equal-balance example

Skills: 40%
Experience: 40%
Education: 20%

Must-have requirements count more than preferred requirements in the requirements check.
An education result does not remove a candidate or limit the candidate’s total score.
A clearly lower qualification is marked as below the requirement. Missing education, or education the system cannot read, is marked as unverified instead.

See a short worked example

This fictional example shows a backend engineer position with no education requirement. Skills and experience contribute equally to the final score.

Illustrative example: not a real candidate or a promised result.

Backend engineer requirements

Five technologies, including TypeScript and PostgreSQL, are confirmed must-have requirements
Relevant experience is assessed from structured technology and date fields
There is no education requirement, and skills and experience have equal weight

Recorded candidate profile

Three of the five required technologies are recorded
Structured technology and date fields are recorded on two project entries
Written descriptions of the migration work do not directly affect this sample score. The profile does not record any leadership of on-call work, so a recruiter would need to check that point.

Illustrative arithmetic

Skills
74
Evidence assessment: 80 · confirmed-requirements check: 60
70% × 80 + 30% × 60
Experience
76
Evidence assessment: 70 · confirmed-requirements check: 90
70% × 70 + 30% × 90

Overall role-fit score

50% × 74 + 50% × 76

A score of 75 falls in the Strong match range. A recruiter still reviews the profile information and decides what happens next.

What does the score mean?

The score from 0 to 100 shows how closely the information in the candidate profile matches the requirements of this job or project. It is not a confidence percentage. It does not predict success, measure the candidate as a person or tell the recruiter whom to hire. The same candidate can receive a different score for a different job or project.

70–100Strong match
45–69Possible match
25–44Weak match
0–24Low match

These are sorting labels, not hiring decisions. In particular, “Low match” never means automatic rejection.

Why a saved score can change

A saved score reflects the job or project requirements, candidate-profile information, scoring weights and scoring method used at that time.

The role description or confirmed requirements change.
Relevant information is added to or removed from the candidate profile.
The skills-to-experience balance changes.
An applicable education requirement is added, changed or removed.
The scoring method or its rules change.

Any of these changes can make an earlier score out of date. Run the evaluation again before relying on it. Because a new run is a fresh assessment, its result is not guaranteed to be exactly the same.

What happens when the system cannot produce an answer?

If Talent Autopilot cannot complete a valid evaluation, it may ask for human review, skip the profile or show no score. This can happen because information is missing or unsuitable, a requirement needs review, or a technical error occurs. It does not make up a score.

A missing result is not a hidden low score.

Why a result may be unavailable

Sometimes Talent Autopilot cannot complete an evaluation or produce a score. These are the main reasons why.

A requirement needs human review

If a role appears to use a protected characteristic or a proxy for one, scoring pauses and asks a person to review the requirement.

The job or project is not ready

The job or project may not have enough confirmed requirements for a meaningful comparison.

The profile has too little usable information

The system may skip a candidate when the profile does not contain enough relevant information. This is not a rejection or a negative score.

A required score is invalid

If a required scoring response is missing or invalid, the system does not make up a replacement number.

A technical error occurs

A technical error can stop one or more candidate evaluations. If other candidates have valid results, the run finishes with errors and keeps those results. If no candidate has a valid result, the run is marked as failed. The system does not create a replacement candidate score.

If the score is valid but the written explanation cannot be produced, the score may remain available with a message asking the recruiter to review it. The system does not make up an explanation.

Safeguards and human review

Evaluation is one input to a recruiting process. It does not replace professional judgement or the organisation’s responsibilities.

Potentially discriminatory criteria are paused

A role that appears to rely on a protected characteristic, directly or indirectly, is sent back for human review instead of being scored. This is a safety check, not a legal judgment.

People inspect and decide

A recruiter can review the profile information used, investigate anything unclear, change the confirmed requirements or scoring weights, and run the evaluation again. A high score does not mean “hire this person.” A low score does not mean automatic rejection.

Read more about the surrounding controls and data handling:

Security and trust Privacy policy

Questions and answers

Common questions

Short answers to the questions recruiters and candidates most often ask.

Is the score a success probability?

No. It describes role fit from the information available. It is not a prediction that someone will succeed, a confidence percentage, or a hiring recommendation.

Does a low score reject someone?

No. Score bands help organise results. The software does not autonomously make the final hire or reject decision. A person reviews the evidence and decides what happens next.

Does missing information prove that something is absent?

No. A profile can be incomplete. Missing evidence identifies something to verify; it does not prove that the candidate lacks the skill or experience.

Does every candidate receive a detailed evaluation?

Not necessarily in a standard run. A large pool may first be narrowed by preliminary relevance. A deep run does not narrow the chosen pool, but individual profiles can still be skipped or fail before a completed score exists. Neither state is a rejection or negative score.

Can education affect the score?

Only when the role has a meaningful, explicit and recruiter-confirmed education requirement. Otherwise education has zero effect and is not used in the evaluation. Missing or unclear education evidence is unverified, not proof that a requirement is unmet.

What happens when masking is active?

Identity information is reduced before evaluation. This limits what identity signals the evaluator sees, while controlled unmasking remains audited. Masking does not promise a bias-free outcome.

Can the score change?

Yes. Changes to the role, recorded candidate evidence, confirmed requirements, effective weights or evaluator can make an earlier result out of date. Re-run before relying on it as current.

Is customer data used to train shared models?

No. Candidate data is used to serve the specific request and is not used to train shared models. Our Security and Privacy pages explain the surrounding controls and data handling.