Open a candidate and the AI's assessment follows your team's workspace (select Evaluation in the bar under their name to jump there): one scoreboard with the score and, beside its title, the verdict, then the evidence for every requirement. Everything on it can be traced back to a line in the candidate's own material.
AI evaluation scoreboard
Target role bar: 55+ ATSProceed with cautionAditya shows solid distributed-systems design from payments work, backed by a clear line on retries and idempotency. Go appears only in internal tooling, and the PostgreSQL experience the role needs could not be verified in the CV, so both must-haves are only partly met.
- Strong (≥70)
- Moderate (45–69)
- Gap (<45)
- Must-have score
- 50
- Preferred score
- 100
- Experience
- 66
- Domain
- 65
- Leadership
- 82
- Communication
- 82
- Technical
- 65
Step 1: ATS composite score
The overall score from 0 to 100. The bar under it shows the score, and the dark tick marks the role bar: the job's minimum ATS score. Target role bar at the top repeats it, next to the verdict badge, for example Proceed with caution.
Step 2: Next step and confidence
What the AI recommends doing next, for example Recommended: interview, and how confident it was in its requirement verdicts on average.
Step 3: Executive summary
Two or three sentences on what is well evidenced and what is missing. They are written from the verdicts and never repeat the numbers.
Step 4: Requirements fully matched
One dot per requirement, filled when it is fully matched. Hover the dots to see each requirement's name and status.
Step 5: Dimension sub-scores
Seven tiles, each with a bar for where it stands: dark for Strong (70 or more), grey for Moderate (45 to 69), coral for a Gap (under 45). The lowest one is marked Weakest, so you know where to start.
Step 6: Role fit risk
How likely the candidate is to struggle in the role, from Low to High. Hover the small info icon to see what drives it.
The verdict
The verdict is worked out from the ATS score by fixed rules, not chosen by the AI:
| Verdict | When |
|---|---|
| Strong match · Recommended: interview | ATS of 80 or more, and no must-have missing. |
| Proceed with caution · Recommended: manual review | ATS at or above the role bar, but not a strong match. |
| Low match · Below the role bar | ATS under the role bar. |
The role bar is the Minimum ATS score in your screening profile, or 55 if you have not set one. See Tune your screening profile.
A Low match is never a rejection. It puts the application in front of a person, who decides. See The AI never rejects anyone.
The dimensions
Four of the dimensions make up the ATS score, in the proportions your screening profile sets. The others add context.
| Dimension | What it reflects | Counts toward ATS |
|---|---|---|
| Must-have score | The share of must-haves evidenced. A partial match counts half. | Yes |
| Preferred score | The same for preferred requirements. | Yes |
| Experience | The years of experience on the application. | Yes |
| Technical | How confidently the requirements are evidenced overall. | Yes |
| Domain | Currently the same overall evidence confidence as Technical. | No |
| Leadership | Leadership signals the AI found in the CV. | No |
| Communication | Whether the AI found a clear signal of how the candidate communicates. | No |
Role fit risk
Risk is High when two or more must-haves are missing or the ATS score is under 55, Medium when the ATS score is under 68, and Low otherwise. Three factors are listed beside it:
- Tenure stability: Low for candidates with eight or more years of experience, Medium below that.
- Skill verification: how well the requirement verdicts are backed by verified quotes.
- Must-have coverage: how many must-haves are missing.
The evidence behind the score
Directly under the scoreboard, AI evidence & requirements lists every requirement with its verdict and the line behind it.
AI evidence & requirements
- GoPartial
Built two internal tools in Go; the core payment services are written in Java and Kotlin.
- PostgreSQLPartial
Describes database work on the ledger service, but the quoted line was not found in the CV. (evidence unverified)
- Distributed systemsMatched
Designed the retry and idempotency layer for a payments queue processing 3,000 events per second.
- KubernetesMissing
The CV does not mention Kubernetes or any container orchestration.
- Distributed systems designDesigned the retry and idempotency layer for a high-volume payments queue.
- Production ownershipRan on-call for the ledger service and led its incident reviews.
- Below the five-year minimumFour years of backend experience against a role that asks for five.
- PostgreSQL at scale not evidencedThe CV does not show production PostgreSQL work at the scale this role needs.
Moderate risk: four years of experience sits below the five years the role asks for, and one must-have rests on evidence that could not be verified.
Each requirement shows Matched, Partial or Missing, with the quote from the CV underneath. On a candidate's page, click a quote to open the CV beside the scorecard with that line highlighted.
Not every line is a quote:
- A line ending in (evidence unverified) is the AI's own sentence. Its quote could not be found in the CV, so the requirement was marked down to Partial. Check the CV yourself before relying on it.
- From Interview feedback or From GitHub profile (or Portfolio, or Website) means the evidence came from your interviewers or from a page the candidate shared, not from the CV.
- A Missing requirement shows the AI's note on what it looked for.
Next to the matrix, or below it on a narrow screen, are the Strengths, the Concerns and gaps, and the risk meter with its factors.
About the candidate
Further down, About the candidate holds everything else the AI read: the career timeline, leadership, projects, skills grouped into matched requirements, gaps and the company values the candidate showed, any shared links, and the answers to your screening questions.
About the candidate
What the AI read in the CV, with evidence, and what the candidate submitted.
Experience
- Software Engineer II3 yrsRazorpay · 2024-01 - PresentPayments platform: webhook processing, retries and the ledger service.
- Software Engineer2 yrsFreshworks · 2022-06 - 2023-12Backend APIs for the ticketing product.
- Led two incident reviews for the ledger serviceDocumented in the CV.
Projects
1- Payments retry queueContributor
Exactly-once handling for payment webhooks, built on idempotency keys.
JavaKafkaRedisImpact: No duplicate charges over two quarters.
Skills
Evidence
Show 2 claims traced to the CV opens the evidence list: each claim, its quote, its confidence and where in the CV it came from.
The verdict appears once
The scoreboard is the only place the verdict is written out, next to the score it comes from. In the candidates list, a coloured dot beside each ATS score stands for it: green for Strong match, amber for Proceed with caution, red for Low match. The candidate's header leaves it out and keeps to who the candidate is and what you can do; once you scroll past it, the bar of sections at the top keeps the name, the ATS score, Reject and Schedule interview in view.
If the job's requirements have changed since the candidate was scored, a Requirements changed since scoring badge appears under the candidate's name. See Requirements, versions and stale scores.