The opaque score
“Trust the 87%” is not a defence.
A single composite score with no per-criterion breakdown is impossible to debug. If the hiring manager disagrees, there is no productive conversation — only a fight with the algorithm.
AI resume shortlisting
AI resume shortlisting that shows its reasoning.
Most AI shortlisting tools hand you an opaque match score and call it a day. ShortlistTable gives you per-criterion AI verdicts, the resume sentence behind each one, and a one-click recruiter override — so the shortlist survives the hiring-manager debrief, the legal review, and the next quarterly audit.
No credit card · Manual override on every cell · No auto-reject
shortlisttable.app / tables / ai-shortlist-q2Auditable
| Candidate | Fit | Evidence | Confidence | Override? | |
|---|---|---|---|---|---|
Alice Chen Ledger Platform Eng · 7y | Strong | Resume p.1 | 0.91 | — | Call first |
Marco Silva Cloud Arch · 12y | Partial | Resume p.2 | 0.64 | Pending | Review fit |
Nina Patel Full-Stack Dev · 4y | Weak | No match | 0.22 | — | Hold |
Citation · Marco Silva · “Migrated payments backend from monolith to event-driven services on AWS” · resume p.2Source-backed
3-state
Yes / No / Needs review
1-click
Recruiter override
0
Auto-rejections
100%
Verdicts cite a sentence
Why most AI rankers fail
An opaque 87% match score is not a decision.
When the hiring manager pushes back on a rejected candidate, you need to point to a specific reason. When legal asks why someone was screened out, you need a paper trail. When the same pile is re-run a week later, the order should be the same — not subtly different because the model drifted.
Most AI shortlisting tools provide none of this. They produce a leaderboard, ask you to trust it, and leave you with nothing to push back against when it is wrong. We take the opposite approach: every verdict carries the criterion, the source sentence, and a one-click override.
The auto-reject button
Hidden screen-outs are a legal liability.
Tools that auto-reject candidates below a threshold compound bias problems silently. We never auto-reject. The bottom of the list is a hold queue with evidence intact, not a delete queue.
The non-deterministic run
Same pile, different shortlist.
Many AI rankers produce different results across runs because the underlying model output is non-deterministic. Ours produces the same verdicts on the same input every time, so your shortlist is reproducible.
How auditable ranking works
The four properties of a defensible shortlist.
01 · Feature
Per-criterion verdicts
Each candidate is judged column-by-column. The shortlist position is derived from those judgments, not the other way around.
Yes · 5 yrs seniorReview · domain fitNo · stack mismatch
02 · Feature
Source sentence per cell
Every verdict cites the resume sentence behind it. If the citation is wrong, you adjust the column. If the verdict is wrong, you override the cell.
Resume p.1 · Experience“Led platform team of 6 engineers building event-driven services on AWS.”
03 · Feature
Confidence + review queue
Below your confidence threshold, verdicts go to a review queue instead of being auto-decided. The recruiter handles the ambiguous cases, the engine handles the obvious ones.
31
candidates in
review queue
04 · Feature
One-click override
Disagree with a verdict? Click the cell, pick a different value, add a one-line reason. The original AI verdict stays in the audit history.
AI: review · 0.62
Override: yes (manual)
05 · Feature
Review priority, not ranking
We sort by review priority, not by composite score. Humans decide who gets called first; the engine just orders the work.
Call firstReview fitHold
06 · Feature
No auto-reject
There is no “delete” bucket. Low-fit candidates land in a hold queue with evidence intact, so any rejection is reviewable rather than silent.
Hold queueAudit trailRe-runnableReviewer override
How it compares
AI shortlisting vs the alternatives.
| Property | Black-box AI ranker | ATS rank column | Spreadsheet | ShortlistTable |
|---|---|---|---|---|
| Per-criterion verdicts | ✕Composite score only | ✕Single rank | –Manual | ✓Typed verdicts per column |
| Source sentence cited | ✕Rare | ✕No | –Manual notes | ✓Every cell |
| Recruiter override + audit | ✕Limited | –Edit only | ✓Yes | ✓Full audit trail |
| Auto-reject below threshold | ✕Common | ✕Common | ✓Never | ✓Never — hold queue only |
| Reproducible runs | ✕Non-deterministic | –Depends on ATS | ✓Manual | ✓Deterministic |
✓ Supported– Partial / manual✕ Not supported
FAQ
Common questions on AI shortlisting.
Will this replace the recruiter?+
No. The recruiter still makes every hire/reject decision. ShortlistTable surfaces the per-criterion evidence so the recruiter can review faster — not so the recruiter can be skipped.
What if the AI gets a verdict wrong?+
Override the cell in one click. The original verdict stays in the audit history; your override becomes the active value and shows up in every export. There is no retraining loop you have to manage.
Is the ranking reproducible across runs?+
Yes. Same inputs, same column definitions, same verdicts — every time. This matters for audit trails and for any retrospective on a hire that did or did not work out.
How do you handle bias?+
Two design choices: (1) every verdict surfaces the source sentence, so systematic mismatches become visible to the recruiter rather than hidden inside a model; (2) there is no auto-reject — the bottom of the list is a hold queue with evidence intact, not a delete queue.
Does the ranking change if I add or remove a column?+
Yes — the review priority is derived from the per-column verdicts, so changing the columns changes the ordering. Saved column templates make this easy to manage across roles.
Can hiring managers see the evidence too?+
Yes. The CSV / XLSX export includes evidence columns next to verdict columns, and you can invite hiring managers as workspace viewers for the interactive view.
More from us
Keep exploring.
Related
Source-backed screening
Every verdict points to a sentence.
Related
Candidate ranking
Review priority, not leaderboard.
Related
Custom screening columns
Plain-English columns, typed outputs.
Free tool
Try the free JD ↔ CV matcher
See how matching works on one candidate.
Read
Keyword vs semantic screening
What actually changes for recruiters.
Defensible by design
Build a shortlist you can defend in a debrief.
Try ShortlistTable on a batch of up to 25 resumes — no credit card required, no auto-reject, full recruiter override.