AI resume screening software built for recruiters who refuse to ship a shortlist they can’t defend. Per-criterion AI verdicts, the resume sentence behind each verdict, one-click override on every cell. The AI does the reading. You make the call.
| Candidate | Kafka in prod | On-call ≥ 12mo | B2B SaaS | 5 yrs senior | Review priority |
|---|---|---|---|---|---|
Alice Chen Ledger Platform Eng · 7y | Yes | Yes · 18mo | Yes | Yes | Call first |
Marco Silva Senior Cloud Architect · 12y | Review | Yes · 6y | Review | Yes | Review fit |
Nina Patel Full-Stack Dev · 4y | No | No | Review | No · 4y | Hold |
Tomás Almeida Backend SRE · 9y | Yes | Yes · 24mo | No | Yes | Review fit |
A leaderboard is not a screening tool. A composite fit score with no per-criterion breakdown is not defensible in a debrief. A keyword scan of the resume is not a meaningful read of the candidate’s career.
The job of resume screening software is to make a careful, human review possible at the speed your pipeline demands. Not to replace the recruiter. Not to delete candidates the model misunderstood. Not to compress the candidate down to a single number you cannot push back against.
Most ATS systems still rank by literal keyword presence. A candidate who built a Kafka cluster but called it “an event-streaming layer” in the resume is filtered out. The ranking column is treated as authoritative by hiring managers who have no way to audit it.
A manually maintained shortlist sheet is fine for the first five roles. By the tenth role the columns drift, the must-haves stop being consistent, and you cannot tell which judgments were made when. The audit trail is “trust the recruiter’s memory”.
The newer wave of AI screening tools gives you a fit number and asks you to trust it. When the verdict is wrong, you cannot push back. When the hiring manager disagrees, the conversation becomes about the algorithm instead of about the candidate.
Upload a folder of PDFs and DOCX files, paste a single resume, or connect a Greenhouse pull. No ATS migration; we never become the system of record for your candidate data.
Each column is a plain-English question (“Has the candidate run Kafka in production?”), not a prompt. You decide what matters for this role. Save column templates per role family.
Every resume gets a verdict per column with the source sentence we matched against. Confidence below threshold goes to the review queue instead of being auto-decided.
CSV / XLSX presets for Greenhouse, Ashby, Lever, Workable, or a hiring-manager sheet. Evidence and citations are exported as columns, not buried in notes.
ShortlistTable is built around the things that make a screening sheet trustworthy: per-criterion verdicts, source citations, recruiter-editable cells, and exports that survive the round-trip into any ATS.
Every screening column produces a yes / no / needs-review verdict. No composite mystery score. Disagreements happen on a single criterion, not on the model.
Every verdict links to the exact sentence in the resume that supports it. Click any cell to see the highlighted PDF region. No verdict without evidence.
Plain-English columns, typed outputs (yes/no, enum, number, date), reusable templates per role family. You write the question, we read every resume.
Confidence below your threshold lands in a queue, not the bin. The recruiter reviews the ambiguous cases — and only those — in a focused panel.
CSV / XLSX presets for the major ATSs. Evidence columns survive the round-trip. Hiring-manager preset has colour-coded cells for at-a-glance review.
Override any verdict in one click. The original AI value stays in the audit trail; your manual edit becomes the active value. Every export reflects the override.
| Capability | ATS rank column | Spreadsheet | Black-box AI ranker | ShortlistTable |
|---|---|---|---|---|
| Per-criterion verdicts | ✕Single rank score | –Manual, drifts over time | ✕Composite score | ✓Typed verdicts per column |
| Source citation per cell | ✕None | –Manual notes only | ✕Rare | ✓Sentence + page reference |
| Recruiter-editable verdicts | –Limited | ✓Yes | ✕No | ✓Yes — full audit trail |
| Custom screening columns | –Limited | –Yes, but manual | –Limited templates | ✓Plain-English questions |
| Confidence + review queue | ✕No | ✕No | –Sometimes | ✓Per-column threshold |
| ATS-shaped export presets | ✓Native | ✕Manual mapping | –Generic CSV | ✓Greenhouse, Ashby, Lever |
| Setup time | ✕Weeks | ✓Minutes | –Days | ✓Minutes |
You are hiring a backend engineer with Kafka experience. You have 86 inbound resumes from a LinkedIn post, a referral push, and a recent agency submission. The hiring manager wants to see a shortlist on Friday.
The naive approach is to read all 86 resumes against the same five must-haves, copy fields into a sheet by hand, and hope your notes from resume #1 are still consistent with your notes from resume #86. They will not be. By the end of the pass, “senior” means different things to past-you and current-you, and the hiring manager finds out during the debrief.
With ShortlistTable, you drop the 86 PDFs into a table, write your five must-haves as plain-English columns, and run the table. Each resume gets a verdict per column — yes, no, needs-review — and a source sentence per verdict. The 22 candidates with all must-haves green show up at the top of the sheet, in a “call first” queue. The 31 candidates with at least one needs-review land in a focused review panel — these are the ones you read by hand. The remaining 33 sit in a hold queue with their evidence intact, so the rejection is reviewable rather than silent.
You spend the rest of the morning on the 31 review-queue candidates and the top of the “call first” queue. By lunch you have a shortlist of 12 names with a per-criterion grid, an evidence sentence per cell, and a CSV export ready for the hiring manager.
The work the recruiter actually does — disagreeing with a verdict, overriding a cell, asking the candidate a sharp question on the screening call — does not change. What changes is the amount of time spent doing the parts that did not need a human in the first place.
Submit client-ready shortlists faster than the agency next door.
Per-client workspace boundaries, client-branded submission packs.
Per-search rubrics for the bespoke must-haves of retained search.
The missing layer between “resume in inbox” and “interview scheduled”.
Confidence thresholds, parallel batching, focused review queues.
Stack-aware extraction, title-vs-scope checks, public artifact recognition.
Most ATS scoring outputs a single composite number with no per-criterion breakdown and no source citation. We expose the underlying judgments: one verdict per screening column, the sentence from the resume that supports each verdict, and full recruiter override with audit trail. Disagreement happens on evidence rather than on the algorithm.
No. ShortlistTable sits at the screening step before your ATS. Your ATS stays the system of record for candidate data, stages, and offers. We produce a CSV / XLSX shortlist that imports cleanly into Greenhouse, Ashby, Lever, Workable, Bullhorn, and JobAdder via per-system presets.
PDF, DOCX, TXT, and pasted text are supported in the paid product. The free browser tools (jd-cv matcher, missing requirements finder, bulk resume parser) parse TXT only — for PDFs and DOCX you need the paid pipeline.
A few minutes. Create a table, drop the resume folder, write 5–10 screening columns as plain-English questions, click run. Saving column templates per role family makes the second table for the same role family take under a minute.
You click the cell, edit the verdict, and add a one-sentence note explaining the override. The original AI value remains visible in the audit history; the manual override becomes the active value and is reflected in every export. There is no retraining loop you have to babysit.
Two design choices: there is no auto-reject — the bottom of the list is a hold queue, not a delete queue — and every verdict comes with the source sentence, which makes systematic mismatches visible to the recruiter rather than hidden inside a model. Recruiter override is the final say.
Per-workspace, encrypted at rest, with configurable retention. You can set auto-archive after N days and hard-delete after M days to match your jurisdiction’s requirements. We do not train on customer data.
Per-workspace plans with a monthly resume volume. The free tier covers your first 25 resumes per month so you can validate the workflow before committing. Pricing details are on the pricing page.
Drop 25 resumes, write 5 screening columns, get a sheet. No credit card. No ATS migration. The result is yours to keep whether you sign up afterwards or not.