Screen 900+ job applications without a recruiter: a CFO's AI setup
A 30-person firm's CFO got 930+ applications in 36 hours. Workable's AI ranked them. One screening question cut about 20% before a person read a resume. The vendor says the bottom 80% needed no manual resume review.
Who this is about: 30-person company whose CFO also runs hiring
What it takes
- Cost
- From $299/monthStandard is $299/month for 1-20 employees ($3,588/year). Premier is $599/month. Paid accounts start with 3,000 AI credits. A free trial is offered. Pricing page read Oct 24, 2026.Price checked Oct 24, 2026 on workable.com. Prices change; check before you buy.
- Setup time
- An afternoonOur estimate from the source, for one business.
- Skill needed
- No codeApps and settings only.
Explain it simply
Workable is a place to post a job and collect every application. Its AI scores each application against the requirements you wrote, so you start at the top of the list. A question on the form can drop people who miss a must-have before anyone opens a resume. You still decide who moves forward, and you should tell applicants that AI will read their materials.
How can a small business screen hundreds of job applicants with AI?
Put one knock-out question on the job post, then let an applicant tracker's AI grade each application against the job's requirements so you read only the top tier. CacheFly's CFO, who also runs HR, got 930+ applications in 36 hours; one question "eliminates about 20% of applications automatically" (Workable case study, 2026).
| Workflow / use case | Primary AI role / tool slot | Input data | Expected business output / time saved |
|---|---|---|---|
| Screening a flood of job applications | Workable AI Recruiting Agent (grading, ranking) + one screening question | Job requirements, applications, one must-have question | A ranked shortlist; bottom 80% need no manual resume read (vendor case) |
How does this differ from other ways to do this job?
Workable suits a company that hires several salaried roles a year and wants one system for postings, scoring and interviews. Homebase's Hiring Assistant is built for hourly shift workers and links to scheduling and payroll. ChatGPT connected to a jobs inbox costs least but has no applicant tracking, audit trail or job-board posting.
What was done
CacheFly, a company of about 30 people, had its CFO handle hiring. One role drew 930+ applications in 36 hours.
They put one must-have question on the application. Workable says that question eliminates about 20% of applications automatically. That cut is the question, not the AI score.
Workable's AI Recruiting Agent then graded and ranked the rest against the job requirements. Workable's own summary says the bottom 80% needed no manual resume review.
A person still chooses who advances. Tell applicants that AI will read their materials, and check local hiring rules before you turn scoring on. In New York City, Local Law 144 covers some automated employment decision tools.
How it went
The source is Workable's own case study, dated Aug 19, 2026. It describes the flood of applications and the screening setup. It does not give an outside count of hours saved.
- Workable writes that the screening question "eliminates about 20% of applications automatically."
- Keep a person on every rejection. Do not let a score email a "no" on its own.
Prompt
- Context / Role
You are the owner or the one person who handles hiring at a company under 50 people, with an open role that drew more applications than you can read.
- The copy-paste prompt
PromptWritten by our editors from the source You are my hiring screener for the role [JOB_TITLE] at [COMPANY_NAME]. Must-haves (an applicant missing any of these is "Not a fit"): [MUST_HAVES] Nice-to-haves: [NICE_TO_HAVES] Knock-out question on the form and the answer we need: [KNOCKOUT_QUESTION] For each application below, return one table row: | Name | Knock-out answer | Must-haves met (list) | Missing | Nice-to-haves | Score 1-10 | One-line reason quoting the application | Rules: - Judge only on job-related evidence written in the application. Ignore name, age, gender, photo, nationality, school prestige and gaps in employment. - Quote the exact line that earned each must-have. If you cannot find it, mark it Missing; do not guess. - Sort by score. Put "Not a fit" rows at the bottom with the reason. - Do not write rejection emails. I decide who advances and I review every "Not a fit" before anyone hears back. - End with: the 3 strongest applicants, one question to ask each in a phone screen, and any application you were unsure about. Applications: [APPLICATIONS]- Variables to customize
- [JOB_TITLE]
- [COMPANY_NAME]
- [MUST_HAVES]
- [NICE_TO_HAVES]
- [KNOCKOUT_QUESTION]
- [APPLICATIONS]
- Expected output / next step
A ranked table with quoted evidence and phone-screen questions. Next step: the owner advances people in the applicant tracker or a sheet. Do not send a rejection until a person has reviewed it.
Written by our editors for any AI chat. It is not a built-in Workable prompt. A person decides who advances.
Setup notes
- Write the must-haves and one knock-out question before you post the job.
- Collect applications in Workable so they are not scattered across inboxes.
- Let the AI rank people against those written requirements.
- Read the top of the list yourself. Review every "not a fit" before anyone hears back.
- Tell applicants that AI will read their materials, and check rules such as NYC Local Law 144 if you hire in New York City.
Business fit
Who this helps, and how:
Dental office hiring a receptionist: put one must-have, such as a license or a shift, on the form, then rank the rest.
Local & serviceAgency hiring a remote designer: score applications against a written brief and interview only the top of the list.
AgenciesTrades firm hiring an office manager: screen for the software and hours you actually need before a manager reads resumes.
Local & service
Related
- Routines: daily and weekly runs that keep work movingBot structure
- Evidence grade CShortlist 548 job applicants with ChatGPT, then interview the top 20 yourselfVendor or secondhand, using ChatGPT
- Evidence grade CScreen hourly applicants overnight so good ones don't take another job firstVendor or secondhand, using Homebase
- Evidence grade DA Small Recruiting Team Made Sure Every Applicant Gets a Reply With n8n and ChatGPTLight evidence, using n8n, ChatGPT and Airtable
- AI scheduling and paperwork automationDepartment hub: AI for operations, 57 examples
Similar tools
Other tools for this job, and how they differ. Chosen by our editors, never paid for.
- ChatGPT
Ranks a pasted pile for one role. It does not post jobs or keep an applicant record.
See it in use: Shortlist 548 job applicants with ChatGPT, then interview the top 20 yourself
- Homebase
For hourly teams. Hiring ties into scheduling and payroll. The AI assistant is a paid add-on.
- n8n
Workflow automation with AI steps you can self-host. One page here uses it on a jobs inbox.
See it in use: A Small Recruiting Team Made Sure Every Applicant Gets a Reply With n8n and ChatGPT
“Vendor or secondhand” (grade C) means: A tool maker's case study, a promotional demo, or someone else's write-up. How we grade. Spotted a mistake? Suggest a correction.
