What problem does it solve?
Most recruiting effort goes into candidates who have already applied to an open role, while a company's applicant tracking system quietly accumulates thousands of past applicants, people good enough to interview for a different role, who were never hired and are never looked at again. A recruiting team facing a hard to fill role, such as a security engineer in a location with sparse talent, defaults to posting the job and waiting, or to an external search that costs money and takes weeks, without first checking who the company already knows.
Career sites have the same problem from the other direction: most visitors never apply, and the content, form length and channel mix that would convert more of them into applicants is rarely tested or optimized, so sourcing spend goes to buying more traffic instead of converting the traffic already there.
How does it work?
- Build a living talent pool. The system reads every past applicant, employee referral and sourced profile already in the applicant tracking system and infers each person's skills, experience and likely fit for the roles the company hires for, refreshed as new applications and outcomes come in.
- Match against open roles. When a role opens, it ranks the existing pool and any newly sourced candidates against the role's requirements, and surfaces the recruiter's best matches, including people who applied for a different role months or years earlier.
- Optimize the front door. For inbound traffic, it tests and personalizes career site content and application flow to convert more visitors into applicants, and flags where the funnel is losing people.
- Recruiter decides, agent drafts the approach. The recruiter reviews the suggested matches and decides who to contact; the system can draft a personalized outreach message for the recruiter to send, referencing the role and the candidate's relevant background.
- Audience
- Employee facing
- Autonomy
- Assist
- Adoption
- Early adopters
- Channels
- Internal tools
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Cycle time reduction | Too few to pool | 16% | 1 | 1 organization |
| Productivity gain | Too few to pool | 30% | 1 | 1 vendor |
Value drivers: Employee productivity, Speed and cycle time, Lower cost to serve.
Indicative value
A company that fills 500 roles a year and processes 40,000 applications a year
USD 15,000 to USD 157,500
Recruiter time cost avoided from rediscovering existing candidates per year
How this is calculated
Formula: rolesPerYear * rediscoveredShare * hoursSavedPerRediscoveredHire * recruiterHourCost. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Roles filled per year rolesPerYear, roles per year | 500 | 500 | The reference company. |
| Share of roles filled from the existing applicant pool instead of new sourcing rediscoveredShare, fraction of roles | 0.05 | 0.15 | Editorial assumption, replace with your own applicant tracking system data. |
| Recruiter hours saved per role filled by rediscovery instead of a fresh external search hoursSavedPerRediscoveredHire, hours per hire | 15 | 30 | Editorial assumption. Box reports a 16% reduction in time to hire after adopting AI sourcing and rediscovery, https://eightfold.ai/wp-content/uploads/3Sixty-Insights-Anatomy-of-a-Decision-Eightfold-Box.pdf, but does not publish an hours figure, so this input is not taken directly from that source. |
| Fully loaded cost of a recruiter hour recruiterHourCost, USD per hour | 40 | 70 | Editorial assumption. |
What it leaves out: Counts only recruiter time saved on roles filled from the existing pool instead of a fresh search. It leaves out any reduction in job board or agency spend, the value of a faster time to hire itself, the software cost, and any effect on candidate quality or diversity, which the evidence describes qualitatively but does not quantify consistently.
Who already uses it?
2 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Forvia
France · Manufacturing · 2024
Forvia, an automotive technology supplier transitioning from a traditional manufacturer, deployed Eightfold's Talent Acquisition platform to find digital talent and increase sourcing efficiency across its global recruiting organization, opening new sourcing channels and improving applicant quality and diversity.
- Productivity gain: 30%
"30% productivity gains in sourcing"
Claimed by: vendor
Box
United States · Technology and software · 2023
Box, the cloud content management company, began evaluating vendors in 2021 and adopted Eightfold's AI enabled Talent Acquisition solution to integrate with its Greenhouse applicant tracking system and resurface past applicants already in it, instead of relying only on new inbound and outbound sourcing, aiming for more speed and agility in hiring, including for hard to fill roles such as security engineers.
- Cycle time reduction: 16%
"Box reports having reduced time-to-hire by 16 percent."
Claimed by: organization - Employee adoption: 95%
"Ninety-five percent of the recruiting team at Box is utilizing the Eightfold tool."
Claimed by: vendor
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Historical applicant records in the applicant tracking system, with enough structured data to match against new roles
- Career site analytics to see where visitors drop off before applying
- A policy on how long a past applicant's data is retained and how recontact consent works
Systems to integrate
- Applicant tracking system, for historical applicants and open roles
- Career site or job board content management
- Email or messaging tool the recruiter already uses for outreach
Complexity: Medium
The matching technology is available off the shelf and integrates with common applicant tracking systems. The harder part is data quality in years of historical applicant records, and getting recruiters to change from reactive sourcing (working whichever req is open) to proactively building and maintaining pipelines, which is a workflow and incentive change, not just a tool rollout.
- 1
Clean and connect the historical pool first
Rediscovery is only as good as the applicant history behind it. Confirm what is in the applicant tracking system, how old it is, and whether recontact consent and retention rules allow it to be used before switching the matching on.
- 2
Pilot on the hardest roles to fill
Roles where talent is genuinely sparse, such as security engineering in certain locations, are where an existing applicant pool has the clearest advantage over a fresh external search: fewer qualified candidates means every past applicant is worth checking. Box, for example, resurfaced a past applicant to fill an open security engineer role this way.
- 3
Give recruiters a dashboard, not just a ranked list
Box's team built a dashboard to track hires resulting specifically from resurfacing candidates in the applicant tracking system, and separately benchmarked recruiter adoption of the tool, so return on the tool could be measured rather than assumed from general activity.
- 4
Let AI draft outreach, keep a human sending it
A drafted message referencing the candidate's real background saves time over a blank page, but a recruiter should read and personalize it before it goes out, especially to someone who applied once and was rejected.
- 5
Track quality, not only speed
Faster time to hire is not the only goal. Track interview to offer ratio and early tenure outcomes for rediscovered hires against newly sourced ones so speed is not gained at the cost of fit.
Guardrails
- Recruiters decide who to contact and what outreach to send; the system suggests and drafts, it does not contact candidates on its own
- Past applicants can request their data be removed from the pool, honored promptly
- No protected characteristics or obvious proxies used as ranking features
- Regular fairness testing of who gets surfaced and contacted, by group
KPIs to instrument
- Time to hire and time to fill for roles sourced from the pool versus newly sourced roles
- Share of hires that come from the existing applicant pool
- Recruiter adoption of the tool as a share of the recruiting team
- Candidate response rate to AI drafted versus recruiter written outreach
Human in the loop
Recruiters review every suggested match and outreach draft before contacting a candidate; hiring managers make the actual selection decision as they would for any candidate. HR or talent acquisition leadership owns the matching model's fairness testing and the data retention policy.
Common failure modes
- Stale profiles surface a candidate who is no longer available or interested
- A profile from years ago may no longer reflect the person's skills, interest or availability. Confirm interest before advancing someone the system resurfaced, and refresh the pool from outcomes over time.
- Rediscovery narrows instead of widens the pool
- Always pulling from the same historical pool can reduce diversity of the candidate slate over time. Track the mix of rediscovered versus newly sourced candidates and set a floor for fresh sourcing.
- Outreach feels templated
- A drafted message that is not personalized reads as spam and damages the employer brand. Require a recruiter to edit, not just approve, before sending.
What are the risks and rules?
EU AI Act
High risk
Annex III point 4(a) lists AI systems intended to be used to recruit or select natural persons, including to place targeted job advertisements and to analyse and filter applications. Ranking a company's own past applicants and new candidates against an open role, and deciding who a recruiter sees first, is filtering and evaluating candidates for that purpose.
Rules that apply
Guidance
- Annex III, high risk AI systems referred to in Article 6(2) (European Union, Europe). Point 4(a) covers recruitment or selection of natural persons, including analysing and filtering applications and evaluating candidates.
- Automated Employment Decision Tools, Frequently Asked Questions (New York City Department of Consumer and Worker Protection, North America). Local Law 144 requires a bias audit, published results and candidate notice before an employer uses a tool that substantially assists or replaces discretionary decision making. It applies only when the tool screens a candidate who has applied for a specific position; sourcing people who have not applied for a specific role falls outside its scope.
Controls to put in place
- Bias audit of the ranking and matching model before use in a jurisdiction that requires one
- Documented data retention and recontact consent policy for historical applicants
- Named recruiting operations owner for the model and its fairness testing
- Human review of every suggested match and outreach draft before a candidate is contacted
Frequently asked questions
- What results have companies reported from AI sourcing and rediscovery?
- Box reported, in a 3Sixty Insights report published by Eightfold, a 16% reduction in time to hire; the same report separately notes that 95% of Box's recruiting team was using the tool, and that Box resurfaced a past applicant to fill a hard to fill security engineer role. In a separate Eightfold case study, Forvia reports a 3.5 times increase in visitor to applicant conversion on its career site and 30% productivity gains in sourcing.
- How is this different from resume screening software?
- Screening software ranks people who have already applied to a specific open role. Sourcing and rediscovery works the other direction: it searches a company's own historical applicant pool and external sources for people who fit a role before, or instead of, waiting for them to apply.
- Does the AI contact candidates on its own?
- It should not. Neither the Box nor the Forvia source describes the review workflow in detail, but as recommended practice, a recruiter reviews every suggested match and any drafted outreach message before a candidate is contacted: the AI proposes, the recruiter decides and sends.
- Is AI candidate sourcing high risk under the EU AI Act?
- Yes. Annex III point 4(a) covers AI used to recruit or select people, including filtering and evaluating applications, which is what ranking and surfacing candidates for a recruiter does.
How to cite this page
Blits.ai AI Use Case Library, "AI agent for candidate sourcing and talent rediscovery", last verified 29 September 2026, https://www.blits.ai/ai-use-cases/ai-candidate-sourcing-and-talent-rediscovery. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 29 September 2026: First published