Skip to content
All articles

Recruitment Technology

Candidate Rediscovery: How AI Agents Help Recruiters Work Their Database

24 min read

Make your existing candidate relationships part of the next brief. Explore how agents can prepare evidence and outreach for recruiters to review.

By FluxHire.AI · Published 2 October 2026 · 24 minute read

Large glass data cube linked to smaller task cubes by soft blue light.
Conceptual illustration: rediscovery brings relevant records forward for a fresh review. This is not a product screenshot.

A client sends through a brief on a Tuesday morning and wants a shortlist this afternoon. You know the role and you know the market. Somewhere in the CRM are people you’ve already spoken to who deserve another look. Finding them is the first job. Working out whether what you hold on them is still true is the second, and it’s usually the one that eats the morning.

Candidate rediscovery gives that existing pool a proper place in your recruitment workflow. Instead of treating every brief as a fresh sourcing exercise, you revisit the relevant records, read the history and decide who merits a conversation today. The commercial logic is plain: your agency has already paid for those relationships in consultant hours. Use them.

AI agents can prepare parts of that work for review. They can organise what the records say and draft the next action; a recruiter checks the evidence and decides what happens. That split is the approach FluxHire is building around. What you want from it is a prepared piece of work you can assess in a few minutes, not another database tab to maintain.

This guide covers what that looks like on a desk, what to ask of the software and how to judge the result. It also holds one line firmly: what an old record says and what a candidate has confirmed today are different things.

What is candidate rediscovery?

Candidate rediscovery means searching the people already in your ATS or recruitment CRM against a current brief, then deciding who deserves a fresh conversation. Candidate reactivation is the next step: reconnecting where that’s appropriate. External sourcing looks beyond the pool you already hold. A desk can use all three, but they aren’t interchangeable.

A previous application, interview or phone call gives you context. It doesn’t prove current interest. Someone who wanted a move when you last spoke may have settled in. Someone who turned down a role may have a different brief in mind now. Rediscovery hands the recruiter a question to investigate, not a conclusion about the person.

That matters when you talk about what your database is worth. A record is only useful in relation to a decision. An old suburb, a job title and a CV describe the past. Whether that history helps with today’s role depends on the candidate’s current preferences, capability and permission to proceed.

There’s no single technical definition of a rediscovery product, either. Greenhouse’s July 2026 documentation says its Past candidates view surfaces previous applicants through filters rather than AI match scores, and prospects without a past application don’t appear in that view. Before you accept a broad talent-rediscovery label, ask what the tool actually searches. Greenhouse documentation

AI candidate rediscovery adds software help to the preparation and review. For an agency, the working question is wider than whether somebody applied before. Can you find the relationship history, see why the last process stopped and prepare an informed approach for a different brief? For an in-house team, the equivalent question is whether a past applicant merits a look for a new requisition.

That split between a record and prepared work is the standard to take into any evaluation, ours included. The record stays as evidence to inspect. The agent-assisted process should give the recruiter a useful starting point, and the decision rests on what they’ve reviewed and confirmed.

Which candidate records deserve another look?

Start with records that relate to the new role and have a history you can interpret. Previous finalists, people whose timing was wrong and candidates with related experience are sensible groups. Check exclusions, live processes and consultant ownership before anyone picks up the phone. Sitting in your database doesn’t make someone an appropriate prospect for every brief.

A silver medallist is a candidate who reached a late stage of an earlier process and missed out on the role. Their previous assessment can frame the next conversation, but read why they missed out. A salary range the client wouldn’t meet, or a work arrangement that didn’t suit, calls for a different assessment from an unmet essential qualification.

Don’t carry a rejection forward as a permanent verdict. Equally, don’t erase it because the new search puts the person near the top. Work out which requirements have changed, which earlier observations still hold and which facts need confirming again. Keep your reasoning tied to the current brief.

Timing is its own category. The candidate may have been unavailable for the start date, unable to commute to that site or unwilling to take that particular contract. Those are good reasons to ask again when the brief changes. They’re not grounds to assume the constraint has gone away.

Australian workforce data shows why old information deserves a second look. The ABS reported that 1.0 million people changed employer or business in the year ending February 2026, a job mobility rate of 7.2 per cent. Those are national figures, not evidence about your database or the results of a reactivation campaign. The point they support is modest: working circumstances change, so confirm them. ABS job mobility

Look past exact job-title matches when the brief allows it. Related responsibilities can justify a review, provided you can point to the part of the record that supports the connection. If an essential requirement is unknown, mark it unknown. Don’t quietly let an adjacent skill stand in for it.

Then check the reasons not to make contact. The person may be in another consultant’s live process, may have asked you to stop contacting them, or may sit under a relationship restriction your agency has to honour. Verify the agreement or recorded preference that applies. A search result never overrides those boundaries.

So a good starting group is chosen by relevance, a history you can read and appropriate contact conditions. Its size matters less than whether a consultant can explain each name on it. Ask any AI recruitment agent to help prepare that explanation, then check it against the underlying record before acting.

Why does database searching become another job for recruiters?

Database searching turns into another job when the information behind a decision is spread across records, documents and conversations. The recruiter has to rebuild the history, separate current facts from old ones and work out who owns the next step. An AI score on its own resolves none of that. Prepared evidence and a clear hand-off do.

Think about what you need before calling a past candidate. The CV shows experience. A later note records a salary expectation. An email mentions they’d rather not commute across the city. And another consultant may already be talking to them. Searching CV text answers part of that, at best. A useful ATS candidate search has to show the evidence and the gaps.

Treat that as a workflow diagnosis, not an accusation that your consultants use the system badly. If an important preference lives in a conversation and not in a searchable field, someone has to go and find it. Asking consultants to retype the same information in three places just adds a second task to the first.

If you run the agency, ask where preparation time actually goes. Is the desk struggling to retrieve the right people, assess the evidence, find out who owns the record or draft a decent message? Those are four different problems. A demo that solves the first is not proof it solves the other three.

Data quality belongs in this conversation too. Australia’s APP 10 guidance requires applicable entities to take reasonable steps concerning the quality of information they use or disclose, with relevance assessed against the purpose. That makes quality a decision-time consideration, not just an import-time exercise. Applicability and appropriate steps depend on the circumstances. OAIC APP 10 guidance

On the desk, keep the date and origin of important information visible. A preference the candidate confirmed is different from one a recruiter inferred. A new CV doesn’t automatically settle every conflicting field. If the role depends on availability, ask about availability; don’t infer it from the last activity date.

This is where the case for an agent-first workflow gets specific. The opportunity is to prepare the evidence a recruiter needs, flag the gaps and propose the next action, so the recruiter assesses one coherent piece of work instead of six open tabs. The FluxHire AI recruitment overview sets out that approach, with supported capabilities confirmed during early access.

If your team already finds and reviews its database well, keep what works. Agents earn a place by improving one identified part of the process. Start with the recurring preparation task your consultants complain about, and insist on a demonstration that tackles it using authorised information.

How can AI agents prepare rediscovery work?

AI agents can help with a run of preparation tasks: interpreting a brief, organising potential matches, assembling evidence and drafting messages for review. A recruiter checks the scope, verifies the proposed fit and approves anything consequential. For FluxHire, treat these as workflow stages to evaluate during early access, with record access and available operations confirmed for your team.

Start with the brief. Split essential requirements from preferences, pin down the location and working arrangement, and work out what the client actually needs. Research help can bring in relevant context, but the consultant checks where it came from and decides whether it belongs in the search. An unsupported requirement should never quietly become an exclusion rule.

Discovery help can then organise potential matches for review. Ask for more than a ranked list of names. Each proposed inclusion should show the record, the evidence that relates to the brief and the questions still open. The recruiter needs enough context to challenge the suggestion without rebuilding the whole search.

Next comes organising the information. A summary helps a consultant read a record quickly, but it has to keep a source fact, an interpretation and an unknown apart. If the CV lists a responsibility and the summary infers a skill from it, show that step. If two records disagree, leave the conflict on the page for review.

Matching help should prepare a proposed shortlist. It shouldn’t turn similarity into proof of eligibility or interest. Ask how essential criteria are represented, how missing information is handled and how a recruiter can look at people outside the suggested group. A recommendation is only useful if you can argue with it.

Outreach help can prepare a draft that refers to the real relationship history and the relevant parts of the new brief. The recruiter checks accuracy, tone and contact conditions before anything is sent. Reject any message that invents familiarity, implies the candidate is job hunting or credits them with preferences they’ve never expressed.

Follow-up needs the same discipline. Ask how replies, opt-outs and status changes alter the proposed next action, because a draft that made sense on Monday can be wrong once the candidate has answered. When a single approach grows into a planned sequence, you’re designing recruitment drip campaigns, and the consultant still decides whether each further message serves the relationship.

Those stages are the commercial appeal of FluxHire’s approach: agents prepare work that people review and approve. They also make a practical evaluation framework. Confirm which stages can be shown on your authorised records today, rather than assuming every stage exists because it appears in a workflow diagram.

Then compare it with what your consultants do now. Do they get relevant evidence and a usable draft, or do they still search, interpret and rewrite everything? Can they see what happened between stages? Judge the workflow by the decisions it helps a recruiter make. How much animates on screen during a demo tells you nothing.

Connected glass modules surrounding organised data blocks, with blue and lavender light flowing between them.
Conceptual illustration: connected preparation stages should leave evidence a recruiter can inspect.

What does the workflow look like on a recruitment desk?

A workable desk process runs from an agreed brief to reviewed records, appropriate contact and a shortlist the candidates have confirmed. Each step needs an owner and enough evidence for the next person to follow the decision. The walkthrough below is instructional and uses placeholder fields. It doesn’t describe a customer result or a verified FluxHire feature sequence.

Agree the brief before searching

Write down the role’s essential requirements, location, working arrangement and approved remuneration information. Record how flexible the client is, separately. Clear up any ambiguity before asking software to compare people against the brief, or you’ll get a neatly organised answer to the wrong question.

Use a brief field such as [essential qualification] only where it is genuinely essential and appropriate. Keep [preferred experience] separate. For a regulated role, the consultant must establish what evidence is required and how it will be checked. An old CV isn’t a current verification, and neither is an AI summary.

Prepare a relevant group for review

Search the authorised records for evidence related to that brief, using the fields and search methods your system actually supports. If an agent proposes adjacent experience, ask it to point to the source passage and explain the link. The recruiter decides whether that connection is worth a closer look.

Look at an authorised sample of records outside the proposed group as well. It’s the quickest way to catch criteria that are too tight or fields that are rarely filled in. If the search leans on a field your team seldom completes, a short result list tells you very little about everyone who lacks it.

Review the history and assign the next action

For each relevant record, check the previous process, recent contact, owner and any restrictions. Agree who approaches the person. If two consultants are working related briefs, settle how contact and updates will be handled first. The candidate shouldn’t have to referee an internal ownership dispute through duplicate messages.

Prepare a short decision note: [evidence supporting relevance], [questions to confirm], [reason to contact or hold] and [responsible consultant]. These placeholders are a suggested structure, not four more mandatory fields for every agency. Keep only what helps someone make or understand the next decision.

Check the draft and reopen the conversation

Before contact, establish the right basis and channel for that particular message. Check the relationship reference, the role details and the ask. Strip out anything unsupported. The recruiter should be able to say why this person is getting this message, without pointing at their position in a generated ranking.

Keep the ask small. A first conversation can establish whether the opportunity is relevant and what needs updating. Where a campaign is involved, check its boundaries before you switch it on. Bullhorn’s own reactivation recipe warns that omitting its date backstop can immediately queue messages for the whole dormant database. That’s a product-specific warning with a general lesson: look at who will receive a workflow before starting it. Bullhorn reactivation documentation

Confirm the person before presenting the profile

Update the facts the candidate has confirmed, including preferences that bear on the brief. Get permission for the proposed client submission and handle the record through your agency’s approved process. If they’re not interested, record that outcome and any contact preference they state. An accurate update is worth having even when it doesn’t produce a submission.

The final shortlist should explain current relevance, not repeat a historical score. Your client needs a credible recommendation and your candidate needs to be represented accurately. Agents can prepare material for that hand-off; the consultant still owns the recommendation they put their name to.

Glass data blocks beside a clear upright review panel, connected to a larger collection by fine blue strands.
Conceptual illustration: a selected group still needs a recruiter to review the evidence and confirm the person.

What must stay with the recruiter?

The recruiter owns the reading of the brief, the assessment of evidence and the decision to contact or submit someone. Permission, preferences and exceptions also stay with people, handled through the agency’s approved process. AI assistance helps when it prepares material you can review. It becomes hard to trust the moment it hides uncertainty behind a confident recommendation.

Start with the requirements. A client preference doesn’t automatically justify excluding people, and an old rejection reason may have little to do with a different role. The consultant should be able to explain the job-related basis for the search and push back on criteria that are unclear or inappropriate.

Check the source behind any summary. Did the candidate say they’d relocate, or did the system infer it from an old address? Is a note an assessment, or someone repeating a colleague’s opinion? Keep those distinctions intact. A summary shouldn’t promote an old interpretation into a new fact.

Respect the access boundary too. Private notes and restricted records don’t become shareable because they’d help another consultant. When you evaluate a connected workflow, ask which users and systems can see each type of information. The agent’s preparation has to stay inside the permissions set for the task.

Australia’s APP 6 guidance generally limits an APP entity to using or disclosing personal information for the primary purpose of collection, unless an exception permits a secondary purpose. Consent and reasonable expectations are among the circumstances addressed by the guidance, but the correct basis depends on the situation. Don’t treat a prior application as a universal permission for every later use. OAIC APP 6 guidance

If you run the agency, the practical question is who can stop or correct the work. Ask a supplier to show you the actual review and correction steps, including what happens to an inappropriate recommendation. A general promise of human oversight doesn’t tell you where the person steps in.

In-house teams face the same question. A recruiter organises evidence, a hiring manager judges the role requirements and an authorised person approves the next stage. Settle those responsibilities before the workflow starts. An agent shouldn’t blur an approval chain that already does a useful job.

FluxHire’s public principle is that people review outputs and remain accountable for hiring decisions. Hold any rediscovery evaluation to that. Ask for the exact controls available to your team and test them in a bounded workflow. That conversation belongs next to the recruitment privacy checklist, before a campaign goes out rather than after.

How should you reopen the conversation?

Reopen it with a specific, accurate reason for getting in touch, a short explanation of the opportunity and one clear question. Check the contact conditions that apply first. An agent can prepare the draft; the consultant checks every claim and decides whether to send. Adapt the templates below to the real history and that message’s requirements.

For a previous conversation, try this structure:

Hi [name], we spoke about [verified previous role or discussion]. I'm working on [accurate description of current opportunity], and your experience in [evidence-backed area] prompted me to review our earlier conversation. Would you be open to discussing whether it is relevant now? [Sender identification, contact details and any required unsubscribe information.]

If a specific constraint has changed, name the change without assuming the candidate’s position has changed with it:

Hi [name], when we last spoke, [verified constraint] made that opportunity unsuitable. This brief offers [confirmed difference]. If it is relevant to your current plans, would you like to discuss it? [Sender identification, contact details and any required unsubscribe information.]

Both examples give a genuine reason for the approach. Neither claims the candidate is actively looking or that the role is perfect. The reply is where the consultant establishes the facts. Don’t fill a placeholder with invented familiarity just because a warmer message reads better.

Treat candidate re-engagement as a current relationship decision. The next message should reflect the answer. If the person has declined, asked for different work or stated a contact preference, build that into the next action. If there’s no answer, decide whether a follow-up is appropriate; silence isn’t continuing enthusiasm. And if they’re interested in principle but not this year, that’s a job for candidate nurture, not another chaser.

For commercial electronic messages covered by Australia’s spam rules, ACMA guidance requires consent, sender identification, contact details and an easy way to unsubscribe. It distinguishes express and inferred consent, and places responsibility on the sender to substantiate consent. Whether a particular recruitment message falls within those rules depends on its content and context. Get the appropriate advice for your campaign; these templates do not certify compliance. ACMA guidance

There’s one more trap: a message asking for marketing consent can itself be a marketing message. Don’t use an unpermitted blast to repair missing consent records. Keep source, channel and contact conditions visible when you review a proposed reactivation group.

Outreach also needs consistent context. If a candidate replies on a different channel, the next action should reflect what the recruiter has learned there. That’s why the unified recruiter inbox guide sits beside a rediscovery workflow: finding the person and keeping one coherent conversation are the same job seen from two ends.

When you evaluate FluxHire, judge the prepared draft and the review step together. Does the material reflect authorised history? Can the consultant correct it? What happens when new information changes the right next action? Those questions make a better buying decision than a generic promise of personalised outreach.

How does FluxHire's agent-first approach fit?

FluxHire is building an agent-first recruitment platform where agents prepare work and recruiters review and approve it. If the follow-up after a first approach is the part you want to organise, the page on how to re-engage candidates is the better starting point; this guide stays with the search and the first conversation.

For a rediscovery evaluation, bring a representative brief and an authorised set of records, and follow the preparation through to the point where a recruiter decides. Ask what the system can do today and what stays manual, especially in the awkward cases: contradictory details, a candidate another consultant is already speaking to, a missing fact. During early access, teams evaluate FluxHire alongside their current ATS/CRM, and supported records, integrations and review steps are confirmed with each team.

The FluxHire AI agent framework overview and the agency CRM evaluation guide give the wider picture. When you request early access, bring a specific desk problem. Asking how agents could prepare database rediscovery work for your consultants to review is a good brief to start with.

A small glass workspace with a desk and chair linked to a larger archive by a stream of blue light.
Conceptual illustration: begin with one bounded desk workflow before evaluating a wider rollout.

How do you test it on one desk and measure value?

Test rediscovery on one role family with an accountable consultant, authorised records and agreed review steps. Write down how the desk prepares today before trying an agent-assisted alternative. Measure confirmed relevance and downstream progress alongside time and errors. The pilot tells you about that workflow and sample; it isn’t a forecast for every desk in the agency.

Run your recruitment database reactivation pilot on a role your team knows well. Keep the scope small enough that the consultant can inspect every proposed record and explain each decision. Confirm tools, permissions and contact conditions before you start. A pilot doesn’t require contacting everyone the search returns, and it must not turn into an uncontrolled campaign.

Write the current process down in plain language. Who checks the brief? How are records found? Who reviews previous contact? Who writes the message? Where does the candidate’s reply get recorded? A useful baseline captures the work you’ll compare, including the time spent reviewing and correcting.

Keep candidate outcomes separate. A retrieved record isn’t a reviewed prospect. A reply isn’t confirmed suitability. A submission isn’t a placement. Report the transition you actually observed, not the biggest activity count you can find.

Define the denominator for any response measure. Meaningful replies divided by delivered, appropriate approaches measures something different from replies divided by every retrieved record. Decide what counts as meaningful before you look at the results, and record opt-outs and poor matches separately so a busier campaign can’t hide worse decisions.

Measure time the same way. Include preparation, review, corrections and the follow-up needed to reach the chosen stage. If consultants spend less time searching but more time checking irrelevant suggestions, that goes in the comparison. Theoretical machine speed isn’t saved recruiter time.

Check a sample of records that weren’t recommended. Were good candidates missed, and why? Then check the recommended ones for unsupported claims and stale information. That tells you how good the preparation is, which a count of names returned never will.

If you own the agency, connect the pilot to your economics only once you have observations. Compare attributable effort, relevant submissions and eventual fee income using your existing definitions. Account for costs and keep observed results apart from assumptions. There’s no defensible percentage improvement to promise before your team has done that work.

Agree the next decision before you start: continue with the bounded workflow, revise a criterion, fix the data or stop. If you expand, write down what you learned and what’s still untested. That gives you an adoption conversation grounded in how your team recruits, not in someone else’s benchmark.

What should you ask before adopting recruitment AI agents?

Ask whether the agent workflow searches the right records, keeps the evidence visible and gives your recruiters review and correction steps. Confirm supported operations, permissions, costs and where accepted changes go. Then test it against an authorised brief and sample. After a good demo, you should be able to explain how the work reaches a responsible person.

Start with coverage. Which records are included: applicants, prospects, previous placements or another group? Which fields can the search use? What’s excluded or restricted? Greenhouse’s documented split between applicants and prospects shows why a product label can’t answer those questions for you.

Ask to inspect the proposed fit. Can the recruiter see the underlying evidence and tell a fact from an inference? What happens to missing information? Can they look at people outside the recommendation? These are practical questions about the work, not requests for a more impressive score.

Next, follow the draft to the point of action. Who can approve contact? What happens after a reply? How are opt-outs and record changes handled? Where are accepted updates stored? Confirm each path; a connected interface doesn’t imply complete write-back across every system.

Finally, ask what the early-access evaluation includes and what it doesn’t. Agree the records, integrations, review responsibilities and measures before starting. For wider supplier selection, the candidate sourcing evaluation guide helps, while this use case stays specific.

The best reason to consider FluxHire’s AI agents is the prospect of useful work arriving prepared for your review. Test that with a real brief. Inspect the evidence, correct what needs correcting and decide whether the next conversation is better for it. If that’s the part of your desk you want to evaluate, request early access and bring the workflow with you.

Frequently asked questions

What is candidate rediscovery in recruitment?

Candidate rediscovery is the review of people already recorded in an ATS or recruitment CRM against a new brief. It uses previous applications, experience and relationship history as evidence to inspect. The recruiter still needs to confirm current interest, relevant facts and permission before taking the person forward. Rediscovery identifies a reason for a conversation; it does not establish current suitability by itself.

How is rediscovery different from database reactivation?

Rediscovery identifies records worth reviewing for a current requirement. Recruitment database reactivation is the subsequent process of reconnecting with people where contact is appropriate. A record can be relevant without being suitable for immediate outreach. Keep the search decision separate from the contact decision, and review the relationship, current processes, preferences and applicable contact conditions before sending a message.

Can AI agents confirm that an old candidate is suitable?

An agent can help organise recorded evidence and prepare questions for a recruiter to check. An old record alone cannot confirm present interest, availability or preferences. Treat missing or conflicting information as something to resolve with the candidate, rather than a gap to fill with an assumption. The recruiter remains responsible for the recommendation and for approving consequential actions.

Are silver medallist candidates always the best starting point?

Previous finalists are a useful group to review when their experience relates to the new brief. They are not automatically the strongest candidates for a different role. Read why the earlier process stopped, identify which requirements have changed and confirm the facts that matter now. Use the previous assessment to inform a conversation, rather than carrying its outcome into every later search.

Can we email everyone in our recruitment database?

Do not assume that a stored address permits every later message. Review the purpose, relationship, preferences and applicable requirements for the particular campaign. For commercial electronic messages covered by Australia's spam rules, ACMA sets out consent, sender identification and unsubscribe requirements. Privacy obligations also depend on the entity and proposed use. Use the linked ACMA and OAIC guidance and obtain appropriate advice for your circumstances before activating a campaign.

Do all ATS candidate searches include the same records?

Check the documentation and demonstrate the search on an authorised sample. Record coverage, fields, filters and permissions differ between tools and views. For example, the Greenhouse Past candidates documentation describes previous applicants and excludes prospects without a past application. Ask what your chosen workflow includes, which records it omits and how missing information affects the results.

Does FluxHire already provide every rediscovery workflow described here?

Availability must be confirmed with your team. FluxHire is building an agent-first recruitment platform where agents prepare work and recruiters review and approve it. The rediscovery applications in this article are evaluation criteria, not a feature guarantee. During early access, agree the supported records, integrations, operations and review steps for the workflow you want to test alongside your current tools.

How should an agency measure a rediscovery pilot?

Record the current workflow before the pilot and choose a bounded comparison. Track reviewed relevance, appropriate delivered approaches, meaningful replies, candidate-confirmed suitability, submissions and later outcomes separately. Include review and correction effort when measuring time. Define each denominator and avoid attributing a placement or fee to the pilot without evidence. The result describes the tested workflow and sample, rather than a guaranteed return for every desk.

Sources and reading

Sources checked 1 October 2026. The workflow and message examples are instructional. Privacy and messaging guidance is general information; confirm the requirements for your organisation and proposed use.

Evaluate the AI agent framework

Review how FluxHire handles records, ownership and hand-offs

Read the AI agent framework overview, then discuss your records, consultant hand-offs and existing tools with our team. Recruiters review AI outputs and approve consequential actions. Early-access availability is subject to review.

Continue Reading

View all articles