Greenhouse Turns Candidate Profiles Into a Sourcing Channel
On this page 7 sections
AI-generated editorial illustration.
On October 15, Greenhouse put a new name around its hiring software. The Orchestrated Hiring Ecosystem joins an applicant tracking system with sourcing, interviews, scheduling and the profiles that millions of job seekers have already created in MyGreenhouse. The most consequential part is still limited: Greenhouse says sourcing from that pool is in closed beta for select customers. Its candidate portal, Dream Job flag, identity checks and Voice AI interview are existing products or earlier launches, assembled under a larger pitch.
A recruiter opening a role this month can already receive applications through an ATS. The new promise is that the same system may surface people who are looking before they apply. A person who built a profile to reuse a resume or track an application may now enter a discovery channel. Visibility, contact, interest and a completed hire are four different events. Greenhouse has announced the route; it has not published a controlled result for the beta’s completed hires.
CEO Daniel Chait’s October account describes a cycle in which candidates send more AI-assisted applications and recruiters screen more of them with AI. Greenhouse wants to replace volume with what it calls signal: intent, verified identity, structured interviews and the record of an actual hiring process. The company is also moving into a crowded contest for the place where candidates and employers meet. LinkedIn approaches it through its professional network. Paychex approaches it through payroll and HCM customers. Greenhouse starts from the ATS and asks the applicant to stay.
For a buyer, the bundle’s reach matters only when an introduction goes somewhere. Can the sourcing channel find people who want to be found, answer a recruiter, complete a fair evaluation and accept a job? A useful trial has to count the applicants who feed the system and hear nothing back. An employer can run that test. A candidate should be able to see its effects.
An applicant record becomes a discovery route
Greenhouse says its software serves more than 7,500 companies. That is an employer-customer count, not a count of companies enrolled in the new sourcing beta. The release also says millions of people have profiles in MyGreenhouse. It does not say how many are actively seeking work, have agreed to be considered for new roles, match an open job, or can be contacted by a beta customer. Those denominators decide the practical size of a talent network.
MyGreenhouse was not invented on October 15. The candidate service lets a person save information for applications, find jobs, receive alerts and track status where employers share it. Greenhouse’s employer setup documentation describes one login across participating companies. It says account creation is optional for applicants and that the employer chooses whether to enable features such as status sharing, tasks and job-board controls. The portal already helps both sides: a candidate can avoid retyping information, while a recruiter receives a more complete application.
The sourcing beta changes the direction of travel. An application begins when a person approaches a particular job. A sourcing suggestion begins when an employer looks for people who may suit a job and have signaled availability. The release says Greenhouse will surface actively looking job seekers who fit a role inside the recruiter workflow. Its public description does not supply the beta’s matching criteria, profile-visibility settings, consent wording, eligible geographies or contact-rate results. A buyer should ask for those before treating the millions of accounts as a searchable inventory. A candidate should be told when a profile field is used to suggest them for an unrequested role.
The distinction between a person and a profile is especially sharp for a recruiter who hires infrequently. If a role opens once a year, the recruiter may not have a maintained talent pipeline or time to learn another sourcing product. A suggestion inside the existing ATS could help because the role, interview plan and decision owners are already there. The shortcut still depends on a current profile and a recruiter ready to act. Send a generic message to every suggested person and the application flood simply runs in reverse. A trial should record how many relevant contacts the team can make and answer.
One existing feature shows why a two-sided account can be valuable. A person can mark one application each month as a Dream Job. That creates a scarce expression of interest where ordinary applications can be numerous. The signal is more informative than clicking Apply on fifty roles, but it still needs interpretation. A candidate may choose a Dream Job because of location, salary, mission or a personal referral. A recruiter cannot infer skill from intensity of interest. Nor should a candidate who does not use the portal be presumed less interested.
Greenhouse’s home page claims a person is five times more likely to be hired with Dream Job. That is a vendor-reported comparison whose public display does not give the cohort definition, controls or base rates needed to identify a causal effect. Motivated applicants may be more likely to use a scarce flag. Employers may also review those applications earlier. Both can raise the observed hire rate without the button itself improving a person’s qualifications. A procurement team could ask Greenhouse for a matched-role analysis, not multiply its own expected hires by five.
The commercial idea is clear even with those unknowns. An ATS already holds job definitions, interview plans and candidate decisions. If a candidate account supplies fresh preferences and interest, Greenhouse can help a recruiter search before a posting creates an overloaded inbox. That can save staff time if the matches respond and fit. If it only moves the pile from applicants to suggested profiles, the recruiter has bought another queue.
The candidate account has limits
A reusable profile sounds like one place to manage a search. The MyGreenhouse candidate FAQ draws a tighter boundary. An employer controls whether application status appears in the portal; MyGreenhouse cannot change an application or contact a recruiter for the candidate. Applications submitted through LinkedIn, Indeed or other third-party routes may not appear. The account currently supports one default resume, rather than multiple tailored versions. These are ordinary product limits, but each matters when the marketing language suggests a complete view of a job search.
There is also a data boundary. Greenhouse says it controls information a person adds to a MyGreenhouse account and lets the person edit or delete the profile. Once someone submits an application, the employer becomes the controller of that application data; a request to correct or delete it goes to the employer. An identity check can add yet another party. Greenhouse’s CLEAR verification FAQ separates CLEAR’s control of identity artifacts, Greenhouse’s control of portal and result data, and the employer’s control of the application-linked result. Deleting one account is not necessarily deletion of every record touched in a hiring process.
For a candidate asked to complete a profile, verify an identity and sit for an AI interview, these boundaries become immediate. Which steps does this job require? Who can see a new preference after an unsuccessful application? How long will a transcript remain available, and who can correct it? The launch announcement cannot answer those questions for each employer’s configuration. Employers need to publish the process they actually use.
The portal’s status feature makes the trade concrete. Greenhouse says an employer may choose what updates to share and that the candidate-facing stage need not expose internal stage names. The stage documentation describes mapping internal steps to standard external ones. That can make a complicated hiring process understandable. It can also leave a person staring at a broad stage while the employer has already made a decision. An account is useful only if the employer keeps the signal current.
For a worker applying through an employer’s job board, the immediate benefit may be simple: no repeated form, an acknowledgment that the application arrived, and a timely answer. For Greenhouse, the same account can produce preferences, response history and role interest that improve sourcing. Both sides can benefit. The distribution of benefit depends on product settings and employer behavior, not on the existence of a common login.
Employers have a defensible reason to control what appears in that login. Internal interview feedback, references and deliberations are not candidate-facing status updates. Publishing every internal stage could also expose a tentative judgment as a final decision. But a broad external status can become an excuse to send no human message. The practical compromise is to set a response deadline for each stage and identify the employer contact who can resolve an error. A portal can deliver that promise consistently; it cannot write the promise for the employer.
Candidates do not all arrive through the same door. Some may start with a direct job-board form, others with an external network or an agency, and others with an employee referral. Greenhouse’s FAQ says third-party applications may be absent from MyGreenhouse. A hiring team that wants a complete picture should check duplicate records and source attribution before it treats a blank portal view as evidence that someone never applied. For the applicant, the cost of a fragmented view is personal: they may repeat a screening step or contact a recruiter who assumes the system already answered them. Fixing those handoffs may matter more than adding another profile field.
A signal is only useful after a response
Greenhouse’s framing has a factual basis in its own research, though that research should be read as vendor-sponsored survey evidence. Its 2026 AI in Hiring report draws on 1,200 U.S. job seekers, 219 recruiters and 446 hiring managers. In Chait’s summary, 49% of job seekers said they applied to more jobs than in the previous year, 46% said trust in hiring had fallen, and 34% of hiring managers reported spending up to half their week filtering spam or junk applications. These are survey responses about experience, not platform logs that establish how many Greenhouse applications were false or how the new sourcing beta performed.
The company’s October release reports that fewer than 7% of applicants ever get an interview. That figure lacks a publicly comparable role, time and applicant-cohort definition in the release. It is not a conversion rate a buyer should transfer into their own hiring forecast. Application volume differs radically between a retail shift, a specialist engineering role and a remote junior position. A recruiter also sees multiple applications from one person and resumes that may have been polished by software without being fraudulent.
A richer conversation may reveal what a resume misses. It can also create a new false negative. A Dream Job flag rewards people who know the feature exists and choose to spend it on a particular role. A sourcing filter may miss someone whose profile is stale. Identity verification can deter fraud while placing a document or device burden on a legitimate applicant. A voice interview can reveal job-relevant detail; it can also be difficult for a person with a disability, a poor connection or no private place to speak. The employer needs a route for those cases.
Greenhouse says human decision-makers remain accountable. That is a meaningful design principle, but it does not describe the whole journey. A person can be excluded before a human receives a profile. A recruiter can read an AI evaluation without having time to inspect the transcript. A hiring manager can leave a candidate waiting even when the platform sends a prompt. The evaluation unit should therefore be a full application or sourced-person cohort, from exposure to answer, rather than a count of AI actions.
For the candidate, the simplest signal of a functioning system is a reply. A sourced contact should say why the employer believes the role fits, who will read the response, and when the candidate can expect a next step. A rejection should close the loop instead of leaving a profile eternally “active” in a dashboard. For the employer, reply rates and completed hires are harder metrics than impressions or matches. They also reveal whether the system is surfacing people who actually want this particular work.
AI-generated editorial illustration. A profile must still travel through contact, evaluation and an employer decision.
Voice interviews expand the first gate
Greenhouse’s Voice AI adds an interview before the recruiter has committed the same time to every applicant. The company bought Ezra AI Labs in May and says the tool asks the same role-specific questions and uses recruiter-defined criteria. The October release says applicants can have an initial conversation and that hiring teams receive transcripts and explainable evaluations. Greenhouse also says Voice AI can be purchased without its ATS, a useful limit on the claim that every product must be bought as one system.
The word “every” needs a real deployment definition. Does every applicant for a participating role receive an invitation? Does the candidate need to finish the conversation for the application to count as complete? Who offers an alternative? A Greenhouse Zapier customer story describes a 25-minute Voice AI conversation after the application, with 97% of participating candidates rating the experience excellent. That is one vendor-selected customer account, not a cross-employer estimate of accessibility, completion, job performance or comparative hire quality. An applicant who declined or dropped out may not be represented in the satisfaction percentage.
Earlier Greenhouse candidate research shows why disclosure cannot be left to a product page. In a survey of 2,950 active job seekers across five countries, 63% said they had encountered an AI interview and 38% said they had walked away from a process because it included one. Among U.S. candidates with relevant experience, 70% said AI involvement was not clearly disclosed before the most recent interview. The study describes reported experiences across employers, not Greenhouse Voice AI outcomes. It nonetheless shows the risk of turning an optional conversation into an unexplained condition of applying.
A buyer should demand an employer-specific comparison. For similar open roles, how many people were invited, consented, completed, asked for a human alternative, advanced, received a rejection, accepted an offer and remained in the role? How much recruiter time did the tool save after reviewing transcripts, handling exceptions and answering candidates? Did the process widen access for people whose resumes undersold their ability, or make another lengthy requirement before a human would speak? The answers may differ by job family and applicant volume.
Voice AI has a credible use case. A recruiter with a queue of applications may be able to hear from people who would otherwise never reach a first call, and a candidate who works shifts may prefer an interview outside office hours. Structured questions can make comparisons less dependent on which recruiter happened to be free. Those advantages justify a trial. They do not justify making a positive interview rating stand in for an offer, or assuming that the same completion rate applies to every language, disability and connection quality. The employer must decide what accommodations and manual review are available before the queue arrives.
Price is part of that test. Greenhouse’s Voice AI product page says it sells annual subscriptions based on company or scoped-team headcount, with a fair-use interview allowance and often a contract of at least twelve months. The site does not list a public dollar price. The core ATS pricing page also asks employers for a custom quote based on plan, volume and complexity. MyGreenhouse’s employer documentation says portal features are included at no added cost across customer tiers. That does not establish what sourcing will cost after beta. A buyer needs a quote that separates software subscriptions, implementation, identity checks, staff review and any per-interview allowance.
LinkedIn and Paychex take different routes
Greenhouse’s launch is part of a competition over where hiring decisions start. On September 29, LinkedIn announced Hiring Assistant 2. Its agent draws on professional profiles, verification, ATS information and prescreen responses. The new capabilities are scheduled to reach existing English-language Hiring Assistant customers beginning in November. LinkedIn says more than 20,000 companies use its agentic hiring solutions; it does not disclose how many use this version because it has not reached that scheduled rollout. The product’s first-party asset is a large member network, with ATS connections supplying part of the workflow.
LinkedIn’s release reports that its assistant lets recruiters review 83% fewer profiles to find qualified matches and find interview-quality applicants 33% faster. It also says top matches are 2.6 times as likely to be interviewed. Those are LinkedIn-reported metrics with their own selected users and definitions. They cannot be laid beside Greenhouse’s five-times Dream Job claim to declare a winner. One describes recruiter search, another the observed outcome of a candidate flag. Both stop before long-term performance or a counterfactual hire test.
Paychex introduced WISE Hire on September 23 from a different base. It offers sourcing, workflow automation, matching support and hiring expertise both as a standalone product and integrated with Paychex HCM. The company says customers find and hire qualified talent four times faster on average. The release does not publish an independent comparison cohort, rate card or complete method for that average. Paychex’s roughly 840,000 customers and payroll relationship give it distribution among employers that may never start their software search with an ATS. They do not prove WISE Hire has already won those customers.
The routes create different tradeoffs. LinkedIn can discover a member long before that person applies to a particular employer. Paychex can join hiring to payroll and employer operations. Greenhouse can join discovery to a structured interview plan and the record of the hiring decision. A recruiting team may use more than one. Indeed, outside agencies, employee referrals and direct applications remain part of many searches. The count worth testing is the number of qualified people each route introduces who would otherwise have been missed. Their replies matter too.
Greenhouse explicitly says its ecosystem remains open to outside tools. Its May MCP announcement describes controlled connections to third-party AI tools. Voice AI is offered outside its ATS. Those are practical forms of openness. A buyer still needs to test exports, duplicate-profile resolution, permission boundaries and whether a candidate discovered on one service can be managed without forcing a second account. An integration announcement does not answer those workflow questions on its own.
A buyer must count both sides
The October beta offers a chance to measure a candidate network at a manageable scale. A buyer could run a trial on several comparable roles with enough applicants and clear job criteria, holding the recruiter team and decision rules as steady as possible. The trial should separate candidates sourced through MyGreenhouse from direct applicants, referrals and other channels. If a person appears in several channels, attribute the initial introduction and record the overlap instead of counting two hires.
Here is a two-sided scorecard for such a test. The rows are an editorial design, not Greenhouse’s published telemetry or a promise about fields available in its product.
| Stage | Employer denominator | Candidate check | Decision the measure supports |
|---|---|---|---|
| Discoverable pool | People eligible for this role and visible under the trial settings | Was visibility understood and controllable? | Whether the searchable pool is usable rather than merely large |
| Suggested match | Unique people surfaced to recruiters | Was the reason for a suggestion intelligible? | Whether matching reduces review work |
| Contact | Unique people sent a relevant invitation | Did the invitation name the role and next step? | Whether suggestions reach people likely to answer |
| Conversation | People who accepted and completed a first step | Was an alternative available when needed? | Whether automation widens or narrows access |
| Decision | Interviewed, offered and hired people by source | Was there a timely answer after each stage? | Whether the channel changes actual outcomes |
| Follow-through | Hires retained at a defined date | Could a candidate correct or withdraw data? | Whether a fast hire stayed a good hire |
The table has one cost column off the page: staff time. Recruiting leaders should log hours spent reviewing suggested people, editing role criteria, checking identity exceptions, reading transcripts, explaining decisions and answering appeals. A lower cost per invitation may still produce a higher cost per accepted offer. Similarly, an account can reduce a candidate’s form-filling while increasing time spent on screening conversations. Both ledgers belong in a purchase decision.
Do not divide hires by every registered MyGreenhouse profile. Most profiles will never be eligible or active for a particular role. Use the population that the product could actually show under the beta’s rules. Then report contact, reply, interview, offer and hire rates by role, location and source. A matched comparison to the employer’s existing channels will be imperfect because people select where they apply; it is still far stronger than citing a platform’s global account count.
Even a well-run test will have a selection problem. People who save a full profile may be more motivated, digitally comfortable or already familiar with Greenhouse-hosted employers than those who do not. Recruiters may reserve the new channel for hard roles or use it first on easy roles. Either choice shifts the apparent success rate. Record the role mix and the order in which recruiters used each channel. If possible, compare similar requisitions before and after the beta, then disclose the differences that remain. A dramatic percentage without a baseline can hide a small number of hires or a change in the team’s behavior.
The buyer also needs a harm and access review. The EEOC’s guidance on selection procedures treats interviews, tests and other procedures used to make employment decisions as selection methods. It does not mean this particular beta is unlawful. It means an employer should examine how its own criteria, interview rules and accommodations work when software changes who receives attention. A human final click cannot repair a candidate never shown to that human.
An honest trial may find that direct applicants and referrals still outperform the new pool for some roles. It may also find that a profile with fresh availability saves weeks of search for a hard-to-fill job. Either result is useful. The point is to make the units visible to the employer and the person, rather than allowing a single “signal” score to explain every outcome after the fact.
When the first sourcing trial closes
Greenhouse has a sensible reason to connect its two customer sides. People who already use MyGreenhouse to apply and track jobs may be easier to reach than a cold list. Employers using the ATS already have roles, interview plans and decisions that could help evaluate those introductions. A single system can reduce the handoffs in which a candidate disappears.
The October release also leaves material questions open. Sourcing is a closed beta. Public materials do not state its future price, the fraction of profiles that are eligible for discovery, the consent experience for this use, contact-to-hire rates, or a controlled comparison with existing channels. Greenhouse should be judged as a proposed route with some live components, not as a proven cure for application volume. LinkedIn and Paychex have their own vendor-selected speed claims, rather than a shared independent hiring benchmark.
Start with named roles, a visible consent rule and a fixed trial period. Record who was contacted, answered, interviewed, hired and told the result. A candidate should know whether a profile is being used to suggest them; a recruiter should know whether a suggested person wants the role. When the cohort closes, count the jobs and the replies. The database size can stay in the sales deck.