Findem

Findem works alongside the customer's ATS and writes back into it. An Integrations page sits in the primary navigation under the heading "built to work with your existing systems". Findem ingests from customer ATS systems, including EEOC data where provided, writes verification results back into the ATS with context, and delivers hire-ready candidates into the team's existing workflow.

Outward reach runs in two further directions: real time talent intelligence exposed into agentic workflows on the ServiceNow AI Platform, and an Embedded AI line placing Findem's sourcing copilot, job board and labeling engine inside partner products. Multichannel sourcing and outreach span external channels beyond the ATS. The breadth rests on the platform's positioning and the ATS write back rather than on an enumerated connector list.

Sourcefindem.ai navigation, /agents/id-verify-agent and /embedded-airead 2026-09-08

Agents "plan, execute, and improve full hiring workflows, from calibration to hire-ready candidates", in Findem's words, and the execution involves several agents rather than one looping. A live job post drives Intelligent Job Post, the Application Boost Agent runs outreach, the Screening Agent conducts structured screens, the Scheduling Agent coordinates availability and books interviews, and the ID Verify Agent runs identity checks at the start, before candidates are screened, assessed or scheduled. Work passes between named agents with different jobs at ordered stages.

Findem's own tiering makes the same point: Assistive AI streamlines, Agentic AI lets workflows run themselves, and Build AI lets teams construct agents reflecting their own hiring philosophy and standards. No visual workflow builder, branching model or named orchestration runtime is documented, so the orchestration shows up as staged execution across several named agents rather than an exposed engine.

Sourcefindem.ai/meet-the-agents and findem.airead 2026-09-08

Findem's retrieval structure is the product, and each layer has its own page. 3D Data is a time ordered graph across people, companies and time, built on the premise that every person and company has a trajectory. Success Signals are expert labeled indicators of potential, performance and fit derived from unstructured career history. Relationship Signals map who has worked with whom so teams can find warm introductions, and a Data Labeling Engine structures raw data into AI ready signals.

The structure persists, scales far past a context window and stays queryable: it answers questions no résumé database can, such as who was part of a successful exit or who took a company from negative to positive operating margin, and it surfaces explainable signals rather than opaque scores.

Much of the graph is Findem's own corpus built from public professional sources, which is vendor knowledge rather than the customer's. The customer's own knowledge enters it too: the customer's ATS data is ingested into the same structure, candidate rediscovery runs against the customer's existing database, and the Labeling Engine is offered to structure the customer's data.

Sourcefindem.ai/why-findem/3d-data, /success-signals, /relationship-signals and /platform/data-labeling-engineread 2026-09-08

A defined handoff marks where agent output stops and a person decides. Intelligent Job Post delivers hire-ready candidates automatically to the team for review, and Findem's framing is that agents move qualified candidates forward while teams stay focused on judgment and fit. Explainable signals rather than opaque scores let a recruiter see why a candidate surfaced before acting on it, and a Responsible AI page is published as a first class part of the site.

No approval queue, confidence threshold, guardrail configuration or documented hold before an agent acts is published. That matters more here than on a sourcing only product, because these agents act on real people: the Screening Agent conducts structured screens with candidates, the Scheduling Agent books and reschedules interviews, and the ID Verify Agent captures government ID and biometric signals. Delivery for review after the fact is not a review and approve surface over the action.

Sourcefindem.ai, /why-findem/responsible-ai and /meet-the-agentsread 2026-09-08

Security rests on an active SOC 2 Type II, with an annual independent audit and the audit firm named publicly, plus continuous monitoring, role based controls and a published vulnerability disclosure route. A Vanta operated trust center at trust.findem.ai names Information Security as one of four pillars alongside AI Transparency, Employment Fairness and Data Privacy, and a responsible disclosure route to trust@findem.ai is published with a reproduction steps protocol.

The live security page in the site navigation is findem.ai/why-findem/security-and-compliance, and the Data Processing Agreement page redirects into the trust center. That trust center is client rendered, so its certifications, artifacts and access control detail sit behind a rendering layer rather than on the public pages.

Sourcetrust.findem.ai and findem.ai/report-security-issueread 2026-09-08

Reasoning is inspectable at the decision layer: Findem surfaces explainable signals rather than opaque scores, so a recruiter can see why a candidate matched. A named Analytics product covers recruiting performance, diversity and market trends, and the ID Verify Agent records verification results and shares them with context into the ATS, a durable artifact of what one agent did.

What is missing is an agent run trace, reasoning log or audit trail of what the system did on a recruiter's behalf. Explaining why a candidate ranked is not reconstructing what the Screening Agent asked, what the Scheduling Agent changed, or what the Application Boost Agent sent to whom, and on a platform whose agents contact candidates directly that is the gap a buyer would care about. Reporting is not auditing.

Sourcefindem.ai/products/talent-analytics-sourcing and /agents/id-verify-agentread 2026-09-08

Agents carry no published memory: nothing with a stated scope or lifetime, and no per agent state retained across runs. The time ordered talent graph that agents read on every run is a shared data layer for grounding rather than an agent memory store.

Candidate rediscovery reads the persistent graph rather than a memory of prior agent runs, and the customer's ATS records the platform ingests are the application's data model: deleting them would delete the business record, which is what separates a data model from a memory layer.

Sourcefindem.ai/why-findem/3d-data and /meet-the-agentsread 2026-09-08

Where customer data sits is not a buyer choice. No deployment model choice, region list, hosting location, VPC, on premises option or data residency commitment is published. The vendor's architecture messaging speaks of isolating each customer environment rather than letting a buyer choose where data sits, and isolation is a tenancy property, not a residency option.

The trust center at trust.findem.ai holds the Data Privacy and Information Security pillars, and the Data Processing Agreement page redirects into it, but the portal is client rendered, so any residency terms inside it stay out of public view.

Sourcetrust.findem.ai and findem.ai navigationread 2026-09-08

Findem publishes a catalog of prebuilt agents at findem.ai/meet-the-agents, linked from the primary navigation under Agentic AI and from the homepage as "See all agents". Five named agents each have their own page: the Application Boost Agent, engaging the right people with outreach that reflects their experience; the Screening Agent, conducting structured screens to verify strengths that matter in practice; the Scheduling Agent, coordinating availability, booking interviews and managing changes; the ID Verify Agent, checking government ID documents with facial matching and behavioral signals to detect proxy participation; and Intelligent Job Post, turning static posts into hire-ready candidates delivered automatically for review.

Each does a whole job at a distinct stage of one pipeline: remove the Scheduling Agent and the Screening Agent still screens, remove ID Verify and outreach still runs. That makes it a pack rather than tiers of one product. A Build AI tier lets teams create agents reflecting their own hiring philosophy and standards. The ID Verify Agent page invites teams to be among the first to use it, so that agent is in restricted availability.

Sourcefindem.ai/meet-the-agents and the five agent pagesread 2026-09-08

One real event trigger exists: Intelligent Job Post turns a live job post into hire-ready candidates delivered automatically to the team, described as moving from post and pray to progress on autopilot, so the posting itself starts the agent without a recruiter initiating a search. The ID Verify Agent runs at a defined point in the pipeline, at the start before candidates are screened, assessed or scheduled, and the Scheduling Agent reacts to availability changes and reschedules. Multichannel sourcing and outreach and candidate rediscovery against the existing database add reach.

There is no event framework a buyer can address: no webhooks, subscriptions, schedule engine or enumerated channel matrix. Delivery is into the ATS and the Findem workspace rather than across a documented set of channels, and most workflows outside Intelligent Job Post remain recruiter initiated from a search.

Sourcefindem.ai/platform/intelligent-job-post and /meet-the-agentsread 2026-09-08

Customers get no model choice, and no model provider is named. No model picker, admin model setting, routing description, provider list or bring your own model option is published, and naming no provider at all is unusual. Findem's framing runs the other way: it describes a unified, explainable model built on its own expert labeled data and states that one customer's data is never repurposed to train another's. That is a data handling boundary and a proprietary model claim, not customer choice.

AI Transparency is one of the trust center's four pillars, though any provider disclosure there is out of public view because the portal is client rendered.

Sourcefindem.ai/why-findem/responsible-ai, findem.ai/blog/ai-security-in-hr and trust.findem.airead 2026-09-08

An Embedded AI product line, positioned as "your expertise on Findem's infrastructure", carries four named products in the primary navigation: AI Job Board, Sourcing Copilot ("AI actions without leaving your platform"), Labeling Engine, and Agents and MCP ("ship agents with speed and confidence"). A Partnerships program sits beside it as the commercial access path, and Findem has brought real time talent intelligence into agentic workflows on the ServiceNow AI Platform, an outside caller driving this platform.

The API and MCP surface is asserted with no documentation behind it. No developer documentation site, published API reference, endpoint list, authentication guide or SDK is published, so MCP appears as a navigation label rather than a documented interface.

Sourcefindem.ai navigation and /partnershipsread 2026-09-08

Nothing on testing is published: no test environment, simulation harness or scored evaluation of agent behavior.

The notable gap is specific to this category. No independent bias audit summary, adverse impact testing, NYC Local Law 144 audit disclosure or EU AI Act risk classification appears on Findem's own Responsible AI page, in a product category where candidate ranking and automated screening may constitute an automated employment decision tool.

The gap is sharper because the agent set includes a Screening Agent conducting structured screens and an ID Verify Agent applying biometric matching, both squarely in scope for employment decision scrutiny. The trust center names Employment Fairness as one of its four pillars; whether a bias audit sits behind that client rendered portal is not visible from outside it.

Sourcefindem.ai/why-findem/responsible-ai and trust.findem.airead 2026-09-08

Browser and computer use do not appear in the product. The closest feature is the ID Verify Agent, which has candidates upload or capture government ID documents live, checks document authenticity, and uses facial matching and behavioral signals to confirm the person engaging is the same person and to detect proxy participation. That is biometric identity verification through a capture flow the vendor controls, not an agent operating software it does not control, and no browser, desktop session, or remote or local computer control is present.

Elsewhere the platform aggregates and analyzes talent data and acts through ATS and channel integrations, programmatic interfaces rather than screen control. Questions of breakage when UI elements change, or whether control is native, sandboxed, local or remote VM, do not apply to facial matching or an ATS write.

Sourcefindem.ai/agents/id-verify-agent and findem.airead 2026-09-08