Frequently asked questions

Talent Agent is an AI-native recruiting platform. Candidates build a structured profile from their resume, LinkedIn, and other sources. Recruiters define how they want candidates evaluated, and AI agents do the screening - surfacing ranked matches with evidence-backed fit explanations on both sides.

Job boards rely on candidates self-selecting and recruiters skimming resumes for keywords. We invert that: every candidate gets a structured profile, every role gets a clear set of scoring steps, and the platform produces a ranked shortlist with reasoning. Less noise on both sides.

Three audiences. Candidates who want to be matched on substance rather than keyword overlap. Recruiters who want a ranked shortlist before they review a single profile. Admins who want to define exactly how their organization evaluates fit - and have the platform apply that consistently.

No. Recruiters stay in control of every pipeline decision - advance, pass, message. The platform handles the screening groundwork so recruiters can focus their time on the conversations that actually matter.

AI agents handle three layers: extracting structured profiles from resumes and other sources, evaluating candidates against role-specific criteria, and combining dimension scores into a final ranking. Every agent's output is traceable - recruiters and candidates can see the reasoning behind each score.

Create a job draft (basics, requirements, compensation, screening criteria), then attach an evaluation workflow your admin has built. You can tailor the workflow per-job - adjusting weights, thresholds, filter values, or swapping evaluators - without modifying the underlying template. Publish and matches start flowing in.

A workflow is a configured pipeline that defines how candidates are assessed for a role. It pulls hard rules (filters), scoring dimensions (evaluator agents), and combination logic into a single attachable unit. Admins build workflows once; recruiters apply and tailor them per-job.

Yes. Tailoring is always job-scoped. You can describe what's different about the role and let the platform suggest overrides, apply a saved playbook from a similar past role, manually adjust weights and thresholds, override filter values, or swap an evaluator entirely. The underlying template stays untouched.

Two possibilities. They were rejected by a hard rule - check the Filter Rejections section on the job for the exact rule and value. Or they scored below the workflow's match threshold - check the full match detail (visible regardless of score) for the dimension breakdown.

Yes. Rankings are recommendations, not gates. The full pool is visible from the matches view - filter or sort to see lower-ranked candidates. Filter Rejections (hard-rule failures) are also surfaced, just in a separate section so they don't crowd your active list.

The moment you click Interview on a candidate, a direct messaging thread opens for that role with an opening system note - so you can start a conversation right away. Their name and contact info stay hidden until both sides flag interest. Threads are scoped per (candidate, role); interviewing the same candidate on a second job opens a separate conversation. Candidates reply from their portal in real time.

There's a bell icon in your sidebar. It lights up when a candidate flags interest in one of your roles, when a match reaches mutual interest, or when a candidate replies in a thread. Click for the recent list or open the full notifications page. Notifications are in-app only today.

New accounts must verify their email address before signing in. Look for a message from noreply@usetalentagent.com (check spam if needed) and click the link. Links expire after 24 hours; you can request a fresh one from the confirmation page if needed.

The Intelligence Hub. It has tabs for Workflows (the visual pipeline editor), Criteria (evaluator templates, prompts, and model configs), Filter rules (reusable hard rules recruiters can add per job), Questions (reusable screening questions), Runs (every execution with full traces and per-agent observability), and Activity (a live feed of the agents your workflows have invoked). A workflow's own built-in filters are configured inline on each filter node; the Filter rules tab is the reusable layer applied per job.

Fully. You build evaluator agents (prompt + model + output contract), filter templates (with optional tailorable conditions for recruiters to override), and workflows that compose them. Different role types or seniority levels can have different workflows. Recruiters then attach the right workflow per job.

Both agent templates and workflows are versioned. Edits create a new draft; publishing makes it the active version. Old versions remain inspectable, and historical runs always show the version that produced them - even if the template has been edited since.

Yes. The Agent Observability panel on the Runs tab shows per-agent latency p50/p95, token usage, and error rate over a configurable time window (1 hour through 30 days). Useful for spotting prompt regressions or expensive models after a change.

You can start by uploading a resume - the platform will extract a structured profile from it. But the profile goes well beyond a document: it captures your experience, skills, goals, and preferences in a format the matching system can actually reason about. You can also connect LinkedIn, GitHub, and answer follow-up questions to fill in gaps.

When a recruiter publishes a role, the platform runs your profile through the must-have rules (work authorization, location, etc.), then scores you across several areas (skills match, experience depth, trajectory fit), and combines those into an overall fit. Matches are surfaced with an explanation describing what aligned.

Yes. Every match includes a detailed fit explanation broken down by dimension. You'll see what aspects of your profile aligned with the role and where there were gaps - useful both for understanding the match and for figuring out where to strengthen your profile.

We show qualitative fit labels - Strong, Good, Moderate, Weak, or Poor - instead of a raw percentage. The underlying signals aren't precise to two decimal places, and a single number tends to overshadow the per-dimension breakdown that actually tells you something useful. Recruiters see more detail; your view focuses on whether it's worth pursuing.

Each application has its own timeline page that interleaves status updates, recruiter messages, AI follow-up questions, and agent conversations in one chronological feed. Open any application from your Applications list to see the full story for that role.

Only when both you and the recruiter have flagged interest. Up until that point, they see your structured profile and match score, but not your name, email, phone, or any direct contact info. You'll get an in-app notification the moment mutual interest is reached and contact information is shared.

There's a bell icon in the sidebar that lights up for every important event - a recruiter expressing interest, mutual interest reached, a new message, or a conversation that needs your attention. Click the bell for the 10 most recent items or open the full notifications page. Notifications are in-app only today; we don't send email.

Anytime. Completing onboarding or making a major profile change automatically triggers a rerun of your existing matches so the scores reflect the updated profile. Smaller edits flow into future matches automatically and existing scores update on the next workflow run.

GitHub analysis is held in a separate, isolated module. It powers a GitHub-specific assessment for recruiters who opt to use it, but it doesn't silently boost your standard skills list or merge into your main profile. You can connect or disconnect GitHub at any time.

It confirms the address is real and reachable before any data is associated with it - important because matches, messages, and password resets all flow through email. New accounts can't sign in until verification is complete.

Check your spam folder for a message from noreply@usetalentagent.com. If it's still not there, request a new link from the confirmation page (or the sign-in page). Verification links expire after 24 hours and each link can only be used once.

Click "Forgot password?" on the sign-in page and enter your email. If an account exists, you'll receive a single-use reset link. It expires after 15 minutes. Completing the reset signs you out of every device so a stolen session can't outlive the password.

Yes - change it from the settings page while signed in. Your current device stays signed in; every other session is revoked.

Only after a recruiter advances a candidate in their pipeline. Up until that point, evaluation runs against anonymized structured data - recruiters see profiles and scores but not names or contact details.

Profile data is used to match candidates to roles and to generate evaluation assessments. It's not sold, not shared outside the platform, and not used for any purpose unrelated to the recruiting process. Candidates can update or remove their data at any time.

No. Each candidate's data is scoped to their own personal tenant. Recruiters from a given organization only see candidates who matched to roles at their company - and only after advancing them. Profiles are never broadcast across companies.

Every uploaded file is scanned for malware before it's processed, encrypted at rest in object storage, and retained only as long as needed to support active matches. We don't upload your original files (PDFs, DOCXs) to third parties - we extract their text within our own systems. That extracted text is then sent to our AI provider (currently OpenAI) to build your structured profile and power evaluations; under our agreement it is never used to train AI models, and the provider retains it only briefly (up to 30 days) for abuse monitoring before deleting it.

Yes - in transit and at rest. All traffic to the platform runs over HTTPS/TLS with HSTS enforced, so data moving between your browser and our service is encrypted. Stored data - your profile, documents, and database records - is encrypted at rest with AES-256.

No. Your data is never used to train AI models. To run the platform, we do send your data - including the text of your resume and your structured profile fields - to our AI provider (currently OpenAI) so it can extract your profile, evaluate fit, and draft summaries. Under our agreement with that provider, your data is not used for training and is not sold; the provider may retain it briefly (up to 30 days) to monitor for abuse, after which it is deleted. It is never used for anything outside the recruiting process.

Yes. You can enable two-factor authentication (2FA) with any standard authenticator app from your account settings. Once on, sign-in requires your password plus a time-based code, and we issue one-time recovery codes in case you lose your device.

Tenant isolation is enforced at the data layer: every recruiter request is scoped to their organization's tenant, and the platform refuses cross-tenant reads by design. Candidate identities are gated separately on top of that, so even within a single organization a recruiter only sees a candidate's name after mutual interest is reached.

Account deletion removes your profile, source documents, and extraction artifacts. Match records visible to recruiters are anonymized at deletion. Some pipeline state may persist in recruiters' audit logs to comply with employment-law record-keeping, but contains no identifying information.