People · PeopleOS HR · AI with human control

HR that AI runs, and people decide.

Hire, onboard, run, grow and govern your team from one system of record. AI agents screen in shadow mode, run structured interviews, plan onboarding, explain pay and leave and reconcile payroll. A person approves anything that touches money, employment or the law, and can take over or kill a lane with one click.

Employee portal includedCareers site includedExplainable, appealable AIWhole team, one flat price
OpsIQAcme · workspaceAsk OpsIQDO

Autonomy Control Center

Budgets, compiled policies with step-up dual approval, decision escrow, shadow parity, run replay, circuit breakers and one-click kill switches.

PeopleOS
Run · screening.rank · requisition 0418
  • 1Policy verdict allowed in shadowShadow
  • 2Attempted execution yes, sandboxedOK
  • 3Provider outcome 12 candidates ranked, 41 evidence spansOK
  • 4Proof parity 97% · 0 decisions appliedProven
Lane · screening
Risk / ImpactHigh
Budget / day62% used
30-day record412 executed · 97% proven · 38 reviewed
Tool pathhr.agent.step
ShadowLive
Breakers & budget
Circuit breakersAll clear
Escrow2 proposals await a person
Autonomy budget
Automatic decisions0
Human takeoverArmed
Shadow parity97%
01HireRequisition to offer, with evidence
02OnboardJourneys measured from evidence
03RunTime, leave, payroll, service
04GrowPerformance, learning, skills
05GovernAutonomy, fairness, rights
HR Home · People · Organisation · Job Architecture

An org chart drawn from positions, so a vacancy exists before a person does.

Most HR tools draw the chart from the people in it, so an empty seat is invisible until somebody is hired into it. PeopleOS models the position first. A vacancy is a real record with a level, a profile, skill criteria and an effective date, which is why a reorg can be drawn before it happens and a requisition can point at something that already exists.

  • Effective-dated assignmentsEvery move names its date and its author, so the chart can be read as it was on any day.
  • Identity facets, not duplicatesThe payroll provider's record and the job board's record are linked to the canonical person. Conflicts are resolved explicitly; email is contact data, never the identity key.
  • Access is not the recordSuspending access never erases anyone. Restoring it needs a reason and writes a new audit event.
Surfaces: HR Home · Job Architecture · People Intelligence
Take the tour

Six screens that replace an HR stack.
This is the actual product.

Not a concept. Every window below is the PeopleOS admin as it ships, running. Click a tab or let it play, and watch the AI shortlist candidates, run an interview, plan an onboarding, reconcile a payroll run, and report to you at every step.

OpsIQPeopleOS HRAI Screening Review open
OpsIQAcme · workspaceAsk OpsIQDO

AI Screening Review

Shadow-mode screening runs, evidence-bound outcomes, human corrections, evaluations and the fairness gate.

PeopleOS
RunRequisitionModeCandidatesEvidenceFairness gate
#418Senior EngineerShadow1241 spansOPEN
#417Support LeadShadow927 spansOPEN
#409Payroll AnalystShadow3196 spansCLOSED
What a run may do
ranks with evidencedecides nothing0 automatic decisionsparity 97%

The values shown are representative product data. Your workspace uses its own people, policies, jurisdictions and providers.

The power

Two AI agents do the heaviest work in hiring.
You keep every decision.

Screening four hundred applications and running first-round interviews are the two jobs that eat a hiring team alive. PeopleOS does both, around the clock, and hands you a shortlist with the evidence attached instead of a verdict you have to trust.

AI Screening Agent

Read 400 applications in the time it takes to read four.

The agent reads every application against the criteria you set for that requisition, quotes the exact line of the CV or answer behind each score, and ranks the shortlist. It works overnight, so the pile is triaged before your team logs in. What it never does is reject anybody. It runs in shadow mode: it proposes, you decide, and the count of automatic decisions on your workspace stays at zero.

  • Shortlist by lunchtime, not next weekHundreds of applications triaged against job-related criteria while your team sleeps.
  • Every score cites its evidenceClick any number and read the exact sentence it came from, with the agent's confidence beside it.
  • Correct it and it learns, on your termsA reviewer marks what the agent got wrong with a coded reason. Those corrections train nothing until a different authorised person clears the privacy review.
  • Blind to what it must be blind toProtected attributes, appearance, emotion, accent and disability inference are outside the lane by construction.
Shadow mode Screening run · Senior Engineer 412 applications in INBOX overnight no queue by 9am Ola B. “led the rollback and wrote the post-mortem” 79 Noor A. “owned the migration end to end” 74 Lee M. evidence missing · flagged for a person HUMAN Nothing is applied without you AUTOMATIC DECISIONS 0 this quarter SHORTLIST 12 for review
Live room Live interview · Ada Reaudit AI is interviewingA person took over AR CANDIDATE · CAMERA ON CONNECTED INTERVIEWER AI DO Dana O. · hiring manager WHAT THE AI HEARD · VERBATIM Q2 · Tell me about a rollback you led. “We rolled back in nine minutes.” Dana: Which signal told you to roll back? Take over now ends AI speech for good consent: recording · transcript
AI Video Interview

Run a first-round interview at nine on a Sunday night.

Candidates book themselves into a slot and the agent runs the structured interview you designed, asking the questions on the plan and nothing else. You watch it live, read the verbatim transcript as it happens, and take the room over with one click the moment you want to. That takeover is final: the AI does not speak again in that interview.

  • No more scheduling ping-pongSlots come from your real calendar availability and the provider is reconciled both ways.
  • You see exactly what it heardA verbatim transcript and a scorecard anchored to the criteria, never a judgement about the person.
  • Take over at any momentOne click puts you in the room and permanently ends AI speech in that interview.
  • Fair by defaultRecording, transcription and liveness are separate consents, and a text or audio-only route is always offered.
AI at the centre · Agent lanes

The agents do the work. Every one of them is on a leash you hold.

Each capability is a registered typed tool with a purpose, a jurisdiction, a risk class and a maximum autonomy. This is the lane card the Autonomy Control Center shows, for the lanes that matter most.

AIScreeningscreening.rank
Risk / Impact
High
Mode
Shadow
Budget / day
62% used
30-day record
412 executed · 97% proven · 38 reviewed
ranks with evidencedecides nothing0 automatic decisions
AIInterviewinterview.ask
Risk / Impact
High
Mode
Assist
Budget / day
31% used
30-day record
36 executed · 100% proven · 36 reviewed
asks the planned questionstakeover terminalnever scores a person
AISetup Copilotsetup.plan
Risk / Impact
Medium
Mode
Recommend
Budget / day
9% used
30-day record
58 executed · 100% proven · 58 reviewed
plans in plain languagesays what it assumedruns only on confirm
AIAsk HRaskhr.answer
Risk / Impact
Low
Mode
Assist
Budget / day
18% used
30-day record
1,204 executed · 99% proven · 61 reviewed
answers from your articlescites the sourcenever reads sealed cases
AIPay explanationpay.explain
Risk / Impact
Medium
Mode
Assist
Budget / day
12% used
30-day record
318 executed · 100% proven · 20 reviewed
explains the ledgernever changes itnever moves money
HOffer releaseoffer.release
Risk / Impact
High
Mode
Human only
Budget / day
0%
30-day record
0 executed by an agent
never an agentfour eyesconditional gate
Fact 1Policy verdictDid the compiled policy allow this step, in this jurisdiction, at this risk?
Fact 2Attempted executionDid the agent actually try, or was it stopped by a breaker or a budget?
Fact 3Provider outcomeWhat did the tool or the provider return, verbatim?
Fact 4ProofThe evidence that the outcome happened, kept for replay and for the dossier.
Recruiting · AI Screening Review · Interview Studio · Offers

A candidate travels one governed track, and every stage leaves evidence behind.

Requisition, approval, posting, application, structured review, interview, decision, offer, hire. The AI ranks in shadow mode and drafts the interview plan; a person moves the card. An application that arrives by webhook is a hint until the provider is re-queried, so the board never advances on a rumour.

  • Structured review before the candidate is seenQuestions, score anchors, accommodations and reviewer roles are fixed in the plan first.
  • Evidence spans, not opinionsEvery AI score cites the text it relied on, with its confidence, and a reviewer can correct it with a coded reason.
  • Two people release an offerSealed bands, immutable versions, provider-settled checks, and a conditional gate that escrows on any doubt.
Surfaces: Recruiting · AI Screening Review · Interview Studio · Live Interviews · Scheduling · Offers & Hiring
Onboarding · HR Setup Copilot · Frontline

Nine steps. Eight of them proved themselves. The ninth says so out loud.

A journey is data, not a checklist somebody ticks. Each step is owned by the employee, the manager, the company or the system, and it only counts as done when the evidence arrives: a completed envelope, a delivered laptop, an attended meeting, a learning completion. A step that fails is surfaced as a blocker on its own channel, and the chase follows whoever owns it.

  • Least privilege, by constructionEquipment and access come from approved catalogue items and role templates. No model can compose an arbitrary privilege list.
  • Probation is a dated record with a stateDay 12 of 90, check-in booked for day 30, visible to the person it is about.
  • Say it, then confirm itThe Setup Copilot turns a sentence into named recipes and shows what it assumed. Nothing runs until you agree.
Surfaces: Onboarding · HR Setup Copilot · HR Knowledge Network · Frontline
One record per person · HR Home

Every profile section is dated records with a state.

A canonical person carries employment, worker type, manager, team, location, skills, documents and lifecycle state. Every change names its effective date and its author, so the record can be read as it stood on any day. Pick one.

ARAda Reauditemployment record, as at 30 Aug 2026History
Role
Not employed yet
Worker type
Candidate
Manager
Team
Portal access
None
Identity
2 facets linked

What this entry did. Created the canonical person and linked the payroll and job-board identities as facets rather than merging them. Author: Priya S.

ARAda Reauditemployment record, as at 02 Sep 2026Superseded
Role
Senior Engineer · L4 from 15 Sep
Worker type
Employee · full time
Manager
Dana O.
Team
Engineering
Portal access
Not until start date
Documents
Contract v1

What this entry did. Recorded the approved offer with a start date two weeks out. The employment exists on the record before the person starts. Authors: Dana O. and Priya S.

ARAda Reauditemployment record, as at 15 Sep 2026Current
Role
Senior Engineer · L4
Worker type
Employee · full time
Manager
Dana O.
Team
Engineering
Portal access
Provisioned
Documents
Contract v1

What this entry did. Employment took effect on its own date, and portal access was provisioned separately. Suspending that access later would never erase this record. Author: Dana O.

ARAda Reauditemployment record, as at 01 Oct 2026Pending · 1 of 2
Role
Senior Engineer · L5 proposed
Worker type
Employee · full time
Manager
Dana O.
Team
Engineering
Portal access
Unchanged
Documents
Contract v2 · draft

Dated in the future, so it is not in force. The band move exists as a record with its own effective date and is waiting on a second approver. Read today, the same person is still L4. Proposed by Dana O.

Nothing is overwritten. A correction is a new entry with its own date and author. 4 entries on this record
  • Effective-dated assignmentsThe reporting tree and every assignment carry a date and an author, so a reorg can be drawn before it happens.
  • Access is not the recordSuspending access never erases the record. Restoring it needs a reason and creates a new audit event.
  • The org chart comes from positionsA vacancy can be drawn because a position exists before a person fills it.
Hire · Offers & Hiring

Sealed bands, four eyes,
and a gate the AI cannot open alone.

Salary bands are sealed. Offer versions are immutable and a human approves before any send. Background checks, identity verification and right-to-work run through consent-bound connectors and only the provider's own answer changes the record. The conditional auto-offer gate escrows on any doubt.

OpsIQAcme · workspaceAsk OpsIQDO

Offers & Hiring

Sealed salary bands, immutable offer versions, human approval before any send, and the conditional auto-offer gate that escrows on any doubt.

PeopleOS
Band · L4 Engineering · sealedposition in band · figures hidden from this view · within policy
Who may see figures
compensation permission onlynot on the broad API
Approvals required2
Salary bandSealed
Offer versionImmutable
A callback never settles a signature. PeopleOS asks the provider for the authoritative envelope status and stores the hash of the signed result.
OpsIQAcme · workspaceAsk OpsIQDO

Scheduling

Interviews, check-ins and HR appointments on real calendar events: proposed by AI or a human, self-scheduled by candidates, rescheduled and cancelled from either side.

PeopleOS
Meetings · this week
Interview · Ada R. · Tue 14:00Google Calendar · reconciledbooked
Check-in day 30 · Ada R.Microsoft Calendar · proposed by AIconfirm
Interview · Maya C.rescheduled by the candidatemoved
CalendarsGoogle + Microsoft
SlotsLive availability
Provider reconciliation for Google and Microsoft calendars. Slots come from the interviewers' live availability.
Run · HR Inbox, Ask HR, HR Knowledge Network

Employee service is in the box. Not sold per agent.

One private conversation fabric: employee threads HR can start first, case conversations that inherit the case wall, direct messages between the team, announcements with read receipts, and an assistant that answers from your own published handbook with the source cited.

OpsIQAcme · workspaceAsk OpsIQDO

HR Inbox

One private conversation fabric: employee threads HR can start first, case conversations that inherit the case wall, direct messages between the team, announcements with read receipts.

PeopleOS
PSPriya Sato · parental leave questionstarted by HR · 2 replies · privateOpen
MOMarco Obi · shift swap policyAsk HR answered · cited: Shift swapsAnswered
LMCase 0417 · conversationlegal hold · retention paused · sealedHold
ALLAnnouncement · new expense policyread by 231 of 248 · receipts keptSent
Metered agents0
Case wallInherited
Ask HRgrounded · cited
Yes. Swaps need your manager's approval before the shift starts. Request it from Time & leave and your manager gets it in their queue.Source: Shift swaps (your published article)
Answers come from your handbookNever from sealed cases
HR Knowledge Network
Leave & absence packinstalled · 14 articles · 2 edited by you
Frontline & shifts packinstalled · 9 articles
Germany policy packupdate available · keeps your edits
Grow · Performance, Learning & Talent, Skills, Job Architecture

Reviews anchored in evidence.
Careers drawn from positions.

Goals with evidence links, rubric-anchored review cycles, append-only calibration and a private manager coach on one side. Courses, paths, an explainable internal marketplace with employee opt-out and appeal, skills with evidence and succession on the other. Job families, levels and profiles are proposed by the agent from what you already have, approved by a second person.

OpsIQAcme · workspaceAsk OpsIQDO

Performance & Goals

Goals with evidence links, rubric-anchored review cycles, append-only calibration and a private manager coach. No hidden productivity score exists anywhere.

PeopleOS
Goals · Ada Reaudit · Q3
Ship the payroll connectorevidence: 3 changes, 1 incident closed
Mentor two engineers4 sessions logged
Hidden scoreNone
CalibrationAppend only
No hidden productivity score exists anywhere in the product.
OpsIQAcme · workspaceAsk OpsIQDO

Learning & Talent

Courses, paths and mandatory training; an explainable internal marketplace with employee opt-out and appeal; evidence-backed succession planning.

PeopleOS
Paths
Security foundationscompletion received
Leading a teamdue in 30 days
Opt outAnytime
AppealTo a person
Succession slates list people with evidence, not favourites.
OpsIQAcme · workspaceAsk OpsIQDO

Skills Intelligence

The workforce read as skills: what it holds with evidence, what it needs, the gaps between, and the two markets that move people: gigs and mentorship.

PeopleOS
Skills map · Ada Reaudit · with evidence
PHP 5Payroll systems 3Incident lead 4Mentoring 2SQL 4
Proposals the person confirms. Matches explained in words.
OpsIQAcme · workspaceAsk OpsIQDO

Job Architecture

Families, levels, profiles with skill criteria, every position mapped and the paths between levels: proposed by the agent from what you already have, approved by a second person.

PeopleOS
Engineering family
L3 EngineerL4 Senior EngineerL5 Staff EngineerL6 Principal
The org chart is drawn from positions, so a vacancy can be drawn.
Grow · Listening, People Intelligence, Pay Equity, Benchmarks

Groups only. Never a person.

Pulses run themselves on a cadence and at the moments that matter, anonymous by construction. Dashboards suppress any cell below the minimum group size. Pay equity is attested and nobody is named. Benchmarks compare buckets of at least five other opted-in workspaces, and opting out withdraws everything.

eNPS by team and monthAnonymous pulses · suppressed below the group floor
+31 company · 12 months
JanMarMayJulSepNov
Engineering
Support
Finance
Warehouse
Sales
LowerHigher suppressed
Autonomy Control Center · Agent lanes

Every lane has a ceiling and a daily budget. You can see both from across the room.

An agent lane is not a switch. It carries a risk class, a maximum autonomy the policy will allow, a budget that is spent as it works, and a thirty-day record of what it executed, what was proven and what a person reviewed. Offer release stands at zero on purpose: no agent may execute it, in any mode, ever.

  • A reason is required to promote a lanePromotions are audited, and anything above Assist needs step-up dual approval from a second person.
  • Breakers, incidents and kill switchesA breaker, an incident, a human takeover or a jurisdiction change stops a lane where it stands. The kill switch arms on the first press and fires on the second.
  • Run replayEvery step in order: what was read, what the policy said, what the provider returned, and the proof.
Surfaces: Autonomy Control Center · AI Governance · AI Metering
Govern · the autonomy ladder

Autonomy is a ladder with a budget, a breaker and a kill switch.

Human only, Shadow, Assist, Recommend, Approve to execute, Bounded autonomous. A lane climbs only with a written reason and, above Assist, step-up dual approval. Compiled policies, escrow, shadow parity and run replay sit beside budgets, circuit breakers and kill switches, and a human takeover ends an AI lane on the spot.

01

Human only

No agent may execute. Offers, discipline, compensation and anything a policy pack forbids live here.

02

Shadow

The AI drafts and scores in parallel; a person compares and corrects. Nothing it says is applied.

03

Assist and above

Low-risk steps run inside a budget. Recommend, Approve to execute and Bounded autonomous each need a reason and dual approval.

Govern · EU AI Act dossier, AI Quality & Rights, Policy Packs

Readiness is computed. Never declared.

The dossier shows what a deployer of high-risk employment AI must be able to show, per system, computed from live records, with the action that closes each gap and a document to hand to an authority. Policy packs say which law applies where, who signed it off, when it takes effect and when it runs out, and what the product refuses to do in each country until somebody signs.

OpsIQAcme · workspaceAsk OpsIQDO

EU AI Act dossier

What a deployer of high-risk employment AI must be able to show, per system, computed from live records, with the action that closes each gap and a document to hand to an authority.

PeopleOS
AI systemReadinessOpenNext action
Screening assistant1fairness release
Interview agent2worker notice, log retention
Pay explanation0ready
Readiness views are computed from evidence and cannot be set to green by hand.
OpsIQAcme · workspaceAsk OpsIQDO

AI Quality & Rights

Sampled decision quality, override pressure, the appeals desk with its independent reviewers, and the learning loop behind its privacy door.

PeopleOS
Sampled quality96%defects in 250 sampled decisions
Override pressure4.1%reviewer overrides this month
Appeal outcomes1 of 3upheld
Policy packs: Germany signed, expires 2027-03. France unsigned, so the screening lane refuses to run for French candidates.
Built to be defensible

A people system people can argue with.
And win.

Anyone can put a model in front of an HR file. The hard part is a system that holds up when the person it decided about asks why, and that never quietly rewards the fastest path over the fair one.

62%used today
Autonomy budget

Each lane spends a daily allowance. When it runs out, the lane stops.

85%readiness
EU AI Act dossier

Computed per AI system from live records, never declared by hand.

97%parity
Shadow parity

How closely the screening agent agreed with your reviewers this month.

100%human
Sensitive decisions

Money, legal and employment outcomes signed by a person, every time.

Autonomy What the AI is allowed to do, lane by lane Human onlyShadowAssistRecommendApprove to runBounded CEILING policy will not allow this lane past Shadow DAILY BUDGET · SCREENING 62% used · 412 runs · 97% proven BREAKER closed · all clear KILL SWITCH ARMED press again to fire EVERY RUN REPORTS Policy verdict Attempted execution Provider outcome Proof, kept for replay never one tick that hides three HUMAN ONLY, IN EVERY MODE Offers · discipline · compensation · anything a country pack forbidsno agent may execute these, and no setting can change that

Sensitive decisions keep a human path

Money, legal and employment decisions are never executed by a model. AI can draft, explain, summarise, classify or recommend inside the configured risk and jurisdiction policy, and a permitted action is not reported as completed until execution and outcome evidence exist.

Not an assessment input, ever

Protected attributes, appearance, emotion, accent and disability inference are outside the screening and interview lanes by construction. The video connector class has no interface for emotion, attention, attractiveness, race, disability or appearance analysis.

Consent per purpose

Recording, transcription and biometric liveness each require their own purpose-specific consent, and withdrawal applies to that purpose. Candidates can always use a text accommodation path.

Explanations from the record

An explanation is generated from the sealed decision record and the policy evidence, not invented by a model. Appeals go to an independent human reviewer. Correction, access and deletion requests stay available to candidates and employees.

A callback is a hint

Background checks, job boards, e-signature, identity and payroll callbacks are notifications to re-query. The provider's status result, not the callback claim, changes the PeopleOS record. Drift is recorded as a conflict, never silently overwritten.

Corrections train nothing on their own

Reviewer corrections can become a proposed dataset, but nothing is trained until a different authorised person clears the privacy review. The learning loop sits behind its own privacy door.

Sealed stays sealed

Employee-relations material, whistleblower reports and case attachments are never exposed through the ordinary profile, search, analytics or AI context. A broadcast is never a substitute for a confidential case conversation.

Technical green is not legal green

Jurisdiction packs must name an accountable legal owner, cite actual legal sources, pass their tests and be released. Release evidence distinguishes code conformance, sandbox evidence, real provider evidence, security review, legal approval, fairness and accessibility.

Every readiness view is computed from evidence and cannot be set to green by hand. Absence is never converted into a passing record.

Govern · Employee Relations

Sealed cases. A vault nobody can browse.

Grievances on sealed cases, conflict-checked investigations with human-only findings, four-eyes discipline with appeal, and the whistleblower vault. Sealed material is never exposed through the ordinary profile, search, analytics or AI context.

GrievancesInvestigationsDisciplineWhistleblower vault
Five front doors

Employees, managers, candidates, frontline teams and developers
each get their own surface.

The employee portal at /hr/your-workspace, the careers site at /careers/your-workspace, the clocking kiosk at /kiosk, the admin surfaces and the General API. All of them re-check the HR entitlement server-side.

OpsIQWho sees whatEmployee portal open
Hi AdaOverview · Tuesday
2 urgent taskssign the expense policy · book your day-30 check-in
Leave · 3 daysrequested → approved by Dana
Your first days8 of 9 · from evidence
My pay, explainedSeptember · generated from the record
Announcementnew expense policy · read receipt sent
HomeTimePayCompanyMe
Employee portal · /hr/your-workspace

Seatless, mobile-first, scoped to the person signed in.

OverviewMy profileHR SupportMy tasksYour first daysTime & leaveSkills & growthMy pay, explainedCompanyDecisions about me
  • Honest cardsA failed load says so and offers a retry. An outage never looks like "you have nothing".
  • Find a colleague, the org chart, and who may see whatVisibility the employer sets, a phone number hidden by default.
  • Photos, documents and identity papers go through the vaultNever a public upload folder.
  • Your own domainMap a host and the portal answers there, verified against the host map.
Connectors · certified per account

Your providers keep their jobs. PeopleOS keeps the record.

16 connector classes with bounded capabilities and 17 signed packages on the marketplace: a sandbox kit for every class plus the production-capable Provider Gateway. DocuSign, Google and Microsoft Calendar, Zoom, Slack and Teams supply the rest.

ClassPurposeHighest-impact capabilities
Job boardPublish and maintain vacanciespost, update, close, receive application
Background checkConsent-bound checkscreate, receive result, cancel
Identity verificationOptional identity and liveness proofcreate session, verdict, delete subject
E-signatureExternal envelopessend, status, void, verify completion
PayrollWorker facts and reconciliationpush event, read state, exceptions
BenefitsEligibility and enrolmentpush eligibility, read enrolments
LearningCourses and completioncatalogue, assign, completion
CalendarAvailability and interviewsread, create, update, cancel
VideoGoverned interview mediaroom, token, retention, deletion
AssessmentJob-related assessmentscreate, invite, result, accommodation
Tax and right to workLegal verificationcreate, status
Equipment and ITEquipment and accessprovision, status, revoke
Expense paymentReimbursement handoffsubmit, status
DirectoryWorkforce directoryread, lifecycle push, reconcile
Data warehouseAggregate exportexport, status
MessagingSend HR notificationssend, delivery status
Capability atlas

Every one of the 31 admin surfaces,
and what each one actually does.

Grouped by journey. Each card lists the capabilities that surface ships today, in the words the product itself uses.

Hire6 surfaces

Recruiting

  • Requisitions with title, openings, justification, budget context, hiring team, location and jurisdiction
  • Approval before a posting goes anywhere; approve as the only approver when policy allows it
  • Job boards through certified connectors: post, update, close
  • Applications arriving by callback are re-queried before a candidate moves
  • Accommodations and appeals as separate governed workflows, upheld or not upheld on the record
  • Talent pools

AI Screening Review

  • Shadow-mode screening runs that decide nothing
  • Evidence spans, confidence and human corrections with a coded reason
  • Adversarial verification and versioned labels
  • The fairness gate decides whether a run may influence anything
  • Protected attributes, appearance, emotion, accent and disability inference are never inputs

Interview Studio

  • Structured plans with job-related questions and score anchors
  • Accommodations and reviewer roles set before candidate evidence is seen
  • Sandboxed AI sessions with human takeover
  • Transcripts and content-bound scorecards that record evidence, not impressions

Live Interviews

  • Watch an AI-run interview as it happens
  • Take it over the moment you want to; takeover is final for AI speech
  • See exactly what the AI heard, never what it thought of anyone
  • Answer in writing, audio only, a person instead, live captions
  • Consent per purpose: recording, identity check, written transcript

Scheduling

  • Real calendar events for interviews, check-ins and HR appointments
  • Proposed by AI or a human, self-scheduled by candidates
  • Booking links and working hours
  • Reschedule, cancel and provider reconciliation for Google and Microsoft calendars

Offers & Hiring

  • Sealed salary bands and position-in-band views
  • Immutable offer versions and human approval before any send
  • Background checks, identity and right-to-work settled by the provider, not the callback
  • The conditional auto-offer gate that escrows on any doubt
Onboard3 surfaces

Onboarding

  • Templates and dated task plans with employee, manager, company and system-owned steps
  • Progress derived from evidence, never asserted
  • Least-privilege provisioning from approved catalogue items and role templates
  • Probation as a dated record with a state, and the 30/60/90 check-ins
  • Every failure surfaced, never silently fixed

HR Setup Copilot

  • Say what you want set up
  • A plan in plain language: what it understood, what it assumed
  • Nothing runs until you confirm, through the same services as the manual screens

HR Knowledge Network

  • Your own FAQs and curated packs of HR practice that become your published handbook
  • What Ask HR may tell your employees, and nothing else
  • Cited by Ask HR and the portal; pack updates keep your edits
Run6 surfaces

HR Home

  • People with canonical identity, employment record, worker type, manager, team, location, skills and lifecycle state
  • Organisation with effective-dated assignments
  • Cases, rights, documents and records on one operator surface
  • Duplicate external identities linked as facets, never merged blindly

Time & Leave

  • Schedules, shifts and time approvals
  • Leave types, policies, holidays and requests
  • Deterministic auto-approval only where a policy explicitly opts in
  • Conflict rules and calendar behaviour per leave type

Frontline

  • Shared-device clocking kiosks with PINs and badge codes
  • Shift swaps between colleagues, approved by a manager
  • The installable employee app
  • Sign-in design for the device

Payroll & Expenses

  • Imported payroll runs reconciled against the HR record and approved compensation
  • Four-eyes pay changes
  • Benefit enrolments
  • Expense claims handed to reimbursement; OpsIQ never moves money

HR Inbox

  • Employee threads HR can start first
  • Case conversations that inherit the case wall
  • Direct messages between the team and announcements with read receipts
  • Legal hold, retention and deletion on the thread itself

AI Metering

  • Where PeopleOS AI spend goes by feature, day and model, estimated from the usage log
  • Lane budgets
  • A monthly cap that tells a person, never silently
  • A CSV export
Grow9 surfaces

Performance & Goals

  • Goals with evidence links and check-ins
  • Rubric-anchored review cycles
  • Append-only calibration
  • A private manager coach and recognition; no hidden productivity score exists

Learning & Talent

  • Courses, paths and mandatory training
  • An explainable internal marketplace with employee opt-out and appeal
  • Evidence-backed succession planning

Skills Intelligence

  • The workforce read as skills, with evidence
  • Supply and demand, the gaps between, inference the person confirms
  • Two markets that move people: gigs and mentorship
  • A learning map

Job Architecture

  • Families, levels and profiles with skill criteria
  • Every position mapped and the paths between levels
  • Proposed by the agent from what you already have, approved by a second person
  • Build by hand and job descriptions

Listening

  • Pulses that run themselves on a cadence and at the moments that matter
  • What the workforce keeps saying, as themes and a trend, never as a quote
  • Schedules and campaigns

Engagement & Intelligence

  • Anonymous pulses with one-way links
  • Suppressed workforce dashboards and cohort-only turnover
  • Headcount planning, the scenario lab and reproducible compliance packs

People Intelligence

  • Every journey analysed as one system: what is moving, what is unusual, where it is heading
  • Command deck, people, hiring, service and time, engagement and growth, AI governance, behaviour
  • An analyst you can ask; for groups, never for a person

Pay Equity

  • Where people sit in their bands
  • Whether self-declared groups are paid alike for like work
  • Measured, suppressed, attested; nobody is named; a ledger

Benchmarks

  • How this workspace compares with others like it
  • Buckets of at least five other opted-in workspaces, aggregates only
  • Opting out withdraws everything
Govern7 surfaces

Autonomy Control Center

  • Mode and budgets per lane
  • Compiled policies with step-up dual approval
  • Decision escrow and sampling
  • Runs, shadow parity, replay and one-click kill switches
  • Agent lanes with Risk / Impact, Budget / day, 30-day record and tool path

AI Governance

  • Reason-coded corrections and immutable datasets
  • The five-dimension release gate and the independent verifier
  • Drift monitors, system cards and incidents
  • Connectors and readiness

AI Quality & Rights

  • Quality board: sampled decision quality and override pressure
  • The appeals desk with independent reviewers
  • The learning loop behind its privacy door

EU AI Act dossier

  • Readiness matrix per AI system, computed from live records
  • Oversight and notices, impact assessment
  • The dossier to hand to an authority, with the action that closes each gap

Policy Packs

  • Which law applies where, who signed it off, when it takes effect and when it runs out
  • Markets, signing off a pack, law changes
  • What the product refuses to do in each country until somebody signs

Employee Relations

  • Grievances on sealed cases
  • Conflict-checked investigations with human-only findings
  • Four-eyes discipline with appeal
  • The whistleblower vault

Launch & Leadership

  • The fifteen gates a world-leading claim has to clear, each computed from live evidence
  • Personas: what each one actually experiences
  • The org graph, run ledgers and the truth register
One system of record

What you would otherwise buy as eight tools.

An applicant tracking system, an HRIS, an onboarding tool, time and attendance, a learning platform, a performance tool, an engagement survey tool and an employee service desk, each with its own copy of your people. PeopleOS keeps one record and the providers you already trust.

Permission-scopedWorkspace-specific policy packsDeterministic gatesHuman decisions only
ATSHRISOnboardingTime & attendanceLMSPerformancePulse surveysEmployee service desk
OpsIQPeopleOS
HR
One record per person
Payroll providerJob boardsE-signatureCalendars
Full feature list

Everything in PeopleOS HR.

Every admin surface and public surface, grouped by journey. ★ marks a stand-out.

SurfaceWhat it does
Hire
Recruiting ★Requisitions, job postings, the candidate pipeline, accommodations and appeals.
AI Screening Review ★Shadow-mode screening runs with evidence, corrections, evaluations and the fairness gate.
Interview StudioStructured interview plans, sandboxed AI sessions with human takeover, transcripts and scorecards.
Live Interviews ★Watch an AI-run interview as it happens, take it over at any moment, and see exactly what the AI heard.
SchedulingReal calendar events for interviews and check-ins, candidate self-scheduling, reschedule, cancel and provider reconciliation.
Offers & Hiring ★Sealed salary bands, offer approvals, checks, decisions and the conditional auto-offer gate.
Onboard
Onboarding ★Templates, task plans, least-privilege provisioning, probation and new-hire check-ins, with progress measured from evidence.
HR Setup CopilotSay what you want set up; get a plan of named recipes with what was understood and what was assumed; confirm and it runs.
HR Knowledge NetworkCurated packs of HR practice installed as your own published articles, cited by Ask HR and the portal; updates keep your edits.
Run
HR HomePeople, organisation, cases and rights, the PeopleOS operator surface.
Time & Leave ★Schedules, shifts, time approvals, leave policies, holidays and requests.
FrontlineShared-device clocking kiosks with PINs and badge codes, shift swaps approved by a manager, and the installable employee app.
Payroll & Expenses ★Payroll reconciliation, four-eyes compensation, benefits and expense claims.
HR InboxEmployee threads HR can start first, case conversations, direct messages and announcements with read receipts.
AI MeteringPeopleOS AI spend by feature, day and model, lane budgets, a monthly cap that alerts a person, and a CSV export.
Grow
Performance & Goals ★Goals, evidence-anchored review cycles, calibration and the private manager coach.
Learning & TalentCourses, paths, the explainable internal marketplace, skills and succession.
Skills IntelligenceThe workforce as skills: what you have with evidence, what you need, the gaps, gigs and mentorship.
Job ArchitectureFamilies, levels, profiles with skill criteria, mapped positions and career paths, proposed by the agent, approved by a second person.
ListeningPulses on a cadence and at lifecycle moments; eNPS trend, theme counts and drop alerts, anonymous by construction.
Engagement & IntelligenceAnonymous pulses, suppressed dashboards, planning, scenarios and compliance packs.
People Intelligence ★Trends, anomalies, forecasts, a health score, the 3D workforce landscape and an analyst you can ask; groups only, never a person.
Pay EquityPosition in band by unit and level, and like-for-like pay gaps between self-declared cohorts; suppressed, attested, nobody named.
BenchmarksOpt-in comparison with workspaces like yours from buckets of five or more; aggregates only, opting out withdraws everything.
Govern
Autonomy Control Center ★Budgets, compiled policies, escrow, shadow parity, run replay, breakers and kill switches.
AI GovernanceCorrections, datasets, the five-dimension release gate, drift, verifier, cards and incidents.
AI Quality & Rights ★Sampled decision quality, override pressure, the appeals desk with independent reviewers, and the learning loop behind its privacy door.
EU AI Act dossierDeployer readiness per AI system computed from live records, with the document to hand to an authority.
Policy PacksWhich law applies in each country, who signed it off, when it expires, and what the product refuses to do until somebody answers.
Employee Relations ★Grievances, investigations, four-eyes discipline and the whistleblower vault.
Launch & LeadershipThe fifteen computed launch gates, measured persona experience, the org graph and the run ledgers.
Public surfaces & developers
Employee portal ★Seatless, mobile-first self-service at /hr/your-workspace or on your own domain, scoped to the verified person.
Careers siteOpen roles, applications, candidate self-scheduling and AI disclosure at /careers/your-workspace.
Clocking kioskShared-device clock in and out with a PIN or a badge code.
Connector packagesA signed sandbox package for every connector class plus the production-capable Provider Gateway, certified per account from C0 to C4.
General APIThe hr.* actions with contracts, scopes, idempotency, confirmation and dry runs; OpenAPI and Postman from the same registry.
Pricing · PeopleOS HR

Whole team, one flat price.
Sized by headcount, not by agent.

HR agents work inside the seats your plan already has and are never counted. Employees use the seatless portal. Pick the tier that covers your active worker records.

Included in Enterprise 14-day free trial Paid plans only
HR 50 50 people covered

PeopleOS HR for up to 50 people: recruiting, onboarding, time & leave, payroll, performance and the employee portal. Whole team, one flat price.

$19/ month $190 / year
  • 5 open requisitions
  • 200 new candidates per month
  • 300 AI interview minutes per month
  • 500 AI actions per month
  • 5 document storage (GB)
HR 250 250 people covered

PeopleOS HR for up to 250 people: recruiting, onboarding, time & leave, payroll, performance and the employee portal. Whole team, one flat price.

$49/ month $490 / year
  • 20 open requisitions
  • 1,000 new candidates per month
  • 1,500 AI interview minutes per month
  • 2,500 AI actions per month
  • 25 document storage (GB)
HR 1000 1,000 people covered

PeopleOS HR for up to 1,000 people: recruiting, onboarding, time & leave, payroll, performance and the employee portal. Whole team, one flat price.

$99/ month $990 / year
  • 60 open requisitions
  • 4,000 new candidates per month
  • 6,000 AI interview minutes per month
  • 10,000 AI actions per month
  • 100 document storage (GB)

Available to add on Launch, Growth, Business, Scale. Not on free plans, and not while your plan itself is still on trial. One tier at a time; switch tiers from your account when you grow. Enterprise customers have it included with unlimited people. Every figure on this page is the live price in your selected currency.

vs Zendesk Employee Service

Zendesk Employee Service charges per agent. PeopleOS HR charges per team.

Zendesk Employee Service prices its employee service desk per agent per month, billed annually. OpsIQ HR is one flat workspace price on monthly billing, HR agents uncounted, and it includes recruiting, onboarding, payroll reconciliation, performance and the employee portal, which an employee service desk does not sell.

TierOne HR agentThree HR agentsWhole team
Zendesk Employee Service Suite Team $29 $87 grows with every agent
Zendesk Employee Service Suite Growth $59 $177 grows with every agent
Zendesk Employee Service Suite Professional $99 $297 grows with every agent
OpsIQ HR 50 $19 $19 $19 · up to 50 people
OpsIQ HR 250 $49 $49 $49 · up to 250 people
OpsIQ HR 1000 $99 $99 $99 · up to 1,000 people

For a single HR agent, the smallest PeopleOS HR tier already costs less per month than the entry tier of Zendesk Employee Service, and the gap widens with every agent you add. Zendesk Employee Service figures are its published per agent per month, billed annually rate, checked on 2026-09-14 (source). OpsIQ figures are the monthly-billing price of each tier, converted into your selected currency.

PeopleOS HR FAQ

Bring your people in. Keep the decisions.

Packaging, control, rights, providers and the API. Every answer here is also published as structured data.

A complete HR system of record and workflow layer inside OpsIQ: recruiting, AI screening review, structured and live interviews, offers, onboarding, time and leave, payroll reconciliation, benefits, expenses, performance, learning, skills, job architecture, employee relations, listening, people intelligence, an employee portal, a careers site and a clocking kiosk. It is built for companies that want AI to do the drafting and the sorting while a person keeps every sensitive decision.
It is included in Enterprise. On every other paid plan you add it as a module with a tier sized to your headcount. Every tier starts with a 14-day free trial, once per workspace. Free plans cannot add it, and a plan that is still on its own trial has to become active first.
Neither. A tier is one flat price for the whole workspace, sized by the number of active worker records it covers. HR agents work inside the seats your plan already has and are not counted separately, and employees use the seatless portal.
Nothing sensitive without a human path. PeopleOS AI runs through registered typed tools, each with a purpose, a jurisdiction, a risk class and a maximum autonomy. Money, legal and employment decisions keep a person in the loop, screening runs in shadow mode with evidence and corrections, and a live interview can be taken over at any moment. Budgets, circuit breakers and kill switches stop a lane; run replay shows exactly what happened.
Yes. Anyone affected can ask for an explanation, which is generated from the sealed decision record and the policy evidence rather than invented by a model, file an appeal that goes to an independent human reviewer, correct their data, and request access or deletion. Accommodations are a separate governed path.
Employees use the mobile-first portal at /hr/your-workspace, or on your own domain once it is mapped, with their record, requests, documents, pay explanations, leave, learning, the company directory, rights and onboarding scoped to their verified identity. Candidates apply on the careers site at /careers/your-workspace and can schedule their own interviews. Shared devices clock in and out on the kiosk with a PIN or a badge code.
No. PeopleOS is not a payroll ledger, a legal adviser, a background-check bureau or an identity provider. Certified connectors exchange only the bounded facts each workflow needs: approved worker events go to payroll and reconciliation state and exceptions come back, a callback never settles a signature until the provider confirms it, and OpsIQ never moves money.
Yes. The General API publishes the hr.* actions with the same dispatcher, contracts, permissions, OpenAPI and Postman sources as the rest of the platform. Writes carry an idempotency key, high-risk writes need an explicit confirmation, and dry_run validates a call without executing it. Inbound provider callbacks are authenticated and replay-protected, and PeopleOS does not send outbound HR webhook events today.