STEP 01
Role brief
Recruiter drops the brief, must-haves, and domain constraints so the session starts with intent.
HUMAN DATA OS · CLEO HIRING DEMO
See the hire before you trust it.
Watch one real loop. A specialist arrives, Cleo runs the rubric, scores intent, and routes the decision — with the transcript, the score, and the reason attached to the record. Read-only here. Run it on your own roles in a pilot. The counts read from the platform’s live metrics feed.
ACTIVE SESSIONS
0
COMPLETED SESSIONS
0
ABUSE-FLAG QUEUE
0
AVATAR SESSION
The screen your recruiters actually open.
Not a mockup. This is the production layout. In a live session it streams the transcript and rubric as they happen, so the decision is built in front of you, not after.
Live previewreview session
00:12:47Cleo
Audio stream Candidate mic live
Transcript
- 00:11:12Cleo: Walk me through how you would design a rubric for hallucination detection in clinical summaries.
- 00:11:41Candidate: I would start with a three-axis rubric: factual grounding, omission severity, and clinical actionability.
- 00:12:18Cleo: Good framing. How would you weight omission severity against factual grounding?
Session
Candidate N.K.
Clinical NLP Specialist
INTERVIEWINGabuse-flag: clear
Intent classification
- Interview intent92%
- Domain knowledge84%
- Reasoning quality78%
Shortlist · ranked
- 1Candidate N.K.94Now
- 2Candidate M.P.88Now
- 3Candidate R.S.822 weeks
Evidence record
Transcript
3 live turns
streaming
Rubric
weighted
weighted
Route
Interview
ready
Owner
Recruiting ops
assigned
CANDIDATE INTAKE
From brief to handoff in one pass.
No inbox round-trips, no lost context. Four steps, one candidate record. Consent, transcript, and score travel together, so the final recommendation can be retraced to the moment it was made.
STEP 02
Candidate arrives
Cleo greets the candidate by avatar, captures consent, and confirms the brief in plain language.
STEP 03
Structured exchange
Rubric questions run with intent classification live — the recruiter watches the session without intervening.
STEP 04
Clean handoff
Transcript, rubric score, and recommendation attach to the candidate record. No inbox round-trips.
INTENT CLASSIFICATION
Four signals, scored as they talk.
Intent, domain knowledge, reasoning, and communication move while the session runs — so the routing decision rests on what was said, not a gut read. The figures below are illustrative; in a pilot they read live from your sessions.
- Interview intent92%
- Domain knowledge84%
- Reasoning quality78%
- Communication88%
SESSION READING · ILLUSTRATIVE
ROUTING DECISION
Three paths. One record.
Cleo picks one path and writes down why. The evidence and the reason stay on the session, so a reviewer or an auditor can retrace every decision later — including who or what flagged it.
ROUTE TO INTERVIEW
Signal is strong enough for a structured round. Cleo schedules the interview and attaches the rubric.
- · Rubric fit above 0.80
- · Availability confirmed
- · No abuse flag
ROUTE TO ASSIGNMENT
Candidate is ready for calibrated work directly. Cleo hands them into Workforce with the trust-tier attached.
- · Prior calibration on file
- · Skill taxonomy match
- · Workforce opening flagged
ESCALATE TO A HUMAN RECRUITER
Edge case, abuse signal, or mixed rubric. Cleo pauses, logs the evidence, and pages the recruiter on call.
- · Abuse-flag triggered
- · Rubric contradictions
- · Novel role profile
ADMIN SURFACE
Every persona change is reviewed and signed.
You cannot change how Cleo talks to candidates without a record of who proposed it and who approved it. The admin panel governs the persona, holds safety flags for review, and keeps a version history you can show an auditor.
PERSONA VERSION COMPARE
● REVIEW REQUIRED
| Field | Current | Proposed |
|---|---|---|
| Voice | Concise technical interviewer | Warmer specialist coach |
| Fallback | Text-first recovery | Text + recruiter join |
| Abuse review | Pause above 0.72 | Pause above 0.68 |
ROBOTICS JOBS BOARD · MINI
Robotics openings, with the rights built in.
Pay, bounty, and the required hardware stay attached to each job. Cleo routes a calibrated specialist straight into a posting — so the trajectory data they capture arrives rights-cleared, not stuck in a gig-platform contract. Postings shown are illustrative.
01
RBT-104
Bimanual package-sort demonstration
UR5 + depth camera
CLEO LIVE
Candidate N.K.
PAY
$42/hour
BOUNTY
1.4×
94
02
RBT-113
Warehouse picking · edge cases
Franka Emika + stereo rig
ASSIGNMENT READY
Candidate M.P.
PAY
$38/hour
BOUNTY
1.2×
88
03
RBT-127
Assembly-line handoff trajectories
Operator rig · dual-arm
HUMAN REVIEW
Candidate R.S.
PAY
$48/hour
BOUNTY
1.6×
82
PLANS
What ships at each tier.
STARTER
Cleo avatar session, intake flow, rubric library, and AI Interviewer review — billed per completed session.
SCALE
Adds robotics jobs board, auto-recruit routing, webhook integrations, and admin persona versioning.
ENTERPRISE
Adds per-tenant persona governance, abuse-flag escalation SLAs, and deployment-scoped audit export.
NEXT STEP
Run this on your own roles.
This was read-only. In a pilot, Cleo runs your real briefs and the figures read from your live sessions. Start a pilot, or read how the same reviewed work becomes data you keep.