Product

A copilot for the interview itself — not the paperwork around it.

Everything below is generated fresh per role and runs live while you talk. Nothing comes from a question bank, so there’s nothing for candidates to memorise — and every score links back to something that actually happened.

Start free

Live copilot

Probe cue-cards and one-tap scoring, while you stay on the human.

Sharp follow-ups appear as cue-cards — ask them, or dismiss the ones that do not fit. Score each dimension the moment you see the evidence, and the evidence goes on the record with the number.

PROBE · WHY: ESCALATION CRITERIA

“Suppose the client goes silent for three days after your call. What then?”

JUDGMENT UNDER PRESSURE (30%)4/5

12345

EVIDENCE · CHATNamed the client’s exec exposure before offering the fix.

Work-sample simulations

Watch them work, not talk about working.

Six sims, generated for the role, playable mid-interview. Kandevo AI plays the other side live and turns what the candidate does into evidence. The conversational sims carry a mandate card — the resources and authority the candidate actually has — so a promise made outside it is flagged as ungrounded for you to weigh. The triage exercises have nothing to offer, so they carry no mandate. The flag is evidence, not a score: every number on the report is one you set.

URGENT INBOX

URGENT — Meridian threatening to cancel

“Renewal is in 6 weeks. I'm in back-to-backs until 4. What do we do?”

Tests: judgment under time pressure

CLIENT CHAT · PLAYED LIVE BY AI

“What are you going to do about it?”

“Here's what happens in the next 24 hours…”

Tests: de-escalation, ownership

BACKLOG TRIAGE

5 emails, 4 actions, no right answers on the label

Reply now · Delegate · Schedule · Ignore

Tests: prioritisation trade-offs

NEGOTIATION

Hidden brief · only the interviewer sees it

Their real bottom line is 12% — they'll move for a longer term.

Tests: reading interests, holding a line

PRIORITY BOARD

Rank your Monday

1 · Call the at-risk key client 2 · Reply to the small account 3 · Prep the QBR

Tests: what they protect under load

CASE STUDY

Read it, summarise it, then defend it

Three questions follow. One of them the case does not answer.

Tests: whether they claim what they didn't read

INTEGRITY LOG · SIGNALS, NEVER VERDICTS

03:12Paste event in inbox task — 214 characters inserted

07:48Tab/window focus lost during client chat

11:05Long pause followed by burst completion — worth probing

14:30Writing style shifted sharply mid-answer — worth reading back

Integrity signals & length mode

Honest about AI — in both directions.

Real paste, tab-switch and typing-cadence detection, plus AI-likeness checks — always framed as prompts to probe, never verdicts. We don’t hand the candidate an AI to answer with — except in the one exercise where using one is the task — and we don’t pretend a signal is proof.

And every kit fits the calendar you actually have: the length mode scales scenarios, probes and sims from a 15-minute screen to a full hour.

Screening mode

Screen before you meet them.

Send one link. The candidate answers the kit’s inbox task in their own time, the same analyser that reads it during a live interview reads it here, and you decide. No scheduling, no call.

Every score links to the candidate’s actual work, and every shortlist or pass is a human click — nothing is ranked or auto-rejected. Available on Business and up.

On the roadmap

Coming next — and labelled as such.

These are built into the plan and the data model, but they are not shipped yet. We’d rather tell you that than let a demo imply otherwise.

Calibration analyticsIN DEVELOPMENT

See which interviewers grade harsh or soft, per dimension — so a 4/5 means the same thing across your whole team.

See it run for your role.

Start free Try the live demo