There’s no script to memorise.
Unfound’s AI job simulation drops each candidate into a live version of the actual job. AI agents push back, the scenario forks at every decision, and every call lands on the record.
Two perfect answers. One of them can’t do the job.
You just flipped a coin.
So does every interview. Both answers are polished, plausible and unverifiable, and either one takes a prep tool seconds to write. In one study, 61% of candidates caught using AI still scored high enough to advance. A transcript shows you who talks well about work. It can’t show you who does it.
Don’t ask how they’d handle it. Watch them handle it.
Generate the job
Unfound builds a scenario from the role: its stakeholders, its data, its pressure points. A new one for every candidate.
Let the agents loose
Autonomous AI agents play the cast. Each has goals, memory and a temper, and each reacts to what the candidate just did.
Score the decisions
Every choice is logged against a consistent rubric. You get a judgment profile, and every score links to the moment behind it.
There’s no script to memorise. Try it.
You’re the candidate. Five decisions. Watch the world fork around every one of them.
decisions. possible worlds. You lived through one.
{{endCounts}} on the record. There’s no right path, only evidence. A real run forks at every decision the candidate makes, so there’s nothing to leak and nothing to rehearse.
{{q.txt}}
{{q.txt}}
{{q.txt}}
{{q.txt}}
{{q.txt}}
Built from the role, not a question bank.
Unfound reads the role you’re hiring for and builds the job around it: who the candidate works with, what data they have, what’s due, and what’s about to go wrong. Every candidate gets a new seed, so no two runs are the same.
Senior Product Lead, B2B analytics. Own the launch plan for our reporting suite. Work with an enterprise client team and engineering. Make the call on scope when data and deadlines disagree.
Agents with goals, memory and a temper.
The candidate negotiates with autonomous AI agents playing the people they would work with. Each wants something, remembers what was said earlier in the run, and reacts to what the candidate just did. They aren’t scripted to be difficult; they’re written to be realistic.
The skeptical VP
Wants the quarter’s story to hold. Escalations land on their desk, and they ask why before they agree to anything.
“The client escalated to me. Why not just give them export?”
The blocked engineer
Wants to ship something that works. Short on time, honest about the trade-offs, and quick to push back on promises they didn’t make.
“Export by Thursday means zero testing. Your call.”
The impatient customer
Wants what they were promised, on the date they were promised it. Remembers every commitment and will walk if it slips.
“We heard there’s a delay. Export ships Thursday, or we walk.”
The analyst
Owns the data and says what it shows, whether or not anyone wants to hear it.
“The Q3 data contradicts the launch plan. Launch is Thursday.”
A real run forks at every decision.
Same scenario, same first message, two candidates. Three decisions later they are living through different jobs. That is why there is nothing to leak and nothing to rehearse, and why the record shows how each person actually works.
“The Q3 data contradicts the launch plan. Launch is Thursday.”
- 03:40 · DecisionPause the plan and re-pull the dataVerifies before committing
- Client agent“We heard there’s a delay. Export ships Thursday, or we walk.”
- 09:12 · DecisionTrade scope: export in phase two, early access nowNegotiates under pressure
- VP agent“The client escalated to me. Why not just give them export?”
- 15:58 · DecisionWalk the VP through the data and the tradeHolds a position with evidence
Kept the client and the date. The record shows verification and negotiation under pressure.
- 02:55 · DecisionKeep the date and flag it after launchTrades certainty for speed
- Engineering agent“The data bug is live in staging. Fix it or ship export. Not both.”
- 08:30 · DecisionShip export and patch laterAccepts a known risk
- World“Launch day. 14% of users hit the data bug.”
- 12:06 · DecisionOwn it: roll back and brief the clientTakes accountability
Shipped on time, then rolled back. Strong on accountability, with verification risks on the record.
One rubric, every candidate.
Scores come from what candidates did, not how they described it. Every decision is logged against the same rubric for the role, and every score links to the moments behind it.
Checks the data before committing to a date, a number or a promise.
07:15 · paused the launch to re-pull contradictory data
Puts the important thing first when everything is urgent, and says what it cost.
19:34 · briefed the VP with the data and kept the account
Trades instead of conceding, and holds a position under pressure.
15:58 · traded scope and put a date on the trade
How consistency is enforced
- One rubric per role, fixed before the first candidate runs.
- Same dimensions on every fork. Scenarios differ between candidates; what gets scored doesn’t.
- Evidence or no score. Every number cites the timestamped decisions that produced it.
- Replayable. Anyone on the panel can check a score against the run.
Reading a score
Did this reliably, including when an agent pushed back.
Did it most of the time. Check the moments where it slipped.
Worth probing in the interview; the replay shows exactly where.
Repeatedly chose against this skill when it mattered.
Show what you can do, not what you can recite.
If a hiring team has invited you to an Unfound simulation, here’s what to expect.
What you’ll see
A short brief, then messages from AI agents playing the people you’d work with: a client, an engineer, a VP. You decide what to do; they react to it.
How long it takes
About as long as a first-round interview. The sample run on this site covers 20 minutes, from the brief to the last decision.
Nothing to cram
There’s no script, no trick question and no single right path. Prep tools won’t help much; your judgment will.
Access
It runs in a web browser, with nothing to install. If you need an adjustment, tell the recruiter who invited you before you start.
People decide
Unfound recommends; the hiring team decides. Every score points to what you actually did, so you’re judged on your work, not a transcript.
