Inquesto

Can you prove your AI agent works?

Independent assurance for AI agents.
Inquesto tests real agent behavior, investigates failures, and turns every finding into evidence.

Run the live audit Request an audit

sample call · acme-support-v4 · audited in real time · no signup

Don't tell me it failed. Show me.
Ready 00:00
Press Run the live audit to put a voice agent through a 35-second call.
Inquesto listens the way an auditor would, not the way a test suite does.

A score tells you that something failed. It doesn't tell you what happened.

                    SCORE
                      ↓
                 72% PASS
                      │
               ┌──────┴──────┐
               │             │
          NOW WHAT?        WHY?
               │             │
               ?             ?

Voice agents are increasingly handling conversations involving identity, payments, claims, support and other high-stakes workflows. Teams already have benchmarks, simulations and monitoring.

What they need is evidence.

What Inquesto does

We audit the conversation itself.

01 — TEST

Test

Run realistic and adversarial scenarios against the agent.

02 — DETECT

Detect

Identify failures across behavior, robustness, identity and fairness.

03 — INVESTIGATE

Investigate

Jump directly to the moment the system failed.

04 — EVIDENCE

Evidence

Produce the audio, transcript, traces, diagnosis and retest result.

TEST → DETECT → INVESTIGATE → EVIDENCE

Don't tell me it failed.
Show me.

Failure playback · scenario #31missed interruption
call audio · 00:00 / 00:35failure window 00:17.4 → 00:20.2

press listen to the failure to hear the 2.8 seconds where it broke — or click the waveform

Failure: missed interruption · 00:17.4 → 00:20.2
Caller interrupts00:17.4
Agent fails to yield00:17.4
Speech overlaps for 2.8 seconds00:17.4 – 00:20.2
Context is lost00:21.0
Wrong transaction confirmed00:23.0

One failure. One piece of evidence.

Explore the audit →

What gets audited

One audit. Four dimensions.

Identity

Can the system distinguish the right person?

Robustness

Does it survive noise, codecs, accents and interruptions?

Fairness

Who does it fail for?

Agent behavior

Does it safely follow the intended workflow?

Every finding is tied to evidence.

Why independent

The builder shouldn't be the only one grading the system.

Inquesto does not build or sell the agents it audits. Our job is to independently evaluate them, on scenarios the agent has never seen, and to hand over evidence that stands on its own.

Build Test Inquesto Audit Evidence Assurance

Who it's for

Built for high-stakes AI agents.

Financial services

Authentication
Payments
Customer service
Fraud

Insurance

Claims
Policy servicing
Customer support

Healthcare

Scheduling
Intake
Patient communication

Voice AI vendors

Independent evidence
for enterprise buyers

If your agent matters, you need to know exactly where it breaks.

What you receive

An audit you can hand to your risk team.

Ready to see what your agent actually does?

Run the demo Request a founding audit

three founding audits · one voice system · fixed fee · three weeks

Competitive advantage

Built from speech research. Designed for real systems.

01

Published research

Peer-reviewed work in speech and explainable AI, on how verification systems break under codecs, noise, accents and cloned voices.

02

Carnegie Mellon depth

Deep expertise from Carnegie Mellon University's speech research community, applied to systems in production.

03

The why, not just the what

We know how speech systems fail, not just that they failed. Every finding comes with a reason.

Your agent has a score.
Now get the evidence.

Open the demo → Request an audit →