Containment is not resolution.
Your deflection agent says it handled the call. The customer called back angry. Roark scores whether the issue was actually resolved, whether the escalation was clean, and whether the caller was getting more frustrated — on the audio, not just the words.
Backed by YCCaller: Fine. Whatever. Let’s just do it your way.
Agent: Great! Let’s continue.
Rising frustration — Words read neutral; the audio model hears a customer about to churn.Emotion
Scoring production voice AI for teams at
§01 · When the call goes wrong
Here's how a support call quietly fails.
Each one counts as a contained call in your dashboard and a repeat call in your queue — and most are inaudible to a tool that only reads the transcript.
Vanity containment
Caller: So my refund still hasn’t shown up.
Agent: Thanks for confirming! Is there anything else I can help with?
The agent closed the call without solving anything and logged it as contained. The customer calls back in an hour. Real resolution — not hang-up rate — is the metric that matters, and it’s the one most teams never score.
ResolutionThe dropped handoff
Caller: Just put me through to a person.
Agent: I can help with that. What is your account number again?
A clear escalation request got swallowed by another collection loop. The handoff never carried context, so the human starts from zero and the customer repeats everything. Roark fails the turn that ignores or fumbles the escalation trigger.
EscalationThe confident wrong answer
Caller: Is the annual plan refundable after 30 days?
Agent: Yes, you can get a full refund anytime in the first year.
Stated with total confidence and completely invented. A hallucinated policy answer becomes a chargeback, a complaint, or a promise your team has to honor. Roark flags assertions that aren’t grounded in the knowledge base.
HallucinationThe loop that won’t end
Caller: I already gave you the order number twice.
Agent: No problem — can you read me the order number?
The agent re-asked for information the caller already provided, three turns running. Repetition is the fastest way to make a calm customer furious, and it reads as a perfectly polite transcript.
RepetitionThe slow boil
Caller: Fine. Whatever. Let’s just do it your way.
Agent: Great! Let’s continue.
“Fine” and “whatever” read neutral on the page. In the audio the caller is clipped, tense, and about to churn. Only an audio model hears the rising frustration a transcript-only tool scores as agreement.
Emotion§02 · From caught to fixed
Roark catches every one of these — and proves the fix.
Each failure above is filed with its evidence, becomes a repeatable simulation until a candidate passes, and is verified on your next thousand live calls.
Your fix, replayed against the exact failures above.
Every change explicit and diffed — you apply it.
You ship — Roark confirms the metric moved on live calls.
you ship it — Roark verifies every call, with no change to your CCaaS
…and the loop runs again on the next call.
§03 · Simulate before launch
Break it in staging,
not in production.
Run your agent against hundreds of simulated callers — realistic personas, accents, background noise and edge cases — and get every conversation scored before a customer ever dials in.
Scenarios & personas
Hundreds of simulated callers — the angry one, the rambler, the interrupter — built from your real call types.
45 languages & accents
Native accents, code-switching and background noise — in every market your agent answers.
Load & health tests
Peak-volume concurrency and always-on health checks, so the agent that passed in staging survives launch day.
Run it in CI
Every prompt or model change runs the suite before it merges — quality gates for conversations, not just code.
1 failure filed as an issue — fix it before launch, not after
§04 · Evals & observability
64+ metrics. Your models,
not just an LLM.
Every production call scored as it lands — issues filed, alerts fired, dashboards and OTEL traces on tap, for voice calls and chat threads alike. And where most tools grade a transcript with an LLM, Roark runs purpose-built audio models on the call itself, measuring what your customer actually heard.
Everyone else
LLM reads the transcript
“The agent said the right words.” Misses how it sounded — the mispronounced drug name, the flat apology, the rushed close.
Audio models hear the call
Pronunciation, accent, emotion and vocal stress measured from the waveform — the signal an LLM grading text can never see.
Audio-native
custom models
- Emotion
- Vocal stress
- Accent clarity
- Pronunciation
- Pace & pauses
- Interruptions
Conversational
LLM + rules
- Resolution
- Escalation
- Repetition
- Hallucination
- Task success
- Tone
Compliance
policy
- Disclosures
- Identity check
- PII exposure
- Script adherence
Performance
latency
- Time-to-first-word
- Turn latency
- ASR WER
- Barge-in handling
§05 · Get started
First call scored in under a minute.
One click on any platform below and production calls stream in on their own — or send any recording with three lines of code.
Read the quickstartimport Roark from '@roarkhq/sdk'const roark = new Roark({ apiKey })await roark.calls.evaluate({recordingUrl, agent: 'support_v2',}) // scored in seconds
Works with
Also built for
Finance
Verify before you disclose, never expose an account number, and read the disclosure your regulator wrote.
ExploreInsurance
Quote the coverage right, read every required disclosure, and never sound cold to a claimant in crisis.
ExploreRetail
Quote the policy as written, never invent stock, and hold your brand voice when a returns caller is furious.
Explore
HIPAABAA availablePII handling & data residency
Configurable redaction and retention, SSO, and audit logs — with PCI and TCPA script checks scored on every call.
Bring a recording.
We’ll score it live.
See your own agent measured on the audio it actually produced — in the demo, in real time. Stop guessing whether your voice AI works.
founders@roark.ai · we reply fast