We place real test calls into your inbound campaigns, score every script step, and verify the message reached the right on-call. Don't say your agents follow the script. Prove it, every call path, every shift.
Works with Amtelco and Startel platforms, and with any AI or human answering.
For seven days we run Canary Calls against one inbound campaign. You receive a scored audit report with every finding evidenced by timestamps, audio, and transcript. No integration, no cost, no commitment.
Day 0
You send the campaign number, the script, and the escalation rules. We confirm authorization and agree the scenario set.
Days 1–7
Scheduled and randomized Canary Calls cover the happy path and the failure paths: emergency escalation, after-hours rollover, on-call not answering, caller hang-up.
Day 8
We walk through the scored audit report and every finding with you.
Test calls use fictional callers and fictional details. Dispatch goes to a CanaryTel-owned test on-call number.
Request the free auditInternal QA samples calls after the fact. It does not tell you whether the emergency path works tonight.
A client updates its escalation rules in March. By June, half your agents still use the old ones. Nobody notices until a caller does.
Supervisors listen to a handful of daytime calls. The 2 AM shift, where the urgent calls land, is the least reviewed.
A prompt edit or model update changes how your voice agent handles an emergency. The dashboard stays green because nobody tests the failure paths.
A single missed on-call page is the call the client remembers, and the one they cite when they cancel.
Anyone who has to prove, or verify, how calls are handled when nobody is watching.
Prove script adherence to clients and catch drift across accounts.
Verify the vendor you pay follows your escalation rules.
Regression-test failure paths after every prompt or model change.
Audit in-house after-hours teams the same way.
Evidence that after-hours escalation works.
Independent proof that the hotline works from first ring to closed report.
Independent secret-shopper surveys of after-hours and emergency lines.
An independent quality baseline during due diligence.
Consistent call handling across locations and vendors.
CanaryTel is independent of the answering service, the AI vendor, and the client, so one test produces one set of evidence every party can trust.
Synthetic monitoring for the phone line. A real call goes in, a timed, evidenced result comes out.
We turn each inbound campaign script into scenarios with expected steps: the happy path and the failure paths, such as emergency escalation, wrong number, caller hangs up mid-script, and on-call not answering.
Canary Calls run on a schedule you approve and at random times, including overnight and holidays. Callers are fictional and use fictional details only.
Each call is recorded, transcribed, and scored against the script step by step. Each test account's on-call roster points to a CanaryTel-owned test number or inbox, so we capture and time the dispatch ourselves. CanaryTel never contacts your real on-call staff.
Every deviation becomes a finding: what, when, call ID, expected versus actual, with audio and transcript attached. Monthly audit reports are built to hand to your clients and auditors.
Start with Canary Calls on a few campaigns. Add Canary Watch when you want your real calls scored. Any plan that handles patient health information includes a BAA. Pricing is scoped to your account volume.
Start with a free one-week audit on one campaign.
Tier 1
End-to-end campaign testing. The secret shopper for your answering service, automated.
Tier 2 · Early access
Always-on monitoring of the calls you actually handle, from the sources your platform exposes.
Enterprise
For operators and vendors with large account portfolios and formal compliance needs.
Medical comes first. The same method applies wherever a call has an escalation path.
On-call physician escalation, urgent versus routine triage, HIPAA-sensitive message handling. A missed on-call page is a patient safety event.
Failure scenario: chest pain at 2:14 AM
Caller reports chest pain. The script requires an urgent classification, a 911 advisory, and a page to Dr. Patel within 3 minutes.
Actual: The agent takes it as a routine message. The page goes out at 6:40 AM.
Conflict-sensitive intake, accurate capture of lead details, speed to callback.
Failure scenario: opposing party named, number mistyped
Caller names the opposing party in a custody matter. The script requires the conflict question, a read-back of the callback number, and a handoff to the on-call attorney.
Actual: The agent skips the conflict question and transposes two digits. The lead sits unreachable for three hours.
Correct emergency classification (no heat, gas leak, flood), the right technician paged, on-call rotation honored.
Failure scenario: gas smell reported at 11:48 PM
Tenant reports a gas smell. The script requires an emergency classification, a gas utility advisory, and a page to the on-call technician, Rivera this week.
Actual: The agent logs a maintenance request for 8 AM. Nobody is paged.
CanaryTel sits outside your platform. We call in like any caller and, for Canary Watch, read the records your platform exposes, by webhook, API, or scheduled export. Nothing to replace.
Synthetic monitoring and SLOs, applied to the phone line. Define what good looks like for each client account, then watch the numbers burn down or hold.
Time-to-dispatch on urgent and emergency calls, measured from the caller's first word to the page landing.
The share of a campaign's call paths exercised in a period, so untested branches are visible instead of assumed.
The percentage of required script steps performed correctly, per call, per agent, per campaign, per shift.
We are not part of your platform, your agents, or your AI vendor. We have no stake in the result, and we do not train or staff your agents. Scenarios are agreed with you in advance. Scoring is applied the same way to every call, and each score links to audio and a transcript so you can check it yourself.
Test calls use fictional callers and fictional patient details only. We never use real caller or patient information.
During onboarding we agree on dedicated caller IDs you can allowlist and a call-record tag, so test calls can be excluded from agent statistics and from client billing. Agents are not told which call is a test, because that would defeat the point.
CanaryTel never contacts real on-call staff. Each test account's on-call roster points to a CanaryTel-owned test number or inbox, so we capture and time the dispatch ourselves.
CanaryTel is built for HIPAA-sensitive workflows. Canary Calls use fictional patient details only. Canary Watch reads real call records, so it is scoped with you in advance and access is limited to what scoring needs.
Any plan that handles patient health information includes a BAA, and we flow the same obligations down to every provider that touches the data. We do not claim any certification we do not hold.
For a pilot on one or two campaigns:
You do not need to install anything on your platform to begin with Canary Calls.
Tell us about the campaign. We reply within one business day with a proposed scenario set.
The free one-week audit covers one inbound campaign, with no integration, no cost, and no commitment. You need to own the line or have the operator's agreement to test it. Test calls use fictional callers and fictional details.
A pilot starts with the scripts and escalation rules for one or two campaigns. We build the scenarios, you approve them, and the first Canary Calls run on the schedule you set.
All contact goes through this form. There is no public inbox, and we do not publish staff addresses.