PRODUCT

Everything Ringfence checks, and how it checks it.

Ringfence sits between your call source and your review team. It transcribes, checks against your rules, cross-references live data where needed, and produces one audit trail per call — regardless of whether the call was handled by an AI agent or a person.

TRY IT

Paste a line an agent might say.

This is a simplified version of the same rule-check Ringfence runs on live calls — it looks for unauthorized-claim language against a sample rule set. Try something like "this plan comes with a lifetime guarantee" or "your appointment is confirmed for Tuesday."

DETECTION

Six rule categories, applied to every call.

Unauthorized promises

Matches spoken claims against your approved offer sheet — discounts, refund terms, "lifetime" language, guarantees. Flags anything not on the approved list, and separately flags anything that contradicts it.

Factual hallucinations

Cross-checks specific, checkable claims — a price, a coverage limit, an account balance, an appointment slot — against your CRM or product data over API, not against the model's training data.

Script deviation

Confirms required disclosures, disclaimers, and opt-out language were actually said, in the required form, at the required point in the call.

Pressure and tone

Flags patterns linked to high-pressure tactics: repeated urgency framing, interrupting a stated objection, continuing after a customer says no.

Consent and PII

Confirms required consent language preceded recording, card capture, or personal data collection — flags calls where collection started before consent was given.

Cross-agent consistency

Compares flagged claims across your different agent configurations or reps, surfacing when the same question gets materially different answers.

REVIEW QUEUE

Flags land in a queue built for compliance work, not engineering work.

Every flag shows the exact quote, the timestamp, an audio clip of just that segment, the rule it broke, and the severity Ringfence assigned based on how far the statement is from the approved language. Reviewers clear, dispute, or escalate — disputes route to a second reviewer automatically, escalations can trigger a Slack or email alert to a supervisor within the call window if you want live intervention.

Reviewers typically clear a flagged segment in under 40 seconds once they're used to the queue — most of that time is listening to the 10-15 second clip, not reading a transcript.

AUDIT TRAIL

One exportable record per call, kept as long as you need it.

Every call — flagged or clear — gets a permanent record: full transcript, audio, every rule checked and its result, reviewer decisions and timestamps, and the version of your rule set that was active at call time. Export as PDF for a single dispute, or as a bulk CSV/JSON feed for a compliance audit covering thousands of calls.

98.7%
of flagged segments reviewed within 24 hours across current customers
SETUP

What onboarding actually involves.

01

Connect a call source

API key or webhook for Vapi, Retell, or Synthflow — typically live same day. Dialer recording feeds (Twilio, RingCentral, and most on-prem systems) take slightly longer to route.

02

Upload your script and terms

Your approved script, offer sheet, and any regulatory disclosure requirements. We turn these into a structured rule set with your team over 2-3 working sessions.

03

Shadow mode

Run monitoring silently for 5-7 days against real call volume to tune thresholds and cut false positives before flags start reaching your review team.

04

Go live

Flags start routing to your review queue. Most teams see their false-positive rate drop by roughly half within the first three weeks as the rule set gets tuned against real calls.

See it on your own call data.

Start a trial and connect one call source — most teams see their first flagged call within a day.