Every call your team makes,
read and on the record.
Speech Rec turns recorded customer calls into compliance evidence and revenue intelligence — the risk phrases, intent signals and QA scores that matter, surfaced automatically.
Example: a brokerage call. The watchlist behind it is editable per client.
Catch the phrase before the regulator does
Every call is scanned against a severity-graded watchlist — guarantee, misrepresentation, pressure, suitability and evasion — deterministically, with no model in the loop. The phrases are yours: edit the list in the app, or add a tracker term for a product line or a competitor. Flags roll up per agent, escalate when they cross the line, and land in a tamper-evident audit trail.
- ✓ Deterministic phrase surveillance on a watchlist you edit
- ✓ QA scoring and one-click escalation to a reviewer
- ✓ Monday-morning compliance digest, by email and Slack
Same detector, your vocabulary
The scanner does not know your sector — it knows the phrases on your watchlist. One list ships by default, written for retail FX and CFD conduct. Entries are yours to add or remove in the app — each with a phrase, a severity and a category.
-
Brokerage & investment
Guaranteed returns, pressure to fund, unsuitable funding sources, off-record requests. Instruments and asset classes are recognised out of the box.
-
iGaming
Affordability wording, responsible-gambling phrasing, bonus and wagering terms, self-exclusion. Your own game verticals go on the tracker list.
-
Insurance
Cover and exclusion claims, renewal pressure, demands and needs, vulnerable-customer signals — alongside the policies you actually sell.
-
Wealth & advice
Risk-profile language, fee and charge disclosure, capacity for loss, vulnerability signals. Products and wrappers go on the tracker list.
Five conduct categories ship as standard — guarantee, misrepresentation, pressure, suitability and evasion — with a sixth, tracker, for anything else you want counted: a product line, a competitor, a campaign.
See the money on every call
Speech Rec reads intent and stance, the products and objections that came up, and whether a follow-up was actually booked — then correlates each call to your CRM contact so a manager knows exactly who to call back first.
- ✓ Intent scored 0–100, with the reasons it landed there
- ✓ Instrument recognition ships for trading desks; anything else is a tracker term
- ✓ Pipeline by stance, and the interested calls with no callback booked
Example: a trading desk. Anything else you want counted goes on the watchlist as a tracker term.
Who talked, who listened, who improved
When calls are diarised, Speech Rec measures talk ratio, longest monologue, dead air, interruptions, questions asked and speaking pace — and rolls it into an agent scorecard that trends week over week. No speaker labels, no fabricated numbers.
- ✓ Talk-time balance and monologue detection per call
- ✓ Agent scorecards with a weekly trend line
- ✓ Coaching flags a manager can act on in minutes
Your calls stay on your infrastructure
Transcription runs on your own machine with faster-whisper. The write-up — titles, summaries, action points — runs through your own Claude, not a third-party billing account. The app states, honestly, what leaves the network, and nothing about a recording has to.
- ✓ On-machine ASR — no audio upload required
- ✓ Role-based accounts, PII redaction, DSAR export
- ✓ Runs local, or on your own cloud — your choice
- ✓ Fully white-label — your name, logo and colour throughout
Three steps from a call to an answer
No rip-and-replace. Point your recordings at it and read the intelligence.
Record or import
Drop in a file, or wire up Twilio, Telnyx or your VoIP call-recording webhook. Bulk import is one queue away.
Transcribe on your machine
faster-whisper transcribes and diarises locally. Nothing is sent to a transcription vendor.
Read the intelligence
Compliance flags, sales signals, QA and coaching land on one dashboard — and in a morning digest.
Put a real call through it
Open the live demo, or sign in. It runs a real recorded call end to end — flags, intent and all. On your own instance the watchlist behind it is yours to edit.