Compliance scoring for outbound sales and lead-gen teams.
CallGuard AI is outbound sales call compliance software for the channel where one bad call can land you in front of the Information Commission or Ofcom. Objections and TPS rules under PECR, caller identification, no-pressure-language rules and how charges are explained, scored on every call or sale you send us. Mid-call breach alerts on streamed calls.
Outbound is where compliance failures are visible to regulators and competitors.
Customers who receive a non-compliant outbound call complain to the Information Commission. Or they post about it on social media. Or they tip off a competitor who reports you. Inbound mistakes get absorbed quietly. Outbound mistakes are loud, public, and document themselves. The cost of one regulator-grade breach typically exceeds the QA budget that would have caught it.
Sample-based outbound QA
- 1 to 3% of calls reviewed manually
- Ignored objections to marketing discovered weeks later
- TPS list breaches absorbed before the audit catches them
- Pressure-language failures only surface in customer complaints
- Variance between reviewers is meaningful
- Pattern detection across the floor is anecdotal
CallGuard AI on every outbound call
- Every call transcribed automatically, ready to score one by one or as part of a sale
- Caller identification and TPS objections checked on each call scored
- TPS list compliance scored against your dial-time policy
- Pressure-language failures alert mid-call via webhook
- One AI shown your compliance team's corrections as calibration examples
- Patterns surfaced in AI insights briefs, not anecdotes
PECR, TPS, Consumer Duty and your firm's internal closing rules.
CallGuard AI scores against the scorecard you upload. For UK outbound sales operations, the scorecard typically encodes the regulatory regime (PECR objections and TPS rules, caller identification under regulation 24, Ofcom dialler conduct, and how charges are explained under Consumer Duty) plus your firm's internal closing rules and ethical-selling guidelines.
Caller identification and objections
Checks the agent gave their name, which PECR regulation 24 requires, and that an objection to marketing ended the pitch. Once a customer has objected, calling them again for marketing breaches PECR, so the objection needs to reach your suppression list.
TPS / CTPS list compliance
Telephone Preference Service checks scored against your dial-time policy. The dial itself happens upstream of CallGuard, but we score whether the agent acknowledged customer objections appropriately when the customer mentions they are TPS-registered or asks how their number was sourced.
No-pressure-language detection
Urgency tactics, scarcity language ("only available today"), manufactured authority ("I'm calling from your bank") and coercive close patterns flagged in the live transcript. A breach is only sent when the model rates it at 0.75 confidence or higher, which keeps false positives down.
Caller identification + recording disclosure
Agent gives the caller's name, as PECR regulation 24 requires, with an address or freephone number if the customer asks for one. Stating the purpose of the call early is good practice rather than a PECR rule, and is scored if your policy requires it. Recording disclosure given where required, which is a UK GDPR transparency duty rather than PECR. An objection to further marketing respected without more pitching.
Consumer Duty: charges explained
For FCA-regulated outbound (financial products and insurance), fair value is assessed by the product's manufacturer and evidenced in the file, and for non-investment insurance PRIN 2A.4 does not apply at all. What the call can show is the consumer-understanding side: charges stated clearly, and alternatives mentioned where relevant.
Vulnerable customer indicators
Health, life event, resilience and capability vulnerability signals screened on every call. Disclosed-but-unaccommodated indicators flagged as critical breaches even when the procedural compliance was fine.
Catch the urgency-language close before the customer hangs up.
For outbound, "we found this in next month's audit" is too late. The customer has already lodged a complaint with the Information Commission. CallGuard AI streams audio live from your dialler, runs scoring against the rolling transcript as the call happens, and emits high-confidence breach alerts via webhook, HMAC-signed once you set a signing secret, to your dialler, your CRM or a Slack channel.
Pressure-language alert
If the agent uses urgency tactics, false scarcity or coercive close language, a webhook fires to the floor manager. They can step in before the customer agrees to a sale they were pressured into.
Objection ignored alert
If the customer objects to marketing and the agent keeps pitching, a webhook fires for floor-manager intervention. The manager can step in before the call goes further and make sure the objection is recorded.
Vulnerability disclosure alert
If the customer discloses a vulnerability indicator and the agent continues at the same pace using the same close, a webhook fires. The standard expected by Consumer Duty.
Complaint trigger alert
Customer raises a complaint or asks how their number was sourced. Webhook fires to your complaints team's queue so they can pick up the call before the customer hangs up.
Common questions, answered.
Does CallGuard AI score TPS list compliance?
Indirectly. The TPS list check happens at dial-time in your dialler, before CallGuard AI sees the call. What we score is the conversation that follows: did the agent acknowledge objections appropriately, did they handle a customer mentioning TPS registration correctly, did they explain how the number was sourced if asked. The dial-time list check is upstream of us.
Will the AI flag every "act now" as urgency language?
No. Whether urgency language counts as a failure is set by your scorecard and written guidance: "This offer ends Friday" on a genuinely time-limited campaign can pass, while "decide now or lose this forever" on a non-urgent product can be defined as a fail. When your compliance officer corrects a verdict, the most recent corrections on that criterion are shown to the AI as examples the next time it scores it.
How do you handle field sales operations versus call-centre outbound?
The scoring engine is the same; the channel differs. Call-centre outbound uses dialler streaming. Field sales (doorstep, in-person) is on our roadmap pending the first design partner. The streaming protocol works today for any audio source; what's coming is the native iOS / Android capture layer. More on the roadmap here.
Can mid-call alerts be configured per agent or per campaign?
Yes. Different agents (new joiners, escalations) can have different alert thresholds. Different campaigns can have different scorecards. Floor managers can configure who gets which alert type so the right person sees the right breach in real time.
How does this integrate with Five9, Genesys or Twilio?
First-class adapter for Twilio Media Streams, and Amazon Connect live media streaming via a customer-run Lambda bridge. Other diallers can connect via our generic WebSocket protocol for live streaming. For any dialler, including Five9 and Genesys, you can also export recordings and bring them in via upload or our REST API.
Further reading: PECR and TPS rules for outbound calling teams, AI vs human call scoring: when each makes sense and how to score 100% of contact centre calls without hiring more QA staff. If you are comparing real-time tools, see our Balto alternative page.