Call compliance monitoring for FCA-regulated advice firms.
CallGuard AI is FCA call monitoring software for mortgage brokers and protection advisers. Score every advice call against your own scorecard, with checks such as ICOBS 5 demands and needs (and suitability where advice is given), FG21/1 vulnerable customer indicators, and how charges are explained. Evidence on every scored call, not a sample.
Sampling leaves most calls unreviewed.
Most financial planning compliance teams still review a small manual sample of calls. Each one takes a senior reviewer's time. The rest of the advice conversations go unreviewed. Consumer Duty requires firms to monitor the outcomes their customers receive (PRIN 2A.9), and a small sample shows only a small part of them.
Sampled file checking
- A small fraction of calls actually reviewed
- Variance between checkers is meaningful
- Findings land weeks after the recommendation
- Vulnerability indicators slip past unflagged
- Outcomes evidenced anecdotally, not systematically
- QA capacity caps how fast you can hire advisers
CallGuard AI on every advice call
- Every call transcribed automatically, ready to score one by one or as part of a sale
- Your team's corrections shown to the AI as examples
- Critical breaches logged in a breach register, with alerts
- Vulnerability indicators screened on every conversation
- Outcomes evidenced systematically with hard numbers
- Adviser headcount can grow without QA bottlenecks
Your scorecard, built around what a call can evidence.
CallGuard AI scores against the scorecard you upload. For protection advisers that typically includes ICOBS 5 demands and needs (and suitability where advice is given); for mortgage advisers, the MCOB checks your scorecard sets; and for every firm, FG21/1 vulnerable customer indicators, the Consumer Duty outcomes visible on a call, and your firm's own conduct rules. Each verdict shows the transcript evidence it was decided on, or a note that none was found. Protection sold by phone has its own page: protection telesales.
Demands, needs and suitability (ICOBS 5)
Demands and needs established before any recommendation. Existing cover, eligibility and affordability checked. The recommendation explained against the needs the customer actually stated, not the ones implied for them. Pass or fail per criterion, with the transcript evidence where the call has it. For a protection sale, Reconciliation then checks those same disclosures against the answers submitted on the insurer's application, on the Pro plan.
Vulnerable client indicators (FG21/1)
Health, life event, resilience and capability vulnerability signals screened on every call. Flagged for adviser awareness and surfaced in your breach register if the call did not adapt to the disclosed vulnerability.
Product features explained
Were the product's main features, exclusions and limits explained in a way the client understood? Your scorecard sets what has to be covered for each product line, and CallGuard AI scores the call against it.
Charges explained
Fair value is assessed at product level, not on a call; for non-investment insurance that sits in PROD 4 rather than PRIN 2A.4. What the call can evidence is how the charges were explained: charges disclosed clearly, and any ongoing fee explained against the ongoing service the client receives.
Consumer Duty: consumer understanding
Did the adviser check, where appropriate, that the client understood the recommendation? Was technical language adapted to the disclosed financial knowledge level? Confirmation captured in the transcript, not assumed.
Consumer Duty: consumer support
Checks your scorecard can set: cancellation rights explained in plain English, next steps agreed, contact details given for questions or complaints, and specialist referrals made where required.
An AI that's shown your compliance reviewer's own corrections, not a generic baseline.
Generic AI scoring reads only the scorecard's words. CallGuard AI is also shown your team's own corrections and the exemplar calls you mark, so it sees how your firm has read "suitable", "clearly explained" and "good outcome" before, not only the scorecard's wording. When your compliance officer corrects a verdict, up to five of the most recent corrections on that criterion are shown to the AI as examples the next time it scores that criterion. They are examples for the AI to weigh, not rules it must follow. More on how scoring calibration works.
Compliance officer corrections feed back
When your compliance officer disagrees with an AI verdict, they choose Pass or Fail, add a reason if they want to, and save. The correction — with the transcript excerpt it was based on and any reason given — becomes one of the examples shown to the AI the next time it scores that criterion.
Exemplar advice calls
Mark gold-standard advice calls or sales as exemplars. The AI is shown the first 1,500 characters of the two most recently marked exemplars — for a sale, that's its closing call — as what good looks like at your firm. The library is per-firm, not shared across customers.
Per-adviser coaching memory
On calls scored individually, each coaching draft is written with that adviser's last three coaching drafts in view. Personalised coaching, not template feedback.
Strategic insights for the CO
An AI insights brief, generated on request, summarising what changed, the patterns worth attention, and 3 to 4 specific actions a compliance director will actually take. Written for the CO, not as boilerplate.
Common questions, answered.
What evidence does CallGuard produce for an FCA visit?
For each scored call you get the full transcript, the per-criterion verdicts with transcript evidence, the breach register entry (if applicable) with severity and resolution status, the AI's coaching draft, and any compliance corrections made by your CO, with any reason given. Exportable as a per-call PDF or via the API for your case-management system.
Can compliance officers override AI scores?
Yes. A compliance officer can flip any AI verdict: choose Pass or Fail, add a reason if they want to, and save. The corrected verdict becomes the official record, the correction is shown to the AI as a calibration example on future calls, and the original AI verdict is retained alongside for audit purposes.
How is client data secured?
All audio is encrypted at rest with AES-256-GCM. Transport is TLS 1.2+. Customer data is stored in the UK; some sub-processors that process audio and transcripts are outside the UK, and they are named on our sub-processor list. UK GDPR Article 28 DPA is published at /dpa and our sub-processor list at /sub-processors. We do not train any model on your data.
Does this work alongside our existing case management system?
Yes. Push call recordings into CallGuard via REST API, and pull the scoring result back via the same API to render inside your case-management UI. Integrators can also subscribe to webhooks, HMAC-signed once you set a signing secret, to be notified when live breach detection flags a breach on a streamed call (Pro plan); each alert carries the criterion's severity.
Further reading: what AI call QA is and how it actually works, AI vs human call scoring: when each makes sense, scoring ICOBS 5.2 demands and needs as a criterion and vulnerable customer call monitoring under FG21/1. To start from a ready-made scorecard, see our free compliance scorecard templates, including the 34-criterion MCOB mortgage advice scorecard. If you are weighing CallGuard AI against other call QA vendors, our comparison hub sets out where each one fits, including the Aveni alternative and Recordsure alternative pages for UK advice firms.