Score every advice call or sale. See the evidence behind each verdict.

CallGuard AI scores protection and mortgage advice calls, one by one or as a sale, against your own compliance scorecard. Each verdict shows the transcript evidence it was decided on, or tells you none was found, so you can review every call or sale instead of a sample.

Request a demo See how it works
Protection advice call
18:42 · Protection sales scorecard
1 breach
  1. Pass Explained what the policy does not cover 09:12
  2. Pass Checked the premium is affordable for the customer 11:48
  3. Review Responded to possible signs of vulnerability 06:03
  4. Fail Asked for consent before submitting the application 14:31
    1. 14:24Customer:Yes, it's [NAME_GIVEN_1], that's all correct.
    2. 14:31Adviser:Lovely, I'll submit that to the insurer for you now.
    3. 14:36Customer:Oh. Okay.

    The application was submitted without the customer being asked to agree to it.

Illustrative example with synthetic call data.

A 5% sample leaves 95% of calls unreviewed.

Reviewing calls by hand means choosing which few to listen to. Whatever the sample misses is never reviewed: the disclosure that was skipped, the customer who sounded unsure, the pressure to decide today. CallGuard AI scores every call or sale against your scorecard, so none of it depends on which calls were picked.

95%

of calls are never reviewed when a team samples 5% by hand

From the call to the evidence.

CallGuard AI sits alongside the dialler and CRM you already use. Nothing about how your advisers work has to change.

  1. Calls come in

    Calls arrive automatically from your dialler, or upload recordings and send calls through the API.

  2. Checked against your scorecard

    Each call is transcribed and redacted, then scored against your scorecard, alone or as part of a sale. Each criterion gets a verdict with its evidence or a note that none was found, or goes to your review queue.

  3. Evidence you can act on

    Breaches alert your team by email, Slack or webhook, and results can be written back to your CRM.

Works with any dialler, or without one if you upload recordings, and with any CRM that has a suitable API, which we connect during setup. See all integrations.

Scored as a sale, not just a call.

A protection sale is rarely one conversation. When the sale is recorded in your CRM, CallGuard AI takes the calls from that customer's phone number within the sale window and scores them together, so what was said on the first call is in front of the AI alongside the call where the application goes in. Every plan

Sale · Level term life cover
3 calls with one customer
  1. 1
    Fact find
    22 min
  2. 2
    Advice
    31 min
  3. 3
    Application
    12 min
14 of 15 criteria met across the sale 1 breach
Illustrative example with synthetic data.

Checked against the application.

Once Reconciliation is switched on for your firm and the insurer's application PDF is attached to the sale in your CRM, it compares each answer with what the customer said on the calls. A mismatch is flagged for your team to raise with the insurer, and later amendments are classified so a routine fix isn't treated like a real discrepancy.

Get the protection consent-gate checklist

More on how Reconciliation works

Reconciliation
Health and lifestyle questions
1 does not match
Have you smoked in the last 12 months?
No
Call 1 · 16:05
Customer:I only stopped smoking in March, to be honest.
Does not match
Have you been treated for a heart condition?
No
Call 1 · 17:48
Customer:No, nothing like that.
Matches
What is your occupation?
Office manager
Call 2 · 04:12
Customer:I'm an office manager at a builders' merchant.
Matches
Illustrative example with synthetic data.

Breaches flagged while the call is live.

For calls streamed from Twilio, Amazon Connect or our WebSocket API, CallGuard AI flags breaches it is confident have already happened and sends them to your own systems over the stream or a webhook signed when you set a secret, re-checking the transcript every 30 seconds. Pro plan

When a streamed call ends, its full transcript is scored against your scorecard like a recorded call. Consent questions on streamed calls always go to your review queue, because a live stream can't confirm who said what. How live streaming connects

Live call in progress
Streamed from Twilio
Breach
Avoided pressuring the customer to decide · 07:42
Adviser:It's a great price, but you'd really need to decide today.
Flagged over the live streamSent
Webhook to your CRMDelivered
Illustrative example with synthetic call data.

Scored to your firm's standard.

Every firm writes its own scorecard for what a good advice call looks like. When your compliance team corrects a verdict, CallGuard AI uses your most recent corrections as examples when it scores later calls on that criterion. Growth and Pro

Unclear consent goes to a person
Where it is not reliably clear which speaker on a call is the customer, consent items go to your review queue and a person decides.
Your best calls as the benchmark
Mark a call or sale that shows good practice, and the AI is shown the first 1,500 characters of the two most recently marked exemplars as what good looks like at your firm.
Coaching that remembers
On calls scored individually, each coaching draft is written with that adviser's last three coaching drafts in view, and an insights brief, generated on request over a period you choose, summarises patterns for your compliance team.

More on how scoring calibration works

Score correction
Explained what the policy does not cover
AI verdictFail
Corrected by your compliance teamPass
Reason
The adviser read out the exclusions from the policy summary. We count that as covered.
Later calls on this criterionUsed as a calibration example
Illustrative example with synthetic data.

Personal data is redacted before scoring.

Names, payment details and health information are replaced with tags in the transcript before it is stored or scored. Call audio is kept, encrypted, and our transcription provider is named on the sub-processor list. Every redaction category is on by default. Your firm can ask us to turn some off, backed by a DPIA, but payment details can never be turned off.

Stored in the UK
Customer data is stored in a UK region. Audio and transcripts are also processed by the providers named on our sub-processor list, some of them outside the UK.
Encrypted and scoped
Audio is encrypted at rest with AES-256-GCM, and advisers can only see their own calls.
Published, not on request
Read our Data Processing Agreement and privacy policy before you talk to us.
Transcript as stored
Personal details redacted at transcription
  1. 02:10Adviser:Can I take your full name, please?
  2. 02:13Customer:It's [NAME_GIVEN_1] [NAME_FAMILY_1].
  3. 02:20Adviser:Thank you. And the best number to reach you on?
  4. 02:24Customer:It's [PHONE_NUMBER_1].
Illustrative example with synthetic call data.

Three plans, one scoring engine.

Starter, Growth and Pro, quoted per seat for your team size, with volume discounts and no multi-year minimum.

  • Starter

    Your own scorecard, per-criterion verdicts with the evidence behind them, sale scoring and email breach alerts.

  • Growth

    Adds scoring calibration from corrections, example calls, coaching memory, Slack alerts and CRM write-back.

  • Pro

    Adds live mid-call breach detection, Reconciliation and priority support.

Compare plans

Every plan includes

  • Transcription with the adviser and customer separated
  • Personal data redacted from transcripts before scoring
  • A verdict on each criterion, with the evidence behind it or a note that none was found
  • Coaching drafts for advisers
  • No separate per-minute transcription bill

Made for protection and mortgage advice.

Protection advice

Your scorecard can check ICOBS suitability and demands and needs, consent before an application is submitted, and vulnerability and consumer understanding. Reconciliation, where enabled, checks disclosure answers against the insurer application.

Call monitoring for advice firms Protection sold by phone

Mortgage advice

Our MCOB template checks scope of service and fee disclosure, that the recommendation links to stated needs, and the end-of-rate and payment-shock discussion, including uploaded Teams and Zoom recordings.

MCOB mortgage call monitoring The MCOB mortgage advice scorecard template

The same scoring also runs for contact centres and BPOs, debt collections and outbound sales.

Questions compliance teams ask.

Something else? Email hello@callguardai.co.uk.

Comparing vendors? See how CallGuard AI compares with Aveni and Recordsure.

How does CallGuard AI score a call?

Each call is transcribed with the adviser and customer separated, and personal details in the transcript are replaced with tags. The AI gives each criterion on your scorecard its own verdict, with the transcript evidence it was decided on, or a note that none was found. Your corrections and example calls shape how future calls are scored. More on how AI call QA works.

Does it support Consumer Duty, ICOBS and MCOB?

CallGuard AI scores against your firm's own scorecard, so it checks what your compliance team sets: ICOBS 5 demands and needs (and suitability where advice is given) for protection, MCOB 4.4A initial disclosure and MCOB 4.7A suitability for mortgages, the vulnerability characteristics described in FCA guidance FG21/1, and call-level evidence for the consumer understanding and consumer support outcomes (PRIN 2A.5 and 2A.6). Each verdict shows the transcript evidence it was decided on, or a note that none was found.

Can it score a sale made over several calls?

Yes, on every plan. When a sale is recorded in your CRM, CallGuard AI takes the calls from that customer's phone number within the sale window and scores them together as one sale, so what was said on an earlier call is in front of the AI alongside the call where the application was submitted.

How is customer data protected?

Names, payment details and health information are replaced with tags in the transcript before it is stored or scored. Call audio is kept, encrypted at rest with AES-256-GCM. Customer data is stored in the UK, and advisers can only see their own calls. The providers that process audio and transcripts, some of them outside the UK, are named on our sub-processor list, and our Data Processing Agreement is published.

Which diallers and systems does it work with?

Calls come in from your dialler, whichever it is: through our CloudTalk connector, SFTP pickup of your dialler's recording export, recording upload, or the REST API. Results are written back to your CRM, which we connect for you during setup. On the Pro plan, Twilio Media Streams calls, and Amazon Connect calls through a small Lambda bridge, can stream in live.

How do we get started?

Share your existing scorecard and we set up your account. We put a DPA in place before you share any real recordings, and connecting your dialler and your CRM can follow once you are happy with the results; sale scoring, and Reconciliation where it is switched on, start once your CRM is connected. There is no 90-day pilot and no integration project to begin.

How are plans priced?

Three plans, Starter, Growth and Pro, quoted per seat for your team size, with volume discounts and no multi-year minimum. What each plan includes is on the plans page.

See calls scored against a scorecard like yours.

In a short demo we score synthetic calls against a scorecard like yours. When you want to try your own recordings, we'll put a DPA in place before you share any.

Request a demo Or email hello@callguardai.co.uk

We usually reply the same working day. Read how we use your details in our privacy policy.