Advice calls · Risk warnings
DashboardWealth and adviceCatch missing risk warnings on recorded advice calls within 24 hours
Sampled call reviews find a missing risk warning weeks after the client acted on the advice. Here is a setup that checks every recorded advice call within a day and puts the quote in front of a compliance reviewer.
Valnora & Elric team12 min read
- Team
- Advice compliance and supervision
- Manual steps today
- 5
- Build steps
- 9
- Review points
- 5
The problem
Advice firms record calls, and their advice process names the risk warnings an adviser must give before recommending a product: that the value can fall, that past performance is not a guide, what the charges are. Recording rules such as the FCA’s SYSC 10A and the record-keeping requirements in MiFID II mean the evidence exists. The FCA’s Consumer Duty raises the bar on whether clients understood what they were told. What most firms lack is a way to check the evidence at the volume it arrives.
Today, compliance listens to a small sample, often weeks after the call. A missing warning on an unsampled call is never found, and one on a sampled call is found after the client has already invested. Listening in full is slow, so the sample cannot grow with call volume.
For example, a firm recording 6,000 advice calls a month and reviewing 3% of them leaves 5,820 calls a month unchecked.
How advice compliance and supervision works today
Today, done by hand · Advice compliance and supervision
- 01Recordings are stored by the telephony or recording platform.
- 02Each month, compliance picks a small sample of calls per adviser.
- 03A reviewer listens to each sampled call in full against a paper or spreadsheet checklist.
- 04Findings are written up and sent to the adviser’s supervisor.
- 05If a warning was missed, someone decides whether the client needs a follow-up.
Where it breaks
- Most calls are never checked, so a missing warning is found by chance.
- Findings arrive weeks after the call, when the client may already have invested.
- Listening in full is slow, so the sample cannot grow with call volume.
- Reviewers apply the checklist differently, and nobody can show trends by product or team.
What the change could be worth
A wealth and advice firm recording 6,000 advice calls a month and reviewing a 3% sample today.
| Assumption | Value |
|---|---|
| Recorded advice calls per month | 6,000 calls / month |
| Share of calls reviewed today | 3% |
| Share of calls missing a required risk warning | 2% |
| Days from call to sampled review today | 21 days |
| Minutes to review a sampled call by listening | 15 min |
| Share of calls flagged for review with the setup | 5% |
| Minutes to review a flagged call from the quote | 5 min |
| Loaded cost of a reviewer hour | 60 GBP |
Modelled from the assumptions above to show how the outcome is calculated; not a customer result. Your results depend on your volumes and process.
In this illustrative scenario, with 2% of calls missing a required warning, checking every call finds about 116 more misses a month, about 20 days sooner. Reviewers move from listening to sampled calls to reviewing flagged calls from the quote, which redirects around £14k of reviewer time a year while coverage rises to every call.
The outcome you want
Every recorded advice call is transcribed and checked for its risk warnings within 24 hours. A missing critical warning raises an alert, a reviewer gets a drafted note with the quote, and a person decides what happens next, including whether the client needs a follow-up.
Acceptance criteria
- Every recorded advice call is transcribed and checked within 24 hours.
- Every failed check carries the verbatim quote it failed on.
- Any missing critical risk warning raises an alert at the next hourly check.
- A compliance reviewer records a decision on each flagged call within one business day.
- No client or adviser is contacted by the system; people own every follow-up.
The solution end to end
Call compliance in Agent Serve is not a separate product. It is a Call Compliance dashboard built with Dashboards, plus workflows, rules, monitors and a Relay around it.
Workflow 1, recordings to transcripts. The dashboard accepts audio uploads too, but new recordings arrive every hour, so the workflow builder automates the step. An hourly Schedule block picks up new recordings through an API block, a Speech-to-Text block transcribes each one, and a Table block inserts the transcript into the conversations table behind the dashboard. With automatic analysis of new rows on, each call is checked as it lands. Every run is traced in Logs.
The Call Compliance dashboard. Created from the Call Compliance template, it has a Transcripts panel where compliance writes checks in plain English, starting from the Financial services sales calls preset. Each check has a severity, and each call gets pass, fail or not applicable with a verbatim quote. Analytics fields such as summary, topics and next steps add context.
Rules, metrics and monitors. The template ships with a critical-breach rule, a breach-rate metric and two monitors: any critical breach, and breach rate above 5% by default. Add a Flag rule for missing risk warnings and a Rule match monitor on it, checked hourly. Monitor alerts show in the app and by email, and a dashboard KPI alert can post a falling pass rate to Slack.
The Relay. The Compliance Reviewer Relay template starts when the monitor alerts. It queries the conversations table and searches your advice procedures in the Knowledge Base, then drafts a note: what failed, the quote, the product and the adviser’s recent history. The note arrives in the compliance lead’s chat. It runs in Observe mode, so it can read and draft but cannot send email, and its instructions say not to contact customers.
Workflow 2, decision and remediation. Each morning it picks up flagged calls without a decision and pauses at a Human block. The reviewer listens back and submits a decision in the Resume Form. A Table block writes it to a remediation log, and a Microsoft Teams block tells the supervisor about confirmed misses.
Chat. Chat can query the conversations and remediation tables for ad-hoc questions, such as which products had the most missed warnings this month.
From recorded advice call to a reviewed finding
Workflow 1 · Recordings to transcripts
- 01Hourly scheduleA Schedule block picks up advice calls recorded since the last run.
- 02Recording platformAn API block fetches each recording with adviser, client reference and call date.
- 03Speech-to-Text blockTurns each new recording into a transcript automatically, so nobody has to upload files.
- 04Conversations tableA Table block inserts the transcript; new rows are analysed automatically.
Dashboard · Call Compliance
- 05Transcripts panelRuns your plain-English checks: pass, fail or not applicable, with the verbatim quote.
- 06Missing risk warning ruleA Flag rule marks any call that failed a risk warning check.
- 07MonitorsAlert on any missing risk warning and on a breach rate above your threshold.
Relay · Compliance Reviewer
- 08Starts on the alertThe Compliance Reviewer template starts when the monitor alerts, in Observe mode.
- 09Review noteWhat failed, the quote, the product discussed and the adviser’s recent history.
Workflow 2 · Decision and remediation
- 10Compliance decisionA Human block pauses until a reviewer submits a decision in the Resume Form.
- 11Remediation logA Table block records the decision, owner and due date for any client follow-up.
- 12Supervisor follow-upA Microsoft Teams block tells the adviser’s supervisor; people contact the client.
Build it
Write and test the checks before connecting live recordings: the checks are the part compliance has to trust.
Configure it step by step
- 01
Create the Call Compliance dashboard
Dashboards → Start from a template → Call Compliance
- Dashboard name: Advice calls · risk warnings.
- Data → Table: a new conversations table; map the role columns for adviser, team, product and call date.
- 02
Write the risk warning checks
Transcripts panel → Configure checks
- 1 Checklist: start from the Financial services sales calls preset and edit it to your advice process.
- Example: “Before recommending an investment, the adviser says its value can fall as well as rise and the client may get back less than they put in.” Severity: critical.
- Add checks for past performance, charges and capacity for loss, each with a severity.
- 2 Analytics: keep summary, topics and next steps. 3 Test on past calls compliance has reviewed. 4 Apply, then Apply and run.
- 03
Upload procedures to a Knowledge Base
Knowledge → New knowledge base
- Name: Advice procedures. Upload Documents with approved warning wording and supervision procedures, then Create.
- The Relay searches it when it drafts review notes.
- 04
Build Workflow 1, recordings to transcripts
Workflow builder · Schedule, API, Speech-to-Text, Table blocks
- Schedule block: Run frequency Hourly. API block fetches recordings made since the last run, with metadata.
- Loop over recordings; Speech-to-Text block transcribes each one.
- Table block Insert Row into the conversations table, with automatic analysis of new rows on. Deploy.
- For a small pilot you can instead save transcripts to a Library folder and use Upload transcripts → Library folder → Import N.
- 05
Add the missing risk warning rule
Rules → New rule
- Rule type: Flag (true / false). Name: Missing risk warning.
- Condition: the result of any risk warning check is fail. Create rule. Keep the template’s critical-breach rule as well.
- 06
Set up the monitors and KPI alert
Monitors → New monitor
- Rule match on Missing risk warning, threshold 1 row, Check Hourly, Severity Critical. Create monitor.
- Keep the template’s breach rate monitor and set its Threshold to your own figure; 5% is the default.
- On the dashboard, Share → New KPI alert… on risk warning pass rate, Drops below your target, Notify Slack.
- 07
Set up the Compliance Reviewer Relay
Relays → New Relay
- Specialization: Compliance Reviewer. Create draft. Autonomy stays Observe.
- Other resources: Table (conversations) and Knowledge base (Advice procedures). Tools in conversations: Query tables, Search knowledge bases.
- Instructions: add the adviser’s recent failures and the product to each note. Keep “Do not contact customers”.
- New responsibility: When a monitor alerts, Monitor Missing risk warning. Publish.
- 08
Build Workflow 2, decision and remediation
Workflow builder · Schedule, Table, Human, Microsoft Teams blocks
- Schedule block Daily at 08:00; Table block Query Rows for flagged calls without a decision; Loop over them.
- Human block: Display Data shows the failed check and quote; Resume Form has a Decision field (Confirmed miss / Covered elsewhere / Check needs rewording).
- Table block Insert Row into a Remediation log table; Microsoft Teams block tells the supervisor about confirmed misses. Deploy.
- 09
Limit access and schedule the report
Dashboard Settings → Row security; Share → Subscribe…
- Add policy so each supervisor sees their own advice team; compliance sees all.
- Subscribe… Weekly PDF to the head of compliance, Mondays at 08:00.
What it produces
A short excerpt shows the pattern the checks are written to catch: a recommendation backed by past returns, with no warning that the value can fall.
What the solution produces
Transcript excerpt from the Speech-to-Text block (illustrative)
MemoAdviser: Based on what you have told me, I would recommend the global equity fund inside your ISA.
Client: And that grows faster than my savings account?
Adviser: Over the last five years it has returned about 8% a year, so you should do well.
Client: Great, let us go ahead.
Check results for the call (illustrative)
Record- Capital at risk warning
- Fail · critical. No statement that the value can fall and the client may get back less.
- Past performance
- Fail · medium. Quote: “it has returned about 8% a year, so you should do well.”
- Charges explained
- Pass · “the ongoing charge is 0.9% a year.”
- Capacity for loss
- Not applicable · discussed on the earlier fact-find call
- Summary
- Recommendation of an equity fund in an ISA; client agreed on the call.
Compliance Reviewer note (illustrative)
MemoCall A-77312 failed the critical capital at risk check. The adviser cited past returns without saying they may not continue.
Same adviser: one medium past-performance failure in the last 30 days, none critical.
Suggested next step: listen back, then decide whether the client needs a follow-up letter with the risk warning.
I have not contacted the client or the adviser.
Where people stay in charge
Where people stay in charge
- Gate 1Compliance writes and owns every check and its severity, and tests changes on past reviewed calls first.
- Gate 2A reviewer listens back and records a decision on every flagged call; a fail is a prompt, not a finding.
- Gate 3Supervisors decide coaching and whether the client needs a follow-up; the system contacts nobody.
- Gate 4Overturned fails are reviewed monthly to reword checks that misfire.
- Gate 5Changes to the workflows go live only through a new deployed version.
The dashboard
Advice calls · risk warnings
Latest failed risk warning checks
| Call | Product | Check | Severity | Status |
|---|---|---|---|---|
| A-77312 | Equity fund ISA | Capital at risk | Critical | Open |
| A-77298 | Pension transfer | Charges explained | High | In review |
| A-77251 | Bond portfolio | Past performance | Medium | Closed |
The dashboard is an illustrative example. Drill from team to adviser to see where coaching is needed, and compare pass rates with the previous period to check that changes to the checks did not move the numbers on their own.
Controls and audit
Workflow runs record each block’s inputs, outputs and errors in Logs, so you can show when a call was transcribed and checked. Relay activity is written to the audit log, which admins export from Activity audit, and reviewers’ decisions sit in the remediation log with their names. Workflows run from a deployed version.
Row-level access limits each supervisor to their own team on the dashboard, in exports and in scheduled reports. People sign in through SSO and are provisioned with SCIM. Recordings hold personal data, so set retention to your policy, and run Agent Serve self-hosted if audio must stay in your environment. Zero data retention is available with supported model providers through your own keys or self-hosted models. Some provider features keep data on the provider's side by design, such as Gemini Deep Research, so they are outside zero data retention.
Controls and governance
Audit trail
Proposals, approvals, declines and executed actions are written to the workspace audit log, which admins can export.
Run logs
Every workflow run records each block’s inputs, outputs and errors, so a reviewer can trace how a result was reached.
Row-level access
Row-level policies narrow what each person sees in tables, dashboards, exports and scheduled reports.
Approvals before action
Relays never send email or book meetings without an approval, under any autonomy setting. An approval covers the exact payload the reviewer saw and expires if nobody decides.
PII detection
A Guardrails block detects personal data and blocks or masks it before text reaches the next step.
Deployed versions
Workflows run from a deployed version, so a change is reviewed before it reaches production runs.
Self-hosting
Run Agent Serve in your own cloud or data centre, next to the systems it reads.
Zero data retention
Zero data retention with supported model providers, backed by your own provider keys and agreements or self-hosted models. Some provider features keep data by design, such as Gemini Deep Research, so they are outside zero data retention.
More on how Agent Serve handles access, data and audit on the security page.
Roll it out
Start by running the checks on calls compliance has already reviewed, then go live for one advice team. Put calls checked, hours from call to check, risk warning pass rate and open misses on the dashboard from the first week. The timings are a plan, not a promise.
Pilot, measure, expand
- Weeks 1–2
Test the checks
- Create the dashboard and checks, and run them on past calls compliance has already reviewed.
- Reword checks until results match reviewer findings.
Done when: Checks agree with reviewer findings on the test set, with differences explained.
- Weeks 3–4
Live for one advice team
- Deploy Workflow 1 for one team and turn on the monitors and the Relay.
- Track hours from call to check and pass rate on the dashboard.
Done when: Every call from the pilot team is checked within 24 hours for two weeks.
- Weeks 5–8
All advice teams
- Add the remaining teams and Workflow 2 for decisions.
- Keep a small random listening sample to test the checks.
Done when: Flagged calls get a decision within one business day across teams.
| Risk | How to handle it |
|---|---|
| Transcription errors cause false fails or false passes. | Reviewers listen back before deciding, and a random sample keeps testing the checks. |
| Checks misread warnings given in different words. | Write checks around meaning, test them on past calls and track overturns. |
| Too many alerts for reviewers to handle. | Alert on critical misses only and leave lower severities to the daily workflow and dashboard. |
| Call audio and transcripts hold personal data. | Use row-level access, set retention to your policy and use your own keys or self-hosted models. |
Timings are an illustrative plan, not a commitment; your pilot sets the pace.
Before you start
What you need
- API access to your call recording platform, or a scheduled export of recordings
- The risk warnings your advice process requires, per product type, in plain English
- A severity for each check, agreed with compliance
- A sample of past advice calls already reviewed by compliance, to test the checks
- Your advice procedures and approved warning wording, uploaded to a Knowledge Base
- Named compliance reviewers and the supervisors who own adviser follow-up
- Your retention period for recordings, transcripts and check results
Questions teams ask
Questions teams ask
Is call compliance a separate Agent Serve product?
No. It is a Call Compliance dashboard built with Dashboards, from the Call Compliance template. The Transcripts panel, checks, rule, metric and monitors are all ordinary dashboard parts.
Can we upload audio files directly?
Yes. The Transcripts panel accepts call audio such as mp3, m4a, wav, ogg, flac, aac or video files, with a choice of transcription model and optional speaker labels, as well as txt, vtt, srt, json, md, docx, pdf and csv transcripts. Workflow 1 automates the same step for recordings that arrive every hour.
Who decides whether a missed warning was a breach?
A compliance reviewer. The check gives pass, fail or not applicable with a quote, and the Relay drafts a note, but the reviewer listens back and records the decision in the Resume Form of Workflow 2.
What if the warning was given on an earlier call or in writing?
Write the check to allow for it, and use not applicable where your process covers the warning elsewhere. Reviewers can also overturn a fail; overturns are the best signal for rewording a check.
Can the Relay contact the client or the adviser?
No. The Compliance Reviewer template runs in Observe mode, where sending email and booking meetings are blocked, and its instructions say not to contact customers. Supervisors and advisers handle any client follow-up.
Does checking every call replace listening to calls?
It changes which calls people listen to. Reviewers start from failed checks with the quote, and many firms keep a small random sample to test that the checks themselves are right.
Related: disclosure and consent checks and full-coverage call QA. Request access