Collections · Conduct checks
MixedConsumer lendingCollections compliance in consumer lending: choosing how to check calls and letters
Collections is where conduct rules meet customers under pressure. This guide compares four ways to check calls and letters, what each one catches, and how they combine in Agent Serve.
Valnora & Elric team7 min read
Compared in this guide
- AManual sampling
- BCall Compliance dashboard
- CPre-send letter check
- DCompliance Reviewer Relay
Collections is where conduct rules meet customers under pressure. In the US, Regulation F implements the Fair Debt Collection Practices Act and sets rules on communications, harassing or abusive conduct, misleading representations and validation notices, for debt collectors as the Act defines them. In the UK, chapter 7 of the FCA’s consumer credit sourcebook, CONC 7, covers how firms treat customers in arrears or default, contact with customers and customers with mental capacity limitations, and the Consumer Duty sits over all of it.
Whatever the market, the operational question is the same: how do you know your calls and letters meet the rules, and how fast do you find out when they do not? This guide compares four ways to check, and explains when each one fits.
The options side by side
Four ways to check collections communications
| Criterion | Manual sampling | Call Compliance dashboard | Pre-send letter check | Compliance Reviewer Relay |
|---|---|---|---|---|
| What it checks | A small sample of calls and letters | Every transcript against the Collections checklist and your own checks | Every outbound letter and email before it is sent | Each breach a monitor raises |
| When it catches a problem | Days or weeks later | After the call, as soon as the transcript is scored | Before the customer receives anything | Minutes after the breach alert |
| Evidence | Reviewer notes | Pass, fail or not applicable per check, with a verbatim quote | The flag, the rule it failed and the draft, in the run trace | A review note citing the quoted evidence |
| Who decides | QA reviewer | QA lead, from the dashboard and evidence table | A person for every flagged draft | QA lead. The Relay only reads and drafts |
| Effort to run | Grows with headcount | Template, checklist, transcript feed | One workflow, tested on past letters | Template set to your monitors and instructions |
| Best for | Calibrating checks and judging edge cases | Coverage of calls at volume | Written contact, where errors can be stopped | Making sure no critical breach waits for a weekly review |
These are not alternatives in the strict sense. Most lenders will end up with two or three of them. The useful choice is which problem each one solves, and which to build first.
Keep manual sampling, but change its job
Sampling a few calls per agent per month is how most collections QA works today. Its weakness is coverage: a breach on an unsampled call is never found, and a pattern takes months to show.
Its strength is judgement. Experienced reviewers understand tone, context and edge cases better than any checklist. So keep them, and move them from finding breaches to calibrating checks and reviewing the hard cases the other options surface.
Choose a Call Compliance dashboard for coverage
When the question is “did every call meet the rules”, you need every call checked. In Agent Serve this is a Call Compliance dashboard built with Dashboards, from the dashboard template of the same name. It is part of Dashboards, not a separate product.
- Transcripts in. The Transcripts upload panel takes txt, vtt, srt, json, md, docx, pdf or csv files, and call audio such as mp3, m4a or wav, with a choice of transcription model and optional speaker labels. For recordings collected on a schedule, a workflow with the Speech-to-Text block transcribes them automatically.
- Checks. Start from the Collections preset checklist, which covers the collection notice, right-party verification, threats or harassment, a clearly stated balance and hardship options. Add your own checks, such as UK-specific forbearance wording, each with a severity from low to critical. Every check returns pass, fail or not applicable, with a verbatim quote as evidence.
- Analytics. Each call also gets sentiment, escalation risk, next steps, talk ratio, a QA score and a summary, plus custom fields you define.
- Alerts. The template ships with a critical-breach rule, a breach-rate metric and two monitors: one for any critical breach and one for a breach rate above 5% by default. Set the threshold to match your own tolerance.
What the solution produces
Collections call check results (illustrative)
Record- Call
- Outbound, day 34 past due, personal loan
- Collection notice
- Pass · “This is an attempt to collect a debt, and any information obtained will be used for that purpose.”
- Right-party verification
- Pass · balance discussed only after name and date of birth were confirmed
- No threats or harassment
- Fail, critical · “If we don’t get it today, this goes further and it won’t be pretty.”
- Balance stated clearly
- Pass · amount owed split into principal, interest and one late fee
- Hardship options offered
- Not applicable · customer did not mention difficulty paying
- Vulnerability cue (custom)
- Pass · caller mentioned a recent bereavement; agent paused and offered a callback
- QA score
- 72 / 100
Row-level policies let each team leader see their own agents’ calls while compliance sees all of them. Run PII detection in the transcription workflow if transcripts should not carry full card or account numbers.
Add a pre-send check where you can still stop the error
Calls can only be checked after they happen. Letters and emails can be checked before. That makes written contact the one place where a mistake, such as a wrong balance, a missing required statement or threatening wording, can be stopped rather than remediated.
The pre-send check is a workflow. Your collections system sends each draft to a Webhook trigger. An Agent block checks the wording against your rules, a Table or API block pulls the account record so a Function block can compare figures exactly, and a Condition block decides. Passing drafts go back to your system to send. Flagged drafts pause at a Human block for a person to fix or release. The workflow never sends anything itself, and every run is traced in Logs.
Add the Compliance Reviewer when breaches cannot wait
A critical breach, such as a threat or a call to a customer who asked not to be contacted, should not sit until the weekly QA meeting. The Compliance Reviewer template is a Relay that starts when a breach monitor alerts. It runs at the Observe level, and its instructions say not to contact customers.
It reads the breach, the quoted evidence and the agent’s recent results, and posts a review note to the QA lead’s chat. In Observe it cannot save files, send email or book meetings. Coaching, remediation and any customer contact stay with people.
Calls checked after, letters checked before
Workflow · Audio to transcript
- 01Call recordingsRecordings from your dialler or telephony platform, collected on a schedule.
- 02Speech-to-TextA workflow with the Speech-to-Text block transcribes scheduled recordings automatically.
Dashboard · Call Compliance
- 03Transcripts panelTranscripts are uploaded to a Call Compliance dashboard built with Dashboards.
- 04Collections checklistPlain-English checks with severity; each returns pass, fail or not applicable with a quote.
- 05Breach monitorsOne fires on any critical breach, one when the breach rate passes a threshold.
Relay · Compliance Reviewer
- 06Starts on a breach alertThe Compliance Reviewer template, in Observe, starts when a breach monitor fires.
- 07Review noteSummarises the breach, the quoted evidence and the agent’s recent pattern for the QA lead.
- 08QA lead actsCoaching, remediation or escalation is the lead’s call. The Relay never contacts customers.
Workflow · Letter and email check
- 09Draft from your systemYour collections system sends each draft letter or email to a Webhook trigger.
- 10Check content and figuresRequired wording, tone and amounts are checked against the account record.
- 11Hold or releasePassing drafts go back to your system to send; flagged ones wait at a Human block.
Which to build first
- If written contact causes most of your complaints, start with the pre-send letter check. It prevents errors instead of finding them.
- If calls are the main channel and QA covers a small share, start with the Call Compliance dashboard and the Collections checklist.
- Once the dashboard is running, add the Compliance Reviewer so critical breaches reach a person within minutes.
- Throughout, keep manual reviewers on calibration and edge cases.
What the change could be worth
A US collections operation sending 15,000 letters and emails a month and sampling calls for QA.
| Assumption | Value |
|---|---|
| Letters and emails sent per month | 15,000 items / month |
| Share of written contacts with an error | 0.4% |
| Share of those errors the pre-send check stops | 80% |
| Cost to remediate one sent error | 250 USD |
| Days until sampled QA finds a critical breach | 15 days |
| Days until the monitor and Relay surface it | 1 days |
- remediation cost avoided a year: Letters and emails sent per month × Share of written contacts with an error × Share of those errors the pre-send check stops × Cost to remediate one sent error × 12
- letter errors stopped before sending a month: Letters and emails sent per month × Share of written contacts with an error × Share of those errors the pre-send check stops
- sooner a critical breach reaches a person: Days until sampled QA finds a critical breach − Days until the monitor and Relay surface it
Modelled from the assumptions above to show how the outcome is calculated; not a customer result. Your results depend on your volumes and process.
Customers in collections are more likely to be in vulnerable circumstances, so read this alongside vulnerable customer signals. For the same dashboard applied to every call in a contact centre, see call QA with full coverage.
Controls and governance
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.
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.
PII detection
A Guardrails block detects personal data and blocks or masks it before text reaches the next step.
Row-level access
Row-level policies narrow what each person sees in tables, dashboards, exports and scheduled reports.
Roles and permission groups
Admin, write and read roles per workspace, plus permission groups that limit which models, blocks and integrations a team can use.
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.
What you need
- Call recordings or transcripts from your dialler or telephony platform
- Your collections procedures and the rules you apply in each market
- The Collections preset checklist, reviewed and extended by compliance
- A feed of draft letters and emails from your collections system
- Account data the letter check can compare figures against
- A QA lead and a compliance owner to receive breach alerts
- Agreed breach-rate thresholds for the monitors