Skills Compliance

Using support conversations as training data

Support transcripts are the most valuable training and evaluation data a customer-service AI could have, and the most legally and ethically loaded. The request usually arrives as an engineering task and it is not one — the gating questions come first, and if they fail, no amount of filtering fixes it.

ComplianceData protectionPlaybookAny helpdeskRead-only
Installnpx rulebase-skills install cx-training-data-eligibility

When to use it

Reach for this when someone says any of these — they are the phrases the skill itself triggers on:

  • can we train on our support data
  • use transcripts to fine-tune a model
  • build an eval set from real conversations
  • is it ok to send our tickets to a model provider

How it works

The method, in the order the skill runs it. The full procedure — tables, worked examples and the edge cases — is in the skill itself.

  1. The gating questions, before any data moves

    Work through these in order and stop at the first unresolved one.

  2. Exclude these before anything else

    Report exclusions with counts. The proportion excluded is itself informative: a very small exclusion rate usually means the detection is not working rather than that the corpus is clean.

  3. Redaction is necessary and not sufficient

    Redaction reduces risk; it does not make a support conversation anonymous.

  4. Bias, which is a separate obligation

    A model trained on your historical conversations learns your historical behaviour, including the parts you would not defend.

Related skills

Free and open source, and vendor-neutral — it reads the conversations from whichever helpdesk you already run. Browse all 149 skills · connect your helpdesk over MCP · source on GitHub

Review every conversation. Act on what it finds.

AI for customer operations, built for financial services. Specialist agents chase every issue to resolution and every stalled customer to activation.

Rulebase dashboard