Seller field guide · 3 min read
What Intercom data can you assess for an AI license?
Begin with permitted conversation content and the decisions your team made, not just an Intercom reporting export. A CSV of response times or conversation counts describes performance, but it may not contain the dialogue a buyer wants to evaluate. Define content, metadata and exclusions separately.
A reporting CSV is not a conversation transcript
Intercom’s export guide distinguishes reporting CSV data from conversation content. It states that CSV reporting exports omit message bodies, and that individual transcript exports omit internal notes. It documents separate routes for retrieving conversation content.
Choose a route based on the requested evidence, then inspect its output. A report can be a useful inventory of time periods and attributes, but do not present it as proof that full historical dialogue is available. Check the history window and permissions for the exact method in your workspace.
Source notes: Intercom: export conversation data
Separate human decisions from automated replies
For an evaluation, identify where a customer request entered the workflow, when automation responded, when a person intervened and how the case ended. A conversation made entirely of templated responses is different from a difficult escalation with new reasoning. Describe that composition explicitly.
If internal notes are absent, the visible transcript may show the answer without explaining the handoff that produced it. Treat this as a content limitation. Do not reconstruct missing decisions from an AI-generated summary or assume that a summary is a verbatim account.
Choose a defensible evaluation scope
This is our suggested scoping framework, not a statement that a buyer accepts every record below. Review contractual rights and the intended use before preparing a sample. Technical access to a record does not establish permission to license it.
| Record or context | What to evaluate | Boundary to keep |
|---|---|---|
| Reporting attributes | Coverage, channels and response metrics | Keep separate from claims about full message content |
| Conversation transcripts | Question, response and chronological outcome | Remove identifiers and restricted customer content |
| Internal notes or summaries | Whether the evaluation needs this additional context | Review scope and label summaries as summaries |
Check the buyer’s intended evaluation unit
Ask whether the buyer wants a full conversation, an escalation segment or a question-and-answer pair. These units need different selection rules. Cutting a thread immediately before an escalation can make an unsuccessful response look final.
Prepare a schema that identifies human and automated participation when that distinction is available and permitted. Explain missing notes, attachments and off-platform follow-ups. A small, transparent sample gives a clearer basis for discussion than a broad export with unexplained omissions.
Illustrative scenario · fictional company
Intercom: a fictional scoping example
A fictional software company evaluates Intercom onboarding conversations. It first produces a reporting CSV and realizes it cannot show the actual questions. The team prepares a permitted transcript sample instead, excluding billing identities and customer-uploaded files. Human escalations and automated responses are identified. Where a resolution occurred by email outside Intercom, the sample says that the outcome is unavailable rather than claiming the conversation was successfully resolved.
When to pause or narrow the scope
Client work and customer files are excluded from this starting scope. Assess company-owned internal records separately from personnel, health, privileged or NDA-restricted material.
- Customer obligations do not permit the proposed use.
- The available content is mostly repetitive automated replies.
- The useful reasoning lives in notes or external systems that cannot be included.
Your preparation checklist
- Distinguish reporting metrics from actual conversation text.
- Confirm the accessible history and export permissions.
- Identify automated and human participation where available.
- Review customer uploads, internal notes and summaries separately.
- Document missing context and agree on evaluation retention.
Start with a description of the records. The quote form does not ask you to upload a dataset.
Discuss your dataSources and further reading
- Intercom: export conversation data · Checked October 26, 2026
This is general information, not legal advice or confirmation that a buyer will accept your records. Published buyer ranges describe their programs, not an offer for your dataset. Buyers may pay QuoteMyData when a deal closes; see how we are paid.
