Small business email and team chat, 52 firms across 12 industries

Summary

Everyday email and public team-chat channels from 52 US firms of 5 to 40 staff in 12 industries, from HVAC service and freight brokerage to architecture and specialty distribution. Threads are rebuilt with quoted text stripped and attachments reduced to extracted text, so a model sees how small teams quote, schedule, chase late payments and handle problems end to end. Every message carries its firm code, industry and staff band, which suits cross-industry generalization tests and per-industry evaluation splits. Names, emails, phone numbers, addresses, account numbers and signatures are replaced with placeholders that stay consistent within each firm.

Collected to order. We collect it from US companies for your order. Scope, volume and timing are confirmed in writing before you sign.

Pricing on request

Request scope and pricing

At a glance

Planned scope. Volume, sources and timing are confirmed in writing before you sign.

Records: 34.5Memails and chat messages
Size: 88 GB
Tokens, estimated: 9B
Years: 2015 to 2026
Firms: 52

Owner profile: 52 US small businesses, 5 to 40 staff each, in 12 industries across every US region
Industry: Multiple industries
Data type: Company operations archive
Headcount: 5 to 40 staff per firm
Region: United States
Languages: English, Spanish
Formats: JSONL
Updated: 2026-10-02

Sources

System	Kind	Records
Google Workspace mail	Email	19.6M emails
Microsoft 365 mail	Email	8.8M emails
Slack public channels	Team chat	6.1M messages

Good for

pretraining and midtraining on realistic small-business correspondence
cross-industry generalization tests with held-out industries
training assistants to draft replies in a firm's own working style
evaluating thread summarization and next-action extraction

Fields

Field	Type	Description
message_id	string	Stable message id
thread_id	string	Thread or channel conversation id
channel	enum	email or chat
firm_code	string	Code of the contributing firm
industry	enum	One of 12 industry codes
sent_at	timestamp	Send time in UTC
sender_role	string	Job role of the sender, never the name
recipients	array<string>	Recipient placeholders
subject	string	Subject line, empty for chat
body	string	Message text with quoted replies stripped
attachments_text	array<string>	Extracted text of attachments
folder_path	string	Original mailbox folder or channel name

Example record

{
  "message_id": "[MSG_771204]",
  "thread_id": "[THREAD_20931]",
  "channel": "email",
  "firm_code": "F-27",
  "industry": "architecture",
  "sent_at": "2024-11-07T15:42:10Z",
  "sender_role": "project architect",
  "recipients": [
    "[EMAIL_118]",
    "[EMAIL_7]"
  ],
  "subject": "RE: Plan review comments, tenant buildout at [SITE_4]",
  "body": "[PERSON_12], the city sent back 4 comments. Three are quick fixes on our side. The fourth asks for a second exit from the training room, because the occupant load on our plan is 54 and anything over 49 needs two. We can add a door to the corridor or cut the room to 49 seats. The door keeps your layout and adds about a week. Can you confirm by Friday so we can resubmit Monday?",
  "attachments_text": [
    "Plan review comments, 4 items: 2 dimension notes, 1 extinguisher location, 1 egress from room 112"
  ],
  "folder_path": "Inbox/Projects/[PROJECT_19]"
}

One record, de-identified. Placeholders such as [PERSON_12] and [COMPANY_4] stay the same across the whole dataset, so threads still read correctly.

Licenses available

Exclusive
One buyer. Nobody else can license it during the exclusive term, usually 2 years.
Non-exclusive
Licensed to more than one buyer.
First look
One buyer gets 2 to 4 weeks to make an offer before anyone else sees the dataset.

What ships with it

Datasheet: what it is, what it is good for, verified size, sources, years, format and limits.
Provenance file: the owner shown by code, staff notice and consent status, exclusions, mapped to the EU AI Act Article 53 training-data summary and California AB 2013.
Manifest with a hash for every file.
QC report and de-identification report.
Your own copy, with unique canary records.

How it is de-identified

Default exclusions removed, files flattened to text, identifiers found in three layers and replaced with consistent placeholders, and the result measured against a 99% bar for each identifier type in each source.

The full method