Custom datasets, sourced from real businesses.
If your AI model needs training data from inside real businesses and that dataset doesn't exist yet, we can build it. Tell us the industry, the kinds of records and the volume you need. We'll find the companies that hold those records, ask for their consent and come back with the scope, timeline and price.
How is a custom dataset sourced?
Training data is the set of examples an AI model studies to learn a task. We source it from businesses that agree to share their own work records, and we pay them for it.
Consented at the source
Every company in a dataset chooses to take part, signs a license and is paid for its records. A license is the written agreement that says what the buyer may do with the records, for how long and for what price. Nothing is scraped from the public web or guessed from other sources.
First-party business records
The records come straight from the software a company runs on, such as its help desk, its email and its accounting system. They belong to the business that created them, so the license comes from the owner and not from a reseller.
Scoped from start to finish
We handle the whole job: finding the companies, getting their consent, checking the records for quality, anonymizing them and delivering them. Anonymizing means removing names, contact details and other personal information. You can pay a flat fee, pay per record or set up a recurring pipeline.
Every company that contributes records is asked first and paid for what it shares. That is the whole model.
What records do businesses keep?
Every business leaves a trail of records behind each job it handles. Together those records show how real work gets done, which public websites rarely show.
Support tickets and their fixes
A help desk such as Zendesk records each customer's problem, every reply the staff sent and the step that finally solved it.
Email and chat threads
Email, Slack and Teams hold the conversations in which staff ask questions, hand off work and make decisions.
Quotes, invoices and orders
The accounting system shows what was sold, at what price, and how each order moved from a quote to a payment.
Dispatch and field service logs
The dispatcher, the person who schedules a company's technicians, records who was sent to each job, what equipment they took and how long the work took.
Sales and CRM histories
A CRM (customer relationship management system) records every lead, call note and stage a deal passed through, from first contact to signed contract.
Documents and procedures
Shared drives hold the contracts, policies and step-by-step instructions that staff follow to do their jobs.
Other kinds of records can be scoped to your brief.
Which industries can the records come from?
We find and recruit companies with targeted search and outreach campaigns, the same work we do for hiring and customer acquisition. That lets us look for the businesses your brief needs in almost any industry.
Other industries can be scoped to your brief.
Frequently asked questions
Who is this page for?
This page is for AI labs and data vendors. AI labs are the companies that build AI models. Data vendors are companies that buy records from businesses, clean and anonymize them, and package them into training material for AI labs. If either one needs records that no existing dataset covers, this is where to ask.
How do the companies give their consent?
Each company signs a license before any of its records are shared. The license says which records are included, what the buyer may use them for, how long the buyer may use them and what the company is paid. A company that doesn't sign isn't included.
Are the records anonymized?
Yes. Names, contact details and other personal information are removed before the records are delivered. What stays is the work itself: the request, the steps taken to handle it and how it turned out.
How is a custom dataset priced?
You can pay a flat fee for the whole dataset, pay per record or pay for a recurring pipeline that delivers new records on a schedule. The price depends on the industry, the number of records, the years they cover and whether you want exclusivity. Exclusivity means the records are licensed to you alone and can't be sold to anyone else.
Can I sell my own company's data instead?
Yes. This page is for buyers. If you run a business and want to be paid for your records, start with Data partnerships, or use the data value calculator to see what your company's data might be worth.