How it works
From operating records to a licensed dataset.
Your systems already record quotes, investigations, inspections, and the decisions behind them. We identify the strongest dataset, prepare it for technical review, and broker a license with an AI team whose task fits the records.
- Assessment
- Free. It asks about your data, never for it.
- Our fee
- Agreed in writing before buyer outreach. Payable only on a closed transaction.
- Your approval
- You approve each buyer, the dataset package, and the license terms.
- First step
- About five minutes for the assessment, followed by a conversation.
01The process
Six stages, in order.
Each stage has a concrete output: a candidate dataset, an assessment, a documented package, a buyer brief, agreed terms, and delivery.
- 01
Identify
We identify datasets and workflows with potential AI value.
We look for records where an expert made a decision and the outcome was captured: a quote and whether it won, a ticket and how it was resolved, an inspection photo and the repair that followed.
- What we ask of you
- The free assessment, then a conversation with someone who knows how the records are produced.
- What sets the pace
- Availability of someone familiar with your records.
- 02
Evaluate
We assess quality, rights, scarcity, scale, and buyer demand.
We assess record quality, scarcity, scale, and fit for a specific AI task. The output is a dataset brief: what it contains, what makes it distinctive, and which training or evaluation tasks it supports. Rights reviewed before any outreach.
- What we ask of you
- Schema descriptions, field lists, record counts, and relevant agreements through an agreed diligence channel under NDA.
- What sets the pace
- Dataset documentation and completion of the review.
- 03
Prepare
We structure the extract, document the fields, and prepare a sample.
We define the extract and write datasheet-style documentation: collection methods, field definitions, the period covered, known gaps and biases, and every processing step. The result is a package a buyer can evaluate against its technical requirements.
- What we ask of you
- An engineer or analyst with source-system access, and your sign-off on the extract and processing rules.
- What sets the pace
- Source-system access, cleaning, and documentation.
- 04
Match
We find AI teams with a relevant training or evaluation task.
We approach AI labs, model companies, and enterprise AI teams around the dataset's specific use cases. You approve the prospective buyers. They review the documentation and sample under confidentiality terms.
- What we ask of you
- Approval of each prospective buyer, and of what each one may see.
- What sets the pace
- Availability of suitable buyers and their review schedules.
- 05
Negotiate
We negotiate license scope, permitted uses, exclusivity, update cadence, and pricing.
Price is one term among several. The license should define what the buyer may do with the data (evaluation, fine-tuning, pretraining, agent environments), for how long and where, whether exclusivity applies, how often updates are delivered, and what happens to the data and any trained models when the license ends.
- What we ask of you
- Your priorities for price, permitted uses, exclusivity, duration, and updates.
- What sets the pace
- Agreement on scope, pricing, and the final terms.
- 06
Close
Sign the license, deliver the dataset, and receive the agreed payment.
You and the buyer sign the license. The dataset is delivered on the agreed schedule and under the agreed security terms. Our brokerage fee is due when the transaction closes, at the rate agreed in writing before outreach.
- What we ask of you
- Signature, and delivery of the licensed extract as agreed.
- What sets the pace
- Signature, closing conditions, and the agreed delivery schedule.
02What is worth packaging
Volume is not the asset. Structure is.
Expert decisions linked to real outcomes give AI teams worked examples and test cases. Corrections, exceptions, and tool-use histories add structure for post-training, evaluation, reinforcement learning, and agent environments.
01
Input
The situation as it arrived: a request, a document, a photo, an instrument reading.
02
Expert decision
What a trained person chose to do: a quote, a classification, an escalation, a correction.
03
Real-world outcome
What happened next: the order won or lost, the repair that held, the customer who came back.
In ordinary operations
Customer support
- Support request
- Investigation
- Resolution
- Customer outcome
Distribution
- Quote
- Negotiation
- Purchase or loss
Field inspection
- Photo
- Technician classification
- Repair outcome
Signals buyers look for
- Expert decisions
- A trained person made a judgment call
- Outcomes
- What actually happened afterwards
- Feedback
- Ratings, reviews, customer responses
- Corrections
- Someone fixed or overrode a first attempt
- Rankings or comparisons
- One option preferred over another
- Multi-step workflows
- Ordered steps from request to resolution
- Software tool usage
- Which systems were used, and how
- Human demonstrations
- Recordings of people doing the task
- Failures & edge cases
- Mistakes, escalations, disputes
03After the first call
A confidential path, one step at a time.
Start with a dataset description and a call. We then agree a mutual NDA, review the records through a dedicated diligence channel, and prepare a package for buyers you approve.
01
NDA
Agree confidentiality terms for reviewing dataset details and documentation.
02
Secure diligence
Agree access controls, recipients, and transfer methods before reviewing structure, quality, and rights evidence.
03
Dataset sample
Review a small approved extract to assess format, quality, and fit for the intended task.
04
Data room
Organize approved documentation, the sample, and supporting evidence in an agreed data room with access controls for buyer review.
05
Buyer outreach
Buyers you have approved review the materials under confidentiality terms, before any full delivery.
04Fees, ownership, timing
The commercial terms, plainly.
- How fees work
The assessment is free. Our data opportunities service earns a brokerage fee only when a transaction closes.
The rate and structure are agreed with you in writing before any buyer outreach, so you know the economics before a buyer hears about the dataset. If no transaction closes, no brokerage fee is owed.
- Who owns the data
You keep ownership. A buyer receives a license for defined uses over a defined period.
- How long it takes
We agree milestones for dataset review, preparation, buyer evaluation, and negotiation. A well-documented schema and linked outcomes make the first review easier.
Start with a preliminary assessment.
Describe your systems in about five minutes. See your data opportunity score, a preliminary range where supported, and the factors that shape the dataset's AI uses.