AI data sourcing
The data your models need, from the companies that hold it.
Acquiry finds the owners of specialist business data, qualifies them against your brief and coordinates the licence or acquisition. We're paid a percentage of the transaction, with no upfront retainer.
- Sourced against your brief
- 7 data types
- Data, IP or the company
- Licence or sale
- How Acquiry is paid
- % of transaction
- Nothing to pay upfront
- No retainer
Data we source
- Data typeSourced to brief
Operational workflows
Agent training, RL environments, evaluation
ServiceNow, Jira, Asana, Monday
- Data typeSourced to brief
Software engineering
Coding models, fine-tuning, executable environments
GitHub, GitLab, Bitbucket, Azure DevOps
- Data typeSourced to brief
Expert knowledge
Fine-tuning, retrieval, evaluation
Confluence, Notion, SharePoint, Google Drive
- Data typeSourced to brief
Customer operations
Support agents, evaluation, fine-tuning
Zendesk, Intercom, Freshdesk, Salesforce
- Data typeSourced to brief
Industrial and logistics
Specialist models, forecasting, evaluation
IBM Maximo, SAP, ServiceTitan, Procore
- Data typeSourced to brief
Audio, image and video
Multimodal training, evaluation
Call platforms, media libraries, inspection cameras
- Data typeSourced to brief
Structured business records
Training, evaluation, environments
NetSuite, Xero, QuickBooks, HubSpot
What Acquiry does
You define the data. We find the owners who hold it.
Buyer-led origination. Not a dataset marketplace, an annotation workforce or a model-training company.
Brief-specific research
We turn your use case, acceptance criteria and exclusions into a target profile: which kinds of company are likely to hold the records you need, and where.
Company identification
We map operators, rights-holders and specialist content owners against that profile, including businesses that have never thought of their records as an asset.
Owner access
We reach the people who can actually decide: founders, owners and executives, under the confidentiality controls agreed with you.
Commercial qualification
We test authority, willingness, high-level data characteristics and known restrictions before anyone spends time on samples or legal review.
Process coordination
We coordinate permissioned introductions, sample requests, evaluation steps and commercial discussions so each owner moves at a clear pace.
What an engagement produces
Agreed target profile
Company types, systems, history, geography and exclusions, signed off with you.
Qualified owner pipeline
Owners who match the brief, have authority and are open to a licence or sale.
Coordinated evaluations
Approved samples and technical and legal review, run to an agreed sequence.
How we're paid
Paid when the deal completes. Not before.
Acquiry charges a percentage of the transaction, with no upfront retainer. Our incentive is the same as yours: a completed deal with the right owner.
A percentage of the transaction
Set by deal size and complexity, and agreed in writing per mandate before work begins.
No upfront retainer
Nothing to pay to start. Fees are payable when a transaction completes.
Separate from owner proceeds
What a data owner receives is agreed directly in their contract with the buyer.
What we source
Seven kinds of data, each qualified on its rights.
None is automatically lawful to share or valuable. Every source is tested against your brief.
Operational workflows
The context of how work happens, not a pile of files.
- Linked requests, approvals and exceptions
- Documents tied to decisions and outcomes
- Typical systems
- ServiceNow, Jira, Asana, Monday
Software engineering
Code with the history that explains it.
- Issues linked to pull requests
- Review comments and test results
- Typical systems
- GitHub, GitLab, Bitbucket, Azure DevOps
Expert knowledge
How specialists say the work should be done.
- SOPs and technical manuals
- Internal knowledge bases
- Typical systems
- Confluence, Notion, SharePoint, Google Drive
Customer operations
How real problems get resolved.
- Authorised support records
- Resolution histories and reopen paths
- Typical systems
- Zendesk, Intercom, Freshdesk, Salesforce
Industrial and logistics
The physical operation, recorded over time.
- Inspection and maintenance records
- Dispatch and inventory events
- Typical systems
- IBM Maximo, SAP, ServiceTitan, Procore
Audio, image and video
Media with the metadata and consent to match.
- Rights-controlled recordings
- Images with useful labels
- Typical systems
- Call platforms, media libraries, inspection cameras
Structured business records
Chronology and relationships that stay useful.
- Transaction and event histories
- Records linked across entities
- Typical systems
- NetSuite, Xero, QuickBooks, HubSpot
Fictional example
A brief for field-service dispatch records
- 1An agent developer needs five or more years of linked dispatch, job and outcome records from field-service operators in English-speaking markets, for an evaluation set.
- 2The target profile narrows to HVAC, electrical and plumbing operators of a certain size using job-management software, excluding firms with restrictive customer contracts.
- 3Owners who are open to a scoped licence, have authority over the systems and can describe their history at a high level go forward to an approved sample and review.
Heightened review:FinanceLegalHealthcareWorkforce and HRSystem names are examples only and imply no partnership.
For data buyers
Built for teams with a funded data requirement.
Foundation-model developers
Labs that need domain-specific records for training, fine-tuning or evaluation.
Specialist AI developers
Teams building vertical models or agents that need deep, narrow supply.
Data procurement firms
Enterprise-data programmes supplying AI labs and needing more qualified partners.
Training-data suppliers
Aggregators and suppliers filling a specific category, language or industry gap.
Environment builders
Teams building agent-training and reinforcement-learning environments from real workflows.
Evaluation teams
Groups building benchmarks and test sets that must reflect real operating conditions.
Usually led by
- Heads of data acquisition
- Data partnerships
- Strategic sourcing
- Research data
- Frontier environments
- Procurement
- Founders of specialist AI businesses
What goes into a brief
- Data type
- Resolved support tickets with full threads and outcomes
- Industry
- Field service, MSPs, freight brokerage, property management
- Systems
- Records held in a named platform, such as Zendesk or ServiceTitan
- History
- Five or more continuous years on the same tools
- Volume
- An approximate record count or range that would be useful
- Geography and language
- English-language records from the US, UK, Canada and Australia
- AI use
- Evaluation only, or training and evaluation, stated explicitly
- Rights wanted
- Non-exclusive licence, or exclusivity for a fixed term
- Exclusions
- No healthcare records, no firms with restrictive client contracts
How it works
Six stages, with a permission gate where it counts.
Each stage ends with something you can review. No raw data moves before an owner approves it.
- 01
Define the brief
We agree the intended AI use, budget range, acceptance criteria, exclusions and the rights you need.
Signed-off sourcing brief
- 02
Research owners
We identify companies and rights-holders likely to hold matching records, and rank them against the brief.
Target company profile
- 03
Approach under controls
We contact owners without naming you unless you have agreed it, under the confidentiality terms set in step one.
Owner conversations
Your approval to be named
- 04
Qualify
We confirm authority, willingness, high-level data characteristics and known restrictions. No raw data at this stage.
Qualified owner shortlist
- 05
Approved evaluation
Where both sides agree, samples move for technical and legal review with the appropriate specialists.
Evaluation packs
Owner approval of each sample
- 06
Commercial terms
We coordinate the discussion of scope, rights, price and next steps between buyer and owner.
Agreed next steps
Written agreement between the parties
Why Acquiry
Finding owners who aren't looking is what we already do.
Acquiry runs buy-side origination. AI data sourcing applies the same discipline to a new kind of asset.
Targeted origination
Our core work is finding specific companies against a brief, not waiting for whoever applies to a programme.
Off-market conversations
Many of the best-matched owners have never considered licensing. Reaching them takes direct, credible outreach.
Commercial qualification
We test whether an owner can and will transact before it reaches your data and legal teams.
Transaction discipline
Confidentiality, sequencing and clear permissions are how we already run buy-side processes.
| Point | Open application programme | Acquiry origination |
|---|---|---|
| Who starts it | Owners apply when they hear about it | We go to owners against your brief |
| Which companies | Whoever finds the programme | Companies matched to your criteria |
| Owners not looking | Rarely reached | Our core work |
| Before your team spends time | Screening happens after the application | Authority and willingness are tested first |
| Your identity | Usually public | Named only when you approve it |
For data owners
Your records may be worth more than the storage they sit in.
Where a buyer brief matches, a scoped licence may be possible without selling the company.
- A scoped licence can be possible without selling the company.
- Ownership, exclusivity, reuse and model-training rights are set by the contract.
- Existing customer, employee and third-party obligations come first.
- You decide what, if anything, is shared, and when.
- Software and SaaS
- MSPs
- Professional services
- Agencies
- Staffing
- Customer support operations
- Manufacturing
- Engineering
- Logistics
- Field service
- Property operators
- Publishers and content owners
Three ways a transaction can work
License selected data
The owner keeps the business and the data, and grants defined rights for a defined use and term. The most common starting point.
Owner keeps the company and the underlying data
Acquire a dataset or IP
The buyer takes ownership of a specific dataset, archive or IP asset. Rights in the asset move, the business stays with its owner.
Owner keeps the company; the asset moves
Acquire the business
Where the data is inseparable from the operation, a buyer may acquire the company itself. This is a full M&A process.
Ownership of the company moves
Usually useful
Support tickets with resolutions
The problem, the back-and-forth and the outcome
CRM history with outcomes
Deals moving through stages, won or lost
Issues linked to code and reviews
Why the code ended up the way it did
Usually less useful
Marketing copy and finished reports
The reasoning has been edited out
Public or scraped material
AI teams can usually reach it already
Files owned by your clients
They are not yours to license
The research behind it
The AI data economy, in five studies.
Primary-source research on why company data became valuable, what it sells for and who is buying.
Questions
What buyers and owners ask first.
What is the difference between licensing and acquisition?
A licence grants defined rights for a defined use and term while the owner keeps the data and the company. An acquisition moves ownership of a dataset, IP asset or, in some cases, the whole business. The right structure depends on the data and both parties.
Does permission for one AI use cover others?
No. Pretraining, fine-tuning, evaluation, retrieval and reinforcement-learning environments are different uses. Each permitted use has to be agreed explicitly in the contract.
How is the value of a dataset decided?
By relevance to the brief, uniqueness, usable coverage, quality, provenance, rights, permitted uses, preparation effort and whether the buyer accepts it after evaluation. There is no universal formula, headcount threshold or payout calculator.
Can a company sell every record it holds?
No. Customer contracts, employee information, privilege, confidentiality, third-party IP and re-identification risk all limit what can be shared. Finance, legal, healthcare and workforce records need heightened review.
Who handles consent, rights review and de-identification?
These are scoped with the buyer and appropriate legal, privacy and security specialists. Acquiry does not perform anonymisation, security certification or legal clearance, and does not guarantee compliance or anonymity.
Does licensing give competitors our trade secrets?
AI teams generally want records of how everyday work gets done, not strategy. What is excluded, how it is removed and who reviews a sample before acceptance are set in the contract, and you approve each sample.
Do we need to upload anything to enquire?
No. The first step is a short, non-sensitive description of your business and the systems you use. Nothing is shared until a buyer brief matches and you approve it.
How are fees structured?
Acquiry charges a percentage of the transaction, set by deal size and complexity and agreed in writing per mandate. There is no upfront retainer. Acquiry’s fees are separate from anything a data owner receives from the buyer.
What do you need from us to start?
Buyers: contact details and a short description of the requirement. Owners: contact details, your role and a short, non-sensitive description of the business. No raw data, ever, at enquiry stage.
Start here
Tell us what data you need.
Contact details and a short brief are enough to start. Add more detail now or later. Data owners can use the second tab.
Please do not paste or describe in detail:
- Customer records
- Employee information
- Privileged material
- Health records
- Credentials
- Confidential datasets
There are no uploads on this form. Raw data is never needed at enquiry stage.