Definitions and comparisons
Going direct to an AI company vs using an intermediary: pros and cons
By SourceX Editorial · Updated
Short answer
Going direct to an AI company means your own team finds the buyer, runs the rights review, prepares the records and negotiates the license; an intermediary runs those steps with you while you keep ownership and approval. Go direct only if you already have a qualified buyer, in-house contract and privacy capacity, and a way to judge fair terms.
Key takeaways
- Going direct removes one party from the deal but moves buyer search, rights work, privacy preparation and negotiation onto your own team.
- A data licensing intermediary manages the transaction for the supplier; a data broker typically acquires or aggregates data and resells it.
- Price discovery is the hardest part of going direct, because enterprise data deals are private and a single inbound offer gives no comparison.
- Whichever route you choose, your company should keep title to the records and sign every license itself.
- Ask any intermediary in writing whether it ever holds, resells or pools your records before you engage it.
What is the difference between going direct and using an intermediary?#
Going direct means your company deals with an AI developer's data team on its own: you answer their questions, scope the records, prepare them and negotiate the license with their lawyers. Using an intermediary means a third party runs some or all of that transaction alongside you, from matching your records to buyer requests through to delivery and payment.
The real difference is who does the work and who carries the knowledge of how these deals usually go. In both routes the records are licensed, not sold outright, and your company keeps ownership. What changes is how many conversations your team runs itself and how much it has to learn in the middle of a negotiation.
Be precise about the word intermediary. A data licensing intermediary acts for the supplier and does not take title to the records. A data broker usually acquires or aggregates data and resells it to many customers, which is a different business model with different risks for the original source.
How do the two routes compare side by side?#
The two routes compare most clearly across finding buyers, rights review, privacy preparation, price discovery, fees and control. Neither wins on every line, and the right choice depends on what your team can realistically carry alongside its day jobs.
Read the fee row together with the price discovery row. A fee is visible on paper; the cost of accepting a weak first offer, or signing a scope wider than it needed to be, is not.
| Area | Going direct | Using an intermediary |
|---|---|---|
| Finding buyers | You rely on inbound interest or your own outreach to model developers' data teams | The intermediary brings buyer requests that match your record types |
| Rights review | Your counsel reviews customer contracts, vendor terms and employee notices without a template | The review follows a set checklist; your counsel still makes the final calls |
| Privacy preparation | Your IT team removes personal and confidential details and documents the method | Preparation runs to a defined standard, and your team approves the result |
| Price discovery | One offer, no comparison, and the buyer usually knows the market better than you | Terms are framed against other transactions the intermediary has seen, though no public price list exists |
| Fees | No intermediary fee; the cost is internal time and outside counsel | A fee or share of proceeds, which should be fixed in writing before work starts |
| Control | You control every term, and every mistake is yours | You approve each step and sign the license; the intermediary should not sign for you |
| Delivery | You negotiate format, transfer method and acceptance tests yourself | Delivery method and acceptance are set up as part of the transaction |
When does going direct make sense?#
Going direct makes sense when the hardest parts of the deal are already solved. That usually means a specific buyer has approached you with a defined request, your legal team has negotiated IP or data licenses before, and your records are already well understood inside the company.
If more than one item on the list below is missing, the time your team spends filling the gap is often larger than any fee you would save. Treat it as a go-direct-only-if checklist.
- A named buyer has made a written request describing the record types, date range and intended use.
- Your counsel has negotiated IP or data licenses and can review an AI developer's paper without starting from zero.
- Someone on your team can remove personal and confidential details from support tickets, emails or code and document how it was done.
- You have a way to judge whether the offered terms are fair, such as advisers who have seen comparable deals.
- You can handle delivery, acceptance testing and any post-delivery deletion or audit requests without pulling engineers off product work.
- Your board, and any investors or lenders whose consent is needed, are aligned before negotiation starts.
What does an intermediary take off your plate, and what stays with you?#
An intermediary takes on the coordination work: matching records to buyer requests, running the rights checklist, organizing privacy preparation, structuring the license and managing delivery. A good one also keeps a written record of what was approved at each step, which matters when a buyer's diligence team asks questions months later.
Some decisions should stay with you on either route. Your company decides which records are in scope, your authorized signer signs the license, and your counsel gives the final view on rights and privacy. An intermediary that asks to sign on your behalf, hold your records indefinitely or resell them to other buyers is behaving more like a broker.
The practical test is simple. After the deal, you should hold the same records you held before, plus a signed license that names exactly what the buyer may do with a prepared copy.
Questions to ask any intermediary before you engage#
The questions to ask any intermediary focus on title, data handling, fees and exclusivity. Get the answers in writing, because each one should become a term in your engagement agreement rather than a sales assurance.
Compare the answers with the engagement agreement itself. If a sales conversation promises that the company keeps title but the draft agreement grants the intermediary broad rights over delivered data, the agreement is what counts, so have counsel read it before signing.
| Question | Answer you want | Warning sign |
|---|---|---|
| Do you ever take title to our records? | No; the company keeps ownership and licenses defined use | The agreement assigns the records or their IP to the intermediary |
| Do you hold or host our data? | Only what preparation requires, deleted after delivery | Indefinite storage or vague retention language |
| Can you resell or pool our records? | No; each license is a separate, approved transaction | Rights to combine your records with other suppliers' data |
| How is your fee set? | A stated fee or share, agreed before work begins | Fees that change once an offer arrives |
| Do we have to work only with you? | Exclusivity, if any, is narrow and time-limited | Open-ended exclusivity over all your data |
| Who signs the license? | Our authorized signer | The intermediary signs as agent without specific approval |
Illustrative: a software company weighs an inbound request#
Illustrative: a fictional mid-sized vertical software company receives an email from a model developer's data team asking about its engineering history. The company has many years of Jira issues, GitHub pull requests and Zendesk tickets linked to releases, and the CEO is tempted to reply and negotiate alone.
The CTO runs the request against the go-direct checklist. The request is specific, which is a point in favor. But the company has never negotiated a data license, its support tickets contain customer names and phone numbers, several enterprise customer contracts restrict sharing support content, and nobody can say whether the proposed terms are reasonable.
The CEO decides to use an intermediary for the first license while keeping the inbound buyer in the conversation. The rights review carves out tickets from the restricted customers, preparation removes personal details, and the company signs a non-exclusive license limited to coding assistant use. With the process documented, the team expects to handle a renewal with far less outside help.
How SourceX approaches direct and intermediated deals#
SourceX acts as a transaction layer for the supplier, not as a broker. Each license runs through the SourceX five-step transaction: Supply, Rights, Preparation, Approval and Delivery. The company keeps ownership of its records and approves every step, and the initial fit check collects metadata only, so nothing is shared at that stage.
Each completed transaction is documented in a SourceX Evidence Packet covering provenance, licensing rights, permitted use, the privacy record and release authorization. A company that later negotiates directly with a buyer can reuse that record to show what was cleared and why.
Frequently asked questions
Can we go direct for one deal and use an intermediary for another?
Yes, as long as no agreement says otherwise. Check any intermediary engagement for exclusivity, and check each license for clauses that restrict further licenses of the same records. Many companies use help on the first license, learn the process, then decide case by case for later requests.
Will an AI developer negotiate directly with a mid-sized company?
Some will, especially when a company holds a record type they are actively seeking. Buyer data teams tend to prefer suppliers who arrive with a clear inventory, documented rights and prepared records, because it shortens their own diligence. A loose offer of all our data rarely gets far.
Does an intermediary's fee mean we receive less?
Not necessarily. The fee is visible, but going direct carries its own costs: outside counsel time, engineering time on preparation, and the risk of accepting weak terms without any comparison. Compare the net outcome of both routes rather than the headline fee alone.
Is a data licensing intermediary the same as a data broker?
No. A data licensing intermediary manages a transaction for the supplier, who keeps title and signs the license. A data broker typically buys or collects data and resells it to many customers, often leaving the original source with little say over who uses it and how.
What should we share at first contact with a buyer or intermediary?
Descriptive information only: the systems involved, approximate years of history, the kinds of records you hold and any restrictions you already know about. Hold back samples and exports until scope, confidentiality terms and rights are settled, whichever route you choose.
Related resources
- IndustryLegal data
- QuestionDo AI labs buy legal documents?
- InsightDo you need a DPA when licensing de-identified data?
- InsightIs an AI data buyer a controller, a processor or a third party?
- InsightData licensing partner vs law firm vs doing it yourself
- SolutionData partnerships between businesses and AI developers
See if your company qualifies
A short company assessment. No data uploads are needed.