The confusion is natural for anyone who lives in admin consoles. If you can get the data out, it feels as though the hard part is done. In SourceX's process the hard part is somewhere else entirely, and your access never reaches it. The nearest record type for many MSP clients is described in technical documentation referrals; this page is about the two reviews any record type passes through.
Two reviews that ask different questions, in a fixed order
SourceX publishes its sequence, and the order is the point. When a company applies, it answers a few questions and lists its systems, its years of history and what can be exported; no records change hands. SourceX then checks fit, completes the inventory with the company, runs the rights check and sets the processing route. Only after a Master agreement and a Dataset Deal Confirmation, which fixes the exact dataset, the payout, rights, the processing route and deadlines, and only once SourceX gives the go-ahead, are records extracted, transferred and de-identified through approved processing routes. SourceX's stated target for that step is 7 to 10 days once access and inputs are ready.
So "what can be exported" enters the process as a sentence the company writes, and SourceX chooses the processing route from that description. Nobody runs an extraction until the end, by which point the dataset is agreed and signed for.
How complete can feasibility be? Google documents a Data Export tool that lets a super administrator export all of an organization's data. Its conditions are about the administrator and the mechanics: the admin account must be at least 30 days old unless the organization itself is newer, 2-Step Verification must be on, Google holds the export for 48 hours as what it calls a security measure, and a full export typically takes 72 hours and can take up to 14 days. That is feasibility at its most total: one person, one console, the whole organization. Not one of those conditions asks who created the mailboxes and files, what the company's customer contracts allow or whether anyone wants to license them.
When an IT project makes someone ask "could we license this?"
The question rarely starts with you. It starts with a project that puts old records in front of someone with budget authority.
| Project | What the client notices | Who tends to ask | Who decides |
|---|---|---|---|
| Migrating to a new platform | Years of history that may not be carried across | IT director or COO | The CEO |
| A storage cost review | A decade of archives costing money every month | CFO or controller | The CEO |
| An acquisition | A second company's systems and records | Integration lead | The acquirer's CEO once the deal closes; until then, the target's CEO |
| A vendor's end-of-life notice | A deadline after which a system can no longer be read | The system's owner | The CEO |
| A new data warehouse | Records gathered in one place for the first time | Head of data or analytics | The CEO |
The person who asks is a useful internal sponsor. They can take the question upstairs, but under the program the CEO alone authorizes a licensing discussion and signs for the company. Your job at this point is a sentence about the program and a sentence about your own interest, not an estimate of what the archive is worth.
The two-stage review comparison
Read each row as one question put to both reviews. The right-hand column is the one that decides whether there is anything to introduce.
| Aspect | Stage 2: technical feasibility | Stage 1: licensing eligibility |
|---|---|---|
| The question | Can the records be extracted, moved and prepared in usable form? | May the company license them, does it fit a current program, and does its CEO want to? |
| Who answers | The company describes its systems; SourceX picks the processing route from that description | SourceX's fit review and rights check, the company's own confirmations and the CEO |
| Evidence used | System documentation, export features, formats, access arrangements | Peak full-time headcount of 50 or more with contractors excluded, years of the company's own records, who created them, customer and employee terms, any earlier AI-training license, location and language, fit with current demand |
| When it happens | Described at application; tested in practice only after the Confirmation is signed and SourceX gives the go-ahead | At application and qualification, before any agreement |
| What a yes proves | The records can move | The company can continue toward a Confirmation; nothing is bought yet |
| What a yes leaves open | Rights, fit, demand and the CEO's wishes | How easy or quick extraction will be |
| The MSP's part in the referral | None; preparation runs through SourceX's approved processing routes, arranged between the company and SourceX | Describing the program and disclosing your interest; the company answers for itself |
The stages are numbered in the order SourceX runs them, which is the reverse of the order a technician tends to think in. Then sort a client into one of four boxes:
| Published tests look passable | A published test fails | |
|---|---|---|
| Easy to export | Introduce; the export adds nothing to the introduction | Do not introduce on the strength of the export; say which test fails |
| Hard to export | Introduce anyway; SourceX chooses the processing route from the company's description | Do not introduce; the difficulty is beside the point |
A published test fails, for example, when full-time headcount peaked at 38, when the records mostly belong to the client's own customers, when they were already licensed to someone for AI training, when the company sits outside the published geography, or when the CEO has said no. For the geography rule, in SourceX's words: Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English.
Method note. Each row of the comparison cites a SourceX document read on 2026-10-27: the published process on how-it-works, the screening page, the referral terms in Version 2026-10-22.2, since replaced by version 2026-11-04 and the partner agreement sample (unsigned, 2026-10-19). Google's export documentation serves only as an example of feasibility; no other vendor was examined. The table describes SourceX's published sequence, not any individual deal, and SourceX has published no figures on how often companies pass or fail either stage.
Three things an administrator account cannot settle
Your credentials answer feasibility questions in theory. They cannot reach three others.
- Who owns what is inside. Records written by the client's staff about its own work are one thing; files its customers uploaded, or records created under a customer contract, are another. SourceX's screening guidance asks whether the company can license material without breaching client confidentiality, employee privacy or third-party contracts.
- Whether licensing is permitted at all. When a company signs with SourceX it confirms that it has the rights, that personal data was collected lawfully, that the data has not been licensed for AI training before and that it was not created or altered with AI for the sale. No console shows any of those.
- Whether the company wants to. That is the CEO's call, and SourceX's screening page treats a company whose decision maker was never told as a poor prospect.
Your service contract and your partner agreement both stop short as well. The contract gave you access to run systems, not to copy records for a deal the client has not decided on. The partner agreement gives you nothing to add: section 4.2 of the unsigned sample withholds any authority to pass on a supplier's inventory, records, raw data or personal information, or to reach extraction systems. Partners never see, upload or handle a company's records, inventory or data room, however much access their IT work gives them. A provider's standing more generally is covered in can an IT provider share client data.
Same export button, opposite answers (Illustrative)
Illustrative. Two clients of one MSP, with invented facts and no outcome implied.
| Client A: 120-person document services firm | Client B: 300-person mechanical contractor | |
|---|---|---|
| Main system | Modern cloud platform with a documented export function | On-premises service system from 2008, read-only since a 2023 migration, no export function |
| Export effort, in the MSP's view | An afternoon | Weeks of work, if anyone ever needs it |
| Who created the records | Mostly its customers, who upload their own documents | Its own technicians and estimators, over 17 years |
| CEO's view | Not yet asked | Interested, after the COO raised it |
| Published screen | Client-owned material usually fails | Headcount, history and CEO pass on the company's own account; rights to be checked in SourceX's review |
| What the MSP does | Nothing: the easy export changes nothing | Introduces the CEO, carrying only facts the CEO confirmed |
Client A is the trap. The export would be quick and complete, and almost everything it produced would belong to someone else. Client B is the one an MSP might talk itself out of because the old system is awkward, which is exactly what SourceX's choice of processing route is for, with nothing extracted until after signing.
Suppose SourceX later paid Client B USD 200,000 under its Dataset Deal Confirmation. The reward on that first qualifying transaction would be the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap: USD 60,000 here, paid by SourceX to the MSP once every payment condition in the terms is met, with nothing deducted from what Client B is paid. The figure is invented, and SourceX does not publish how it prices offers.
What to tell the CEO without running a test first
Keep feasibility out of the pitch entirely. Three sentences cover it:
- "I'm raising this as a SourceX referral partner, separately from our IT work, and my firm could be paid a reward if you sign."
- "Whether these records can be licensed is decided in SourceX's review, from your description; you don't send files to apply."
- "If it ever gets to extraction, that happens after you sign, through routes you agree with SourceX, not through us."
Then send the screening page, or the non-binding company fit checker. Phrases to drop: "I checked, it exports cleanly", "export-ready", "we could have it out in a week", "your data is clean", and anything that sounds like a promise about security or de-identification beyond what SourceX itself publishes. Why "it fits" and "it was accepted" are several stages apart is laid out in fit check versus buyer acceptance.
Where feasibility does feed into fit
The two stages are separate, not unrelated.
- Technical traits are among the fit signals. SourceX lists volume, historical depth, data structure and clear ownership among the signals it weighs. It judges them from the company's description during review, not from a partner's test.
- Records that no longer exist cannot be licensed. A wiped backup or a system decommissioned without an archive ends the question. If your service records show that, tell the CEO, not SourceX.
- Vendor-hosted systems. Where records sit in a vendor's platform, the vendor's terms can affect rights. That is a rights question for SourceX's review, not a feasibility one.
- A shutdown date. An end-of-life deadline is real urgency for the CEO's decision. It is never a reason to run an export early.
- Corporate groups. The USD 50,000 floor applies once per referred company, to its first qualifying transaction only, and whether a parent's, subsidiary's or sister company's earlier dealings with SourceX affect credit is left to SourceX's duplicate review.
Next step
Refer a company. No cap on rewards. Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.
If a client lands in the left-hand column of the four boxes and its CEO wants the conversation, become a partner, run the company past the published screen rather than an export, and keep the program terms to hand when you write the introduction.