A referral email for someone who has never heard of training data: a ten-term plain-language explainer and one email rewritten from industry jargon into the owner's own words

Short answer

Write the email in the owner's vocabulary, not the AI industry's. Most owners have never needed "training data", "evaluation" or "licensing" in this sense, so translate each into something they already understand: examples a model learns from, tests it is checked against, and paid permission to use copies of records under a contract. Lead with the records you know they keep, say plainly what you would earn, and offer a reply that takes five words.

Visual briefFull-size diagram
An annotated email in five numbered blocks: your records, the category in plain words, what SourceX does, your reward and one easy reply, with jargon terms struck out beside the plain words that replace them.
The email structure from this page. The struck-out terms are examples of jargon to translate, not words SourceX uses with owners.

The owner you are writing to may run a 300-person business well and still have no reason to know how AI models are built. That is not a gap to fix with a lecture. It is a translation job.

Why a capable owner stalls on an email full of AI terms

An owner cannot weigh an opportunity they cannot picture. "Proprietary training corpora for fine-tuning" gives them nothing to picture, so they file it with the other messages that promise money for something vague. Jargon does not make the email sound expert to this reader; it makes it sound like spam.

Federal plain-language guidance makes the same point for any audience: technical terms are acceptable when no other wording will do, but the writer should define them, because the aim is to communicate, not to impress. For this email, almost every technical term has a plain substitute.

The translation still has to be accurate. The EU's AI Act gives usable anchors: it defines training data as data used for training an AI system through fitting its learnable parameters, and testing data as data used for an independent evaluation that confirms a system's expected performance before it reaches the market. The plain versions below stay inside those meanings. For longer explanations an owner can read later, link the explainers on AI training data and model evaluation.

The plain-language category explainer

Ten terms you might reach for, the words to use instead and what each plain version must not suggest.

Term you might useSay this insteadDo not let it suggest
AI training dataReal examples of work that an AI model studies to learn how a task is doneThat the company's systems get connected to anything
Evaluation or testing dataReal cases kept aside to check whether a model does the job wellThat the company grades AI products itself
LicensingPaid permission to use copies of selected records, on terms in a signed contractSelling the company, its systems or its customer list
Direct licensingThe company licenses to SourceX itself, under SourceX's agreementsThat the company negotiates with AI developers
Acquisition PartnerThe AI developer that licenses the finished dataset from SourceXThat you know who it is; partners are not told
Operating recordsEveryday records of how work got done: quotes, tickets, work orders, reviewsPersonal data, or records that belong to customers
MetadataA description of the records: what kind, how many years, which systemSending files or samples
Rights reviewSourceX checking that the company is allowed to license what it proposesThat you have checked anything
Actual Supplier PayoutWhat SourceX actually pays the company for a dataset, fixed in a signed ConfirmationA number anyone can quote before review
Referral rewardWhat SourceX pays you, the partner, if a deal completesThat it comes out of the company's payout

Use one or two of these per email, not all ten. The table is for you; the owner only needs the plain column.

Before and after: one email to the same owner (Illustrative)

Illustrative. A 210-person industrial equipment distributor run by its owner as CEO, with quotes, service tickets and returns notes since 2011 in one ERP system. The partner is a channel partnerships manager at the firm's accounting software reseller and has worked with the owner for five years. Both versions use exactly these facts.

Before:

Subject: AI data monetization opportunity

Hi [first name], AI labs are paying top dollar for proprietary training corpora to fine-tune LLMs and build RLHF and eval sets. Your data exhaust could be a major new revenue stream, with seven-figure payouts possible. SourceX is a data licensing platform that can monetize it for you. Can we jump on a call this week?

After:

Subject: Your service records, and a question you can ignore

Hi [first name], when we spoke in June you mentioned that every quote, service ticket and returns note since 2011 is still in your ERP. That's why I'm writing.

Companies that build AI models use real examples of work to teach their models how a job is done (that's "training data") and to check that they do it well ("evaluation"). A company called SourceX pays established businesses for permission to use copies of records like yours for those purposes, under a signed contract. It reviews the records and your rights to them first, handles the agreement and delivery, and pays you.

To be upfront: if you ever licensed anything through SourceX, it would pay me a referral reward. That comes from SourceX and never out of what it pays you.

If you'd like the one-page overview, reply "send it". If not, reply "not for us" and I won't raise it again.

[Your name]

In the first versionWhy it fails this readerIn the rewrite
"proprietary training corpora", "fine-tune LLMs", "RLHF and eval sets"Four terms the owner cannot picture"real examples of work", "teach", "check that they do it well"
"data exhaust"Implies the records are waste; also opaqueThe specific records the owner described
"seven-figure payouts possible"A payout claim nobody can make before reviewNo figure at all
"a data licensing platform that can monetize it for you"Vague about who pays whom for whatWho pays, for what permission, under what contract
No disclosureHides your financial interestOne plain sentence on the reward and who pays it
"jump on a call this week"A meeting ask with a soft deadlineA five-word reply, no clock

Put the company's reason before the category

The order of the rewrite is deliberate. An owner decides in the first two lines whether an email is about them. If those lines are about AI, the email reads as a pitch; if they are about the ERP the owner mentioned in June, it reads as a note from someone who listened.

So the sequence runs: their records, the category in one sentence, what SourceX does, your interest, one reply. Every sentence after the first can be checked against something the owner said or something SourceX publishes. If you cannot write the first sentence from a real conversation, you are not ready to send the email.

Explaining direct licensing and your reward without jargon

What SourceX does. SourceX pays companies directly for permission to use approved copies of their business data in AI training and evaluation. It checks the company's history, systems and rights. The company signs one Master agreement, and a short Dataset Deal Confirmation then fixes the exact dataset, net payout and terms for each deal. SourceX completes its review, the company signs, the data is delivered and SourceX pays the company. The data licensing explainer and the guide to the difference between an introduction and data delivery go further for an owner who asks.

What you earn. On a company's first qualifying transaction SourceX pays partners the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; the USD 50,000 floor applies once per referred company, and further qualifying transactions first agreed within 12 months of the first Closing pay 15%. The reward is paid separately and never deducted from the company's payout. A lead, a meeting or a signed agreement alone does not make it payable. In the email, the base and the payer matter more than any figure; the guide to the reward figure in an introduction email covers when a reward figure belongs in the message.

Put the disclosure in the body, in plain words, before the ask. The FTC's endorsement guidance sets a sensible bar for any paid recommendation: a connection the reader would not expect, and that could change the weight they give your words, should be disclosed clearly and conspicuously. A footnote or a postscript does not meet it.

What the email can say about fit, and what it leaves open

If the owner asks, describe the published checks in the same plain words: a full-time payroll of 50 or more at its peak, not counting contractors; years of records the business produced itself; the right to license them, which material belonging to its customers usually lacks; and a CEO who gives permission for the discussion and signs on the company's behalf. 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.

Send the email to the CEO, or with the CEO's knowledge, because an introduction the decision maker never heard about is usually not a fit for SourceX. What the email cannot say, because nobody knows yet: whether the records match a current SourceX program, what the rights review will find, what any license would cover and what it would pay. Those are settled in SourceX's review and the signed documents.

One reply the owner can send in five words

End with a choice that costs nothing either way. "Send it" brings the overview, your partner link and the program terms. "Not for us" ends the thread, and you honor that. If there is no reply at all, one short follow-up after a couple of weeks is the most you send; after that, the next move is theirs. The not-now guide covers an owner who is interested but busy.

Before you send the overview with your link, sign the SourceX Channel Partner Agreement. The credit belongs to whichever valid introduction came first within the 90-day attribution window; it is confirmed when the company submits a verified application, and each company is credited to one partner only.

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.

When an owner who had never heard of training data replies "send it", register for a partner link to include in the overview, check the company against who qualifies and keep the program terms ready for the questions that follow.

  1. Step 1Share your linkSend your personal link to a company you know.
  2. Step 2Company appliesThe company applies itself at /apply.
  3. Step 3Company signs and is paidThe company signs, delivers its data and is paid by SourceX under its agreement.
  4. Step 4You get your rewardthe greater of $50,000 plus 5% and 15% of the actual supplier payout, with no cap. A reward becomes payable after SourceX has paid the referred company under its signed agreement and the payout conditions in the program terms are met, including finance approval and your tax and payment details.

Common questions

Should the email use the words 'training data' at all?

Once, with its meaning beside it. The owner may meet the term again in SourceX's documents, so naming it helps, but the sentence should work even if they skip the term: real examples a model learns from.

How long should a first email to a non-technical owner be?

Short enough to read on a phone without scrolling much: one sentence on their records, two on the category and SourceX, one on your reward and one easy reply. The overview and the terms can follow when they ask.

What if the owner replies asking what their records are worth?

Say you cannot estimate it and that nobody can before SourceX reviews the records; SourceX fixes the dataset, payout and terms in a Dataset Deal Confirmation for each deal. Offer the application as the way to find out.

Sources

  1. refer.sourcex.si/terms
  2. refer.sourcex.si
  3. refer.sourcex.si/who-qualifies
  4. eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
  5. digital.gov/guides/plain-language/principles/avoid-jargon
  6. ftc.gov/business-guidance/resources/ftcs-endorsement…

Free resources

Know a company with valuable proprietary data?

Become a referral partner, get your link and introduce the company’s CEO (in an owner-run company, usually the owner). Referral partners can join from most countries. We can't work with partners in countries or regions under U.S. sanctions or on U.S. restricted-party lists. The companies you introduce should be based in the U.S., Canada and the EU, or in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia with at least 70% of their data in English.

Refer a company →

I own a business

Explore licensing your company's data directly to SourceX for AI training and evaluation. Start a short assessment; no uploads needed.

Start an assessment