# Industry expertise in data referrals: what it adds | SourceX

Canonical: https://refer.sourcex.si/guides/what-industry-expertise-adds-to-a-data-referral
Last updated: 2026-11-04

Industry expertise adds precision to a data referral, not proof. A specialist knows what records a kind of company keeps, what the company calls them, which systems hold them, where rights tend to get tangled and who actually runs the operation. That turns "they must have data" into questions a CEO can answer in one call. Expertise cannot show that the company owns its records, that a program wants them or that the CEO wants to talk; each contribution has to become a fact the company confirms.

The map below assigns every piece of expert knowledge to the evidence it should produce and the person who confirms it, so a fund can see what its venture partners and operating experts actually add.

## The sector thesis, and the company question it should produce

Funds organized by sector bring in venture partners, operating partners and expert advisors because they know how companies in that sector work. A data supply thesis built on that bench runs like this: AI developers increasingly value work that only experienced practitioners can judge, so companies whose records were produced by experienced practitioners in our sectors are worth asking about, and our experts are the people best placed to find them.

Primary sources support the first half. OpenAI's GDPval evaluation spans 44 occupations drawn from the nine industries contributing most to U.S. GDP; its tasks were built by professionals averaging 14 years of experience, with at least four years in the occupation required, and a typical task takes an expert about seven hours. NIST's AI Risk Management Framework names domain experts as a distinct group of AI actors: practitioners who bring knowledge of an industry sector or application area, guide design and development, and help testing and impact teams interpret outputs.

Neither source says anything about licensing one company's records. They show that expert judgment is treated as an input to building and evaluating AI for professional work, and that is the part of the thesis a network can test. The company question it produces: which companies in our sector have years of records produced by experienced people doing the work, and which expert in our network can describe those records precisely enough for the CEO to confirm or correct them?

## The expertise-to-evidence contribution map

Every row starts with something the expert knows and ends with something the company says. Anything that never makes that crossing stays out of the introduction.

| What the expert contributes | Example from commercial property claims (Illustrative) | The evidence it must become | Who confirms it | How it goes wrong |
|---|---|---|---|---|
| Vocabulary | Knows a claims team calls its decision notes "reserve rationales" | A metadata question in the company's own words | Claims operations lead, through the CEO | Jargon in the note reads as a confirmed fact |
| Systems | Knows which claims platforms mid-size insurers run and what they retain | System name, first year of records, rough yearly volume | System administrator | The sector's usual retention is assumed to apply here |
| Process | Knows where escalations, approvals and reversals get written down | Which layers exist: decisions, reversals, outcomes | Operations lead | The company is assumed to document like the best firm the expert knew |
| Rights norms | Knows that files handled for other insurers are common in this line of work | A rights flag stated plainly in the introduction | The CEO, then SourceX's rights review | The expert's legal reading is presented as an answer |
| People | Knows who really runs claims and whom the CEO listens to | A named relationship owner | That person, by agreeing to ask | The expert knows the COO but not the CEO |
| Timing | Hears that the company is retiring a legacy claims system next year | A dated event with a source | The company | A rumor is treated as an event |
| Beyond expertise | Nothing in the expert's knowledge settles these | Ownership, demand, the CEO's consent, any payout | SourceX's review, the CEO and a signed agreement | Confidence stands in for evidence |

**Method note.** The rows follow the published screening profile (size, operating history, original documentation, rights, an authorized sponsor and relevance checked in review) and the program terms, checked on 2026-10-26. The map tracks where each statement in an introduction came from. It does not score experts, estimate demand for any record family or predict a payout. The insurance example is invented.

### One introduction note, before and after the expert (Illustrative)

A platform lead drafts a note about a 210-person regional commercial property insurer. The fund's venture partner, a former claims executive, annotates it. Each statement is then tagged by its status before anyone asks the CEO.

| Statement | Generalist draft | Expert-annotated version | Status before the first ask |
|---|---|---|---|
| Records | "They have lots of claims data" | "Reserve rationales and adjuster diary notes on each commercial property claim" | Expert knowledge, to be confirmed |
| System and years | "Probably one big database" | "One claims platform since 2013, an older system before that" | Confirmed by the claims VP |
| Volume | Not mentioned | "About 9,000 claims a year, by the VP's estimate" | Confirmed, approximate |
| Ownership | "Their data" | "Claims on its own policies are its own; files it handles for two other insurers under service agreements are theirs" | Flagged for rights review |
| Authorizer | "The claims VP is keen" | "The CEO decides; the claims VP will sponsor and the operating partner will ask" | Path agreed, CEO not yet asked |
| Size | "A big firm" | "About 210 full-time employees, contractors not counted" | Confirmed from the company's own count |

The annotated note is longer, but every line now says where it came from. The generalist draft would have introduced the third-party files as if they were the insurer's to license.

## Relationship and ownership evidence the expert cannot replace

The CEO authorizes a licensing discussion, but the person a sector expert knows may be the functional head who runs the operation. That makes the expert a strong sponsor and a weak authorizer: the right move is usually for the expert to brief whoever owns the CEO relationship, or to ask the CEO directly only when they know each other well.

Ownership is a separate question with its own evidence. A company needs rights to license its material without breaching client confidentiality, employee privacy or third-party contracts, and material that belongs to its clients is usually not a fit. An expert's sense of sector norms is a reason to raise a flag, never a reason to lower one.

Two limits apply to the expert personally. First, knowledge gained during a paid engagement may be confidential to that client; use only what the expert is free to share, and let the CEO describe the rest. Second, discovery stays at metadata level for everyone: record type, rough count, years, system and administrator, with no files, samples or logins. The [decision-maker discovery questions](/resources/decision-maker-discovery-questions) and the [metadata-only opportunity map](/guides/metadata-only-referral-opportunity-map) fit this step.

However much the expert contributed, one partner is credited per company, and dates decide which: the partner whose valid introduction of the new company is dated earliest inside the 90-day attribution window takes the credit, once the company files a verified application. Any earlier lead, account or email correspondence between SourceX and the company counts as an existing relationship, so ask about prior contact first. Decide in writing whether the expert, the platform lead or the relationship owner registers and submits.

## A test of whether expertise improved the introductions

Take ten companies from one sector shortlist and alternate them: the sector expert annotates the odd-numbered rows before the first CEO conversation, the generalist platform lead the even-numbered rows. Alternating, rather than letting the expert choose, stops the expert from taking the easy companies.

| Measure | Counted when | What you expect if expertise helps |
|---|---|---|
| Fit facts still unknown | At the first CEO conversation | Fewer on expert rows |
| Statements in the note the CEO corrected | At the first CEO conversation | Fewer on expert rows |
| Rights issues raised first by the company | Before submission | Fewer on expert rows, because the expert flagged them |
| CEOs who declined a conversation | Within four weeks of the ask | No prediction; record it |

Write the prediction down before the first call. Five rows per group can show only a large difference, so report the two sets of counts side by side rather than as percentages. If the expert rows show as many corrections and surprises as the generalist rows, the expertise added confidence rather than accuracy, and the fund should spend its experts' time elsewhere. For the separate question of whether specialists find candidates sooner, see [the industry experts who can find AI data opportunities first](/guides/the-industry-experts-who-can-find-ai-data-opportunities-first); for tracing a workflow into companies filed under other industries, see the [sector specialist's method](/guides/how-a-sector-specialist-finds-unusual-referral-candidates).

## Where expertise misleads

- **Sector defaults.** Experts generalize from the firms they knew best. A company in the same sector may keep less, or keep it differently.
- **Experts as a substitute for records.** GDPval shows a developer can commission experienced professionals to write and judge tasks directly. If expert judgment can be bought that way, the premium on companies' own records may be smaller than the thesis assumes.
- **Rights overreach.** An expert's reading of what clients usually allow is not a rights review. SourceX's review decides, and it can exclude records the expert was sure about.
- **A narrow list.** Specialists tend to know the best-known companies, which the fund may already know too.
- **Program fit.** SourceX checks relevance to its current data needs in review, so expertise about a sector cannot tell you what a program wants this quarter.

The [knowledge base assessment guide](/guides/how-to-assess-whether-a-knowledge-base-contains-original-expertise) helps an expert judge whether a company's documentation is its own original work, and the [operating partner page](/for/private-equity-operating-partners) covers the same decision from a portfolio operator's seat.

## 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.

Once an expert-annotated note has every statement confirmed or flagged and a relationship owner ready to ask, whoever will submit should [sign up for a partner link](/register), run the company past [who qualifies](/who-qualifies) one last time and read the [terms of the program](/terms) before the CEO is contacted. More network methods sit in the [guides hub](/guides).

Deploying a fund's experts on data introductions is a referral question; evaluating SourceX as an investment is a separate one, and anyone asking it can request investor materials from the [SourceX investor overview](https://sourcex.si/investors).

This article is for information only and is not an offer to sell or a solicitation of an offer to buy securities. Any offering by SourceX is made only to verified accredited investors through its offering documents.

## Sources

- https://refer.sourcex.si/terms
- https://refer.sourcex.si/who-qualifies
- https://openai.com/index/gdpval/
- https://nvlpubs.nist.gov/nistpubs/ai/nist.ai.100-1.pdf
- https://sourcex.si/investors
