# Workflow data network introductions: who to ask | SourceX

Canonical: https://refer.sourcex.si/guides/when-a-thesis-calls-for-workflow-context-who-can-you-introduce
Last updated: 2026-11-04

Introduce the company that runs the workflow for itself. Around any workflow a thesis points to, four kinds of company sit close by: the operator doing the work for its own business, an outsourcer doing it for clients, the software vendor hosting it and the advisor who redesigned it. Usually only the operator owns the trail of how each case moved, so only its CEO can put those records on the table. The other three are how you find the operator, and sometimes how you reach its CEO.

So the first ask goes not to the best software company in the category but to the unglamorous operator that has run the same process for a decade and kept every handoff.

## What the thesis predicts, and the company question it becomes

Workflow context is the record around a task rather than the task itself: who received a case, who passed it on, what went wrong, what was decided and how it ended.

Anthropic's September 2025 Economic Index report, studying business use of Claude through its API, found that complex tasks arrive with long inputs and concluded that deploying AI for complex tasks might be constrained more by access to information than by model capabilities. Anthropic's engineering guidance on agent evaluations adds the testing side: 20 to 50 simple tasks drawn from real failures is enough to begin.

The thesis, then: dated, real cases of work moving between people and systems, failures included, are useful to teams building and testing AI for business work. That says nothing about any one company's records; SourceX checks relevance to its current data needs only in review.

For a network, the thesis becomes one narrow question: which company in our network runs this workflow at volume for its own business, has kept the trail for years, reached 50 or more full-time employees at peak, and has a CEO that someone we know can ask? Each clause removes a different group.

| Clause in the question | Who it removes |
|---|---|
| Runs the workflow for its own business | Outsourcers working on client files, vendors hosting customer data, advisors |
| Has kept the trail for years | Operators that record outcomes but never stored handoffs or decisions |
| 50+ full-time employees at peak, contractors excluded | Small specialist teams, however good their records |
| A CEO someone you know can ask | Companies you can name but could reach only with a cold approach |

## The workflow-owner referral discovery map

Pick one workflow, not a sector. List every company in your network that touches it and give each one of four positions. The example uses warranty claims on commercial equipment.

| Position | Around a warranty-claims workflow (Illustrative) | Whose records, by default | What this position can do for you | Introduce? |
|---|---|---|---|---|
| In-house operator | A 400-person equipment distributor that adjudicates warranty claims on what it sells and services | Its own, subject to customer contracts and privacy law | Its CEO can authorize a licensing discussion | Yes, once its CEO wants the conversation |
| Outsourced operator | A warranty administrator processing claims for 30 manufacturers | Mostly its clients'; client-owned material is usually not a fit | Discuss the records it created for itself, such as procedures and quality reviews | Only for material it owns, never client files |
| Software vendor | The claims platform both of them use | Its customers'; for EU personal data, a processor acts only on the customer's documented instructions | Tell its own customers about the program and let interested CEOs come forward | Not for hosted data; customers who opt in can be introduced |
| Advisor or implementer | The consultant who redesigned the distributor's claims process | None | Ask the operator CEOs she already knows | No; often the relationship owner instead |

Two rules make the map work. The position belongs to the record, not the company: a warranty administrator is an outsourced operator for its clients' claims and an in-house operator for its own hiring, billing and quality records. And the three non-owning positions are often the best route to the operator, through a vendor telling its customers or an advisor asking CEOs she knows, with nothing confidential passed on.

A filled map for one platform team:

| Row | Position | Company (Illustrative) | Trail known at metadata level | Relationship owner | Next action |
|---|---|---|---|---|---|
| 1 | In-house operator | 400-person equipment distributor | Claims, technician notes and approvals in one system since 2014 | Operating partner who sat on its board | Operating partner asks the CEO |
| 2 | In-house operator | 260-person property maintenance group | Work orders in a system; dispute notes in email | Portfolio founder who supplies it | Founder decides whether to raise it at all |
| 3 | Outsourced operator | Warranty administrator with 30 manufacturer clients | Client files; its own quality scorecards since 2019 | LP principal | Ask only about its own procedures and quality records |
| 4 | Software vendor | Portfolio company that builds the claims platform | Customers' data | Portfolio CEO | Vendor tells customers; opt-ins only |
| 5 | Advisor | Claims-process consultant | None | Platform lead | Ask which operator CEOs she can approach herself |
| 6 | In-house operator | 45-person regional repair shop | Complete trail, ten years | Platform lead | Park until full-time headcount passes 50 |

**Method note.** Positions and default owners come from SourceX's screening guidance, the program terms and, for EU personal data, GDPR Article 28, checked on 2026-10-26. The map does not measure demand, record quality or payout, and SourceX's rights review can overturn any default owner.

## Which layers of the trail did the operator keep?

An operator can own its records and still hold little workflow context. Ask, at metadata level, which of four layers survive and for how long:

1. **The case record.** What was asked for, by whom and when. Which system holds it, and since what year?
2. **The handoff trail.** Each change of owner or status, with a time. Does the system keep the status history or only the current status?
3. **The decision notes.** Why a case was escalated, approved or refused. Are decisions written in the system, or in email and phone calls?
4. **The outcome.** How the case ended and whether it came back. Can a reopened case be linked to the first one?

An operator that kept all four for years is the row the thesis describes; one whose decisions live in inboxes may still be worth a conversation if the introduction says so plainly. The live guide on [workflow context versus isolated files](/guides/how-to-identify-useful-workflow-context-instead-of-isolated-files) lists where these records tend to sit, and the [evidence-gap page](/questions/multimodal-workflow-context-evidence-gaps) shows how to describe a missing piece honestly.

## Relationship and ownership evidence before the first ask

Collect two kinds of evidence separately: that the records are the company's to discuss, and that a trusted person can ask the CEO.

| Evidence | What you need before asking | Who supplies it | Still unknown until SourceX reviews |
|---|---|---|---|
| Ownership | The records were created by the company's own staff for its own work | The operations lead, through the CEO | Whether customer contracts, employee privacy or third-party terms limit reuse |
| History | The years each layer covers | The system administrator | Whether those years match a current program |
| Authority | The CEO knows about the conversation and wants it | The CEO | What a license would cover and pay |
| Path | A named person who will ask the CEO | Your relationship owner | Nothing; settle this yourself |
| Prior contact | No earlier lead, account or email between SourceX and the company | The CEO, then SourceX's duplicate review | Whether credit is available |

Every question stays descriptive: record type, rough count, span of years, system and administrator; files, samples, screenshots and logins are out of scope. Ask about prior contact first, because Section 3 of the program terms treats any lead, account or email correspondence the company already had with SourceX as an existing relationship, and a new introduction cannot be credited for one.

One partner per company is credited, namely the one whose valid introduction came earliest within the 90-day attribution window, and that credit settles only when the company submits a verified application; when an operating partner, LP or portfolio founder owns the relationship, agree in writing who submits. For operators filed under a different industry code, the [sector specialist's tracing method](/guides/how-a-sector-specialist-finds-unusual-referral-candidates) finds them, and the [outsourced support page](/for/outsourced-support-providers) covers the outsourcer's own records.

## A narrow pilot, and the results that would change the thesis

Run the map on one workflow for six weeks, with three predictions and their refutations written before the first conversation.

| Prediction | Refuted if | What you change |
|---|---|---|
| In-house operators hold more of the four layers than any other position | Fewer than half the operator rows can describe a handoff trail or decision notes | Choose a workflow where decisions are written down |
| The three non-owning positions lead you to operators | Vendors and advisors produce no operator row you did not already have | Drop the outer positions and map operators directly |
| Operator CEOs asked by a trusted owner agree to a first conversation | Most decline, or the relationship owners will not ask | Treat the path as the constraint and fix relationship ownership first |

Keep declined and parked rows in the count, and report a dozen rows as counts, not rates.

One result would change the thesis itself rather than the map: operators that kept all four layers, introduced with their CEOs' agreement, repeatedly turned away for relevance. Demand for this workflow would then be weaker than the research implies. For the general method of turning a thesis into a short company pilot, see the [thesis-to-introduction canvas](/guides/from-ai-data-thesis-to-a-useful-company-introduction).

## Where the workflow-context thesis could be wrong

- **Simulated workplaces.** Researchers already build simulated companies to test agents: TheAgentCompany benchmark is a self-hosted software company with simulated colleagues and 175 professional tasks, on which the best baseline agent completed about 30% autonomously. If simulations satisfy developers, demand for real trails could narrow.
- **Scattered context.** The same Economic Index report flags dispersed context as a barrier to deployment; decisions buried in inboxes may never assemble into a usable record.
- **Rights.** Customer contracts and privacy law can limit even an operator's own records. SourceX's rights review decides, and it can say no.
- **Program fit.** SourceX applies each program's own criteria, so a strong operator can be declined today and fit another program later.
- **Goodwill.** Every ask spends a relationship owner's credibility with a CEO; record what it costs in your own pilot.

## 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 the map shows an in-house operator with a known trail and a relationship owner ready to ask, [register for a partner link](/register), check the row against [who qualifies](/who-qualifies) and read the [program terms](/terms) before the CEO hears from anyone. More network methods are in the [guides hub](/guides).

Choosing whom to ask is referral work; evaluating SourceX as an investment is a separate exercise, with investor materials available through 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://www.anthropic.com/research/anthropic-economic-index-september-2025-report
- https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents
- https://arxiv.org/abs/2412.14161
- https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A32016R0679
- https://sourcex.si/investors
