Does context engineering data qualify? Scoring dated policy versions and decision notes

Short answer

To check whether a company's decision records could show context engineering, ask an operations, finance or service lead six things: which decision repeats against a written policy, whether policy versions are kept with effective dates, whether decision records name what was relied on, whether the approver and their authority are recorded, how far back history runs, and how the records link. Pass four gates first: size, rights, authority and, where relevant, English share.

Visual briefFull-size diagram
An arc gauge from 0 to 12 beside six criteria cards; as each card is checked the needle climbs and settles at 10 of 12, in the strong band, for a fictional 260-person managed IT provider.
Illustrative score for a fictional company. The score sorts an introduction; it does not predict acceptance or a payout.

This scorecard sits under the guide to context engineering in a real workflow, where one service credit carries three dates and a contract clause decides which policy version applies. That guide shows what a correctly assembled context looks like. This page is a scoring sheet for a 20-minute discovery call: answers only, no documents.

Gates: size, rights, authority and language

All four must pass before the score means anything.

GatePasses when
Size50 or more full-time employees at peak, contractors excluded
RightsThe company owns its decision records, and no vendor holds them under restrictive terms
AuthorityThe CEO, or the owner in an owner-run company, can authorize a licensing discussion
English shareOutside the U.S., Canada and the EU only: at least 70% of the data in English

The rule in full: 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. Documented operating history is checked by criterion 5 below.

The scorecard: six record criteria out of 12

Anthropic describes context engineering as curating the smallest set of high-signal information that makes the desired outcome most likely. A company's records show that skill only if a past decision can be rebuilt with exactly the information that applied at the time. Each criterion below scores 2, 1 or 0, for a maximum of 12. Criteria 2 and 3 also act as conditions, because nothing can be rebuilt without them.

CriterionStrong (2)Borderline (1)Weak (0)
1. Repeating policy-bound decisionA named decision with a yearly count: credits, exceptions, approvalsSeveral decision types, counts unknown"We make lots of judgment calls"
2. Dated policy versionsSeparate versions with effective dates in a document or contract systemVersion history exists but may be trimmedEdited in place, or only the current copy
3. Decision names its basisClause, policy or reference cited in the ticket or approvalSometimes cited; often in emailOutcome only, no stated basis
4. Approver and authorityApprover field plus an authority tableApprover recorded, limits undocumentedApprovals given verbally
5. History depthSeveral years in one systemSplit by a migrationRecent start or rebuilt records
6. LinkageRequest, incident, decision and billing share an IDLinked by customer name and dateRecords in unconnected tools

Bands. Strong is 10 to 12 with neither criterion 2 nor 3 weak. Borderline is 6 to 9, or a higher total with criterion 2 or 3 weak. Weak is below 6. Version trimming is a real borderline case: Microsoft notes that a SharePoint library keeps up to a set maximum number of versions and deletes the oldest when it reaches that limit, so a policy revised often can lose the version a past decision used.

Strong, borderline and weak, side by side (Illustrative)

All three companies are fictional.

Criterion260-person managed IT provider180-person property manager75-person staffing agency
1. Repeating decision2: service credits, about 400 a year2: rent concessions1: several exception types
2. Dated policy versions210
3. Decision names its basis220
4. Approver and authority221
5. History depth112
6. Linkage111
Score10 of 12, strong9 of 12, borderline5 of 12, weak

The managed IT provider keeps its service credit schedule as dated revisions, the service director cites the revision in each decision note, and an authority table sets approval limits, as in the main guide's example. The open items are a help desk migration and a billing system that links to tickets only by customer name and date.

The property manager cites the lease clause on every concession and records the approver against an authority table, but its concession policy lives in a library that trims old versions, and a 2021 system change split its history. Borderline, with "older policy versions may be trimmed" stated as the main open question.

The staffing agency edits its policies in place and records only outcomes. However many decisions it has made, nothing shows which rule applied, so the opportunity is theoretical for now.

Running the 20-minute discovery call

Ask the questions in this order; the early answers tell you whether the later ones matter.

  1. "Which decision repeats often and depends on a written policy or contract term?" Scores criterion 1.
  2. "Are old versions of that policy kept, with the dates they applied?" Scores criterion 2.
  3. "When someone decides, do they note which clause or version they used?" Scores criterion 3.
  4. "Is the approver recorded, and are their limits written down?" Scores criterion 4.
  5. "How far back do these records go, and in which systems?" Scores criteria 5 and 6.
  6. "Who owns these records and who could authorize a licensing discussion?" Checks the rights and authority gates.

Record answers, not documents. The guide to workflow context instead of isolated files explains why linked records count for more than volume, and the business introducer guide to operational data licensing covers the conversation more broadly.

Where a score misleads

  • AI access is not context. A company that connects assistants to its systems through the Model Context Protocol has given access, not a record of which documents were right for a past decision, and access is not a license.
  • A reviewer needs an answer key. Tools such as micro1's flow (described on micro1's blog, checked October 18, 2026) let a reviewer compare a model run's steps and tool calls against a rubric. A company's decision note is useful because it can serve as that key; a score of 2 on criterion 3 is what makes this possible.
  • The score is self-reported. It reflects what the lead believes. Whether old versions truly survived is checked in review.

Owners need 50 or more full-time employees at peak (contractors excluded), documented history, rights and an authorized contact. The score cannot show how many decisions pass a reviewer's check, how customer data will be handled, buyer interest, or any payment. Interest in context engineering across the industry does not mean every ticket archive qualifies. Partners never request policy files, exports or sample tickets. Owners can prepare with questions to ask before licensing company data; weak rows can be phrased using context engineering evidence gaps, and the company later prepares the context engineering review packet.

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 the company passes every gate and scores 6 or more with dated policy versions in place, register as a partner, check it against who qualifies, and read the terms before you introduce the COO or CFO.

  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

Why do criteria 2 and 3 also act as conditions?

Because they decide whether a past decision can be rebuilt at all. Without dated policy versions and a note naming what was relied on, a reviewer cannot tell which context was correct, however many tickets exist.

Does the company need to use AI already?

No. The score is about past decision records, not AI projects. Many firms with no AI plans keep dated policies and decision notes in ordinary ticketing, billing and document systems.

Can a low score improve later?

Yes, for future records. A company that starts keeping dated policy versions and citing them in decisions builds a stronger archive over time. Past records cannot be improved by rewriting them.

Sources

  1. anthropic.com/engineering/effective-context-engineering-fo…
  2. support.microsoft.com/en-us/sharepoint/lists/documents-and-library…
  3. micro1.ai/blog/introducing-flow
  4. refer.sourcex.si/terms

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