Glossary Updates16 new terms added to the glossaries · 12 November 2026, 21:23 CET
AI TechDocKnowledge
  • English (US)
  • English (UK)
  • Deutsch
All context cards

Context card

Who decides which value counts: which system decides a value of product information?

When product information passes through PLM, ERP, content systems, spreadsheets, portals and AI functions, how do you tell which system decides a value and which one only holds a copy?

By knowledge.aitechdoc.world

Reviewed

Review log and changes

The short answer

Describe the landscape by roles, not by products: source, system of record, master data, integration, derived data, document assembly, exchange package, consuming system, system of engagement and acting system, with stewardship and lineage running across them. For every value that a decision rests on, ask one question: who may change it here — and how does the next system know which value counts? The system that writes first is not the system that decides.

For: Technical writers, technical project managers and documentation systems developers

Key points

  • The third reader is not one system: it is every system that reads product information, writes it on or compiles it into something new. AI is one reader among many.
  • A role is defined by what a system may do with one value, not by its product category; one product often plays several roles.
  • The system of record is decided per object or attribute, never once for a whole system.
  • A copy, a synchronization, a newer spreadsheet or an answer is not a source, and a reader that writes back becomes one.
  • A small lineage record that travels with the value — deciding system, revision, release, derivation, scope — is enough to check and correct it.

The context

Why product categories don't answer the question

A spare parts table in a manual looks like text. In fact it is a set of copies: a design revision released in product lifecycle management (PLM), a material and a bill of materials in enterprise resource planning (ERP), a scheduled job that copied them, a topic in a component content management system, a spreadsheet that service keeps and an answer function that reads all of them. Saying "PLM" or "ERP" says what a system can do. It doesn't say what it did with this one value.

The home page describes three readers of documentation: the source, the person and the third reader — the machines and markets in between. These three readers are placeholders. The check strategy Who decides which value counts: roles of product information from source to answer by Saina Veigel opens all three into roles — the source into writing and deciding, the person into using, the third reader into compiling, using and writing back. This card is its recipe card.

The ten roles at a glance

Role What the system does with a value Test question
Source Creates it for the first time Who created it, and is it released yet?
System of record Holds the value that counts, with revision and status Which system decides it — for this object?
Master data Keeps one agreed version across systems Is this the agreed version, or only the consolidated one?
Integration Moves it between systems When was it taken over, and from where?
Derived data Calculates, filters or copies it into lists and reports What was it derived from, and is that still current?
Document assembly Compiles values into a document for a variant or order From which revisions was this document assembled?
Exchange package Packs it for delivery with its metadata Which version and scope does the package carry?
Consuming system Reads it at the other end How is it told when a value changes?
System of engagement Shows it to people or answers with it Which revision does the answer come from?
Acting system Acts on it, sometimes writes back Under whose authority does it write?

Stewardship and data lineage are not an eleventh role: they run through all ten.

Where it usually goes wrong

Two confusions cause most of the trouble. A system that only passes a value on is treated as if it had decided it — the first system to hold a value, or the last one to show it. And a compilation is read as a source because it is the newest thing anyone sees: a spreadsheet updated this morning, a dashboard figure, a chat answer. A source of truth is therefore never named once for a whole landscape, only per value. Consistent master data is not the same as a released value, and a master-data synchronization is not a decision.

AI functions sharpen both confusions. A retrieval-augmented generation answer can mix two revisions unless answer provenance names one, and an agent that writes a status back is a source from then on — it needs an owner and an action approval gate, or the next synchronization spreads its value.

Evidence that is enough

Tracing every value through every system is no answer. For the values a decision rests on — identifiers, technical values, statuses, intervals, warnings and their translations — a short record is enough: the deciding system with revision and status, who released it and when, what it was derived from, its scope (variant, order, language), when and by which integration it was taken over, and for a write-back which system wrote it under whose authority. Most of it already exists in the systems as revisions and change numbers; what is missing is the line between them.

Limits

The roles describe one value's path for documentation. They are not an architecture for a particular product, they don't replace configuration management or data governance, and naming a method or specification doesn't mean that anything conforms to it.

Questions readers ask next

Is the third reader the same as AI?
No. AI functions are one reader among many and often the last in the chain. The third reader covers every system that reads product information, writes it on or compiles it — engineering and business systems, integrations, spreadsheets, portals, dashboards and AI.
Can one system play several roles?
Yes, and it usually does. The same ERP system can receive a bill of materials, decide the material master, export a list and feed a report — four roles for four different values.
Why is a spreadsheet a role and not just an error?
Because teams often believe it more than the governed systems. Naming it as derived data makes visible what it was derived from, so that it is corrected from the system of record and not the other way round.
When does an AI agent become a source?
As soon as it writes a value back, for example a part status. From then on it needs an owner, recorded authority and a gate in front of the system it writes to.

Sources

  1. Saina Veigel, Who decides which value counts: roles of product information from source to answer (check strategy) — knowledge.aitechdoc.world, 11 November 2026
  2. What we cover: nine fields, three questions, three readers — knowledge.aitechdoc.world, 10 November 2026
  3. Saina Veigel, AI-enabled information systems for technical writers (blueprint) — knowledge.aitechdoc.world, 7 October 2026
  4. Saina Veigel, The content lifecycle under AI: stages, gates and boundaries (blueprint) — knowledge.aitechdoc.world, 11 October 2026
  5. PROV-Overview: An Overview of the PROV Family of Documents (W3C Working Group Note) — W3C, 30 April 2013
  6. DAMA-DMBOK: Data Management Body of Knowledge, 2nd edition — DAMA International
  7. iiRDS – the international standard for intelligent information request and delivery — iiRDS Consortium, 10 November 2026
  8. Geoffrey Moore, Systems of Engagement and the Future of Enterprise IT (white paper) — AIIM

Review log and changes

Every context card is checked against its sources before it is published, and again whenever it changes; the date under the byline is the last review. Corrections (something was wrong) and additions (something was missing) are logged below with date and time (Berlin time). Typos, formatting and link fixes are not listed.

Reviewed

No corrections or additions since publication.