Most organisations do not lack data. They lack data they can trust: one figure in the finance report, another in the CRM, a third in someone's spreadsheet, and nobody sure which is right. Data work, as the Guild does it, closes that gap — mapping where data comes from and who owns it, agreeing what each measure means, building pipelines that keep it current, and making it usable by people and by agents.
Agents raise the stakes. A person who doubts a number asks someone; an agent acts on it. So the questions here are the familiar ones — sources, definitions, lineage, quality, access — asked with more care.
Questions this topic answers
- Which of our reports should we believe, and why do they disagree?
- Who owns each source and each definition?
- How do we know a pipeline is still healthy?
- Should an agent read this data from an index, or query it live from the system of record?
- What does an agent need from our data before it can act on it?
Start here
- Operational intelligence is the missing layer — why data only matters when it reaches a decision.
- The hidden cost of disconnected business systems — where the cost of re-keying and reconciliation hides, and why agents make it larger rather than smaller.
- Designing Event-Driven Business Operations — Field Guide 021: events that move work between systems, instead of people acting as the message bus.
- RAG architecture — retrieval for agents, and when it is the wrong pattern for live operational data.
- Memory is not a database — where records, knowledge and memory each belong.
Key ideas
- One data map — every source, its owner, what it holds, how fresh it is and where the gaps are. Most data work starts here.
- Definitions before dashboards — "active customer" or "on-time delivery" means one thing, written down and owned. Conflicting reports usually come from conflicting definitions, not bad arithmetic.
- Lineage — every figure can be traced back to its source and the transformations it went through.
- Quality is continuous — checks run on every load, failures alert an owner, and quality is reported like any other measure.
- The system of record wins — live operational figures are queried from the system that owns them, not copied to places where they go stale.
- Governed access — people and agents see only what their role allows, and every access can be audited.
- Data serves decisions — a pipeline is finished when it feeds a decision or an agent that needs it, not when the data lands in a warehouse.
Who knows this topic
- Atlas — sources and lineage: maps data, knowledge, integrations and dependencies so the Guild acts from trusted context.
- Hephaestus — pipelines and models.
- Oracle — quality and insight: resolves many streams into one trusted lens.
- Sentinel — access and governance.
Data engineers and domain owners work with Atlas and Oracle to make definitions explicit and keep every insight connected to its source. Atlas's human checkpoint is the Platform Wizard; the Data Wizard role builds the pipelines, models and dashboards that decisions rest on.
Where this knowledge is put to work
Data & Intelligence builds the foundation in order: an audit of sources, owners, gaps and decision needs; then a governed platform with canonical models, pipelines and quality controls; then pipeline management as a continuing service. AI Engineering prepares governed data and retrieval for agents.
Not sure which of your reports to believe? Start with the map of where each figure comes from.
