Documents versus records
Most organisational knowledge lives in documents: spreadsheets, Word files, PDFs, emails. AI can read documents, and it is getting better at it. But when information matters, such as a client’s year end, a service’s scope or who approved a decision, documents have weaknesses:
- The same fact appears in several places, often inconsistently.
- It is hard to tell which version is current.
- Access is all or nothing: you share the whole file or none of it.
- AI has to interpret prose each time, and can interpret it differently.
Structured data holds the same facts as records with defined fields: a client with a name, company number, year end, manager and status; a service with a description, sectors and price basis.
Why structure helps AI
- Precise retrieval. “Clients with a March year end whose records are outstanding” is a query, not a search through documents.
- Consistency. One record, one value, used everywhere.
- Currency. Update the record once, and every use reflects it.
- Relationships. Records can link: this contact belongs to this client; this task relates to this project.
- Field-level access. A client’s contact details can be shared while its risk rating stays internal.
- Actions. AI can update a status or create a task because the record defines what those are.
What to structure first
You do not need to structure everything. Good candidates are facts that are:
- Referred to often, such as clients, projects, services and locations.
- Needed by several audiences, such as staff, clients and the public.
- Changing regularly, such as status, deadlines and outstanding items.
- Currently duplicated across spreadsheets and systems.
A simple approach
- List the kinds of thing your organisation works with: clients, matters, projects, services.
- For each, list the fields that matter.
- Import what you already have. A client list in a spreadsheet is a good start, but it should become records, not remain an uploaded file.
- Decide which system is authoritative for each field. Some data belongs in your accounting or CRM system and should be connected rather than copied.
- Decide who may see which fields.
- Connect AI to the records rather than to copies.
Public company data
Structured data also helps outside the organisation. Company facts, services and locations held as records can be published consistently to your website, to structured data markup, to directories and to AI agents that query them.
How Kroy approaches it
KroyDB stores structured entities, such as clients, projects, services, articles and decisions, together with their relationships. Store what belongs in Kroy. Connect what belongs elsewhere. Projections decide which fields each audience receives, and Kroy Publish serves the public portion to your website, API and AI-facing endpoints.