What context actually consists of
- Identity: which customer, which property, which job.
- History: prior visits, prior quotes, prior complaints.
- Commitments: what was promised, by whom, and when.
- Constraints: access, availability, budget, warranty status.
- Financial position: what is owed, what is paid, what is disputed.
Why the same message means different things
'Can someone come tomorrow?' from a first-time caller is a lead. The same sentence from a customer with a job finished last week and a balance outstanding is a warranty question with a collections implication.
The words are identical. Only context separates a sale from a liability.
Why pasting context into a chat box does not work
Manually assembled context is partial by definition — a person chooses what to include, and they choose based on what they already believe the answer is.
It is also stale the moment it is pasted, and it leaves no record. Context has to be retrieved live from the system that holds the truth, at the moment the question is asked.
How context is built structurally
Through relationships. One customer record links to properties, jobs, appointments, documents, messages and payments. Retrieving the customer retrieves the neighbourhood of facts around them.
Flat systems cannot do this. If messages live in one place and jobs in another with no shared identifier, there is no path from a question to its context.
What good context enables
- Answers that account for history rather than restating a field.
- Automations that skip customers where the action would be wrong.
- Pricing that reflects what this customer was quoted before.
- Escalation that recognises a repeat problem on the first mention.