What is business intelligence?

Business intelligence gives record-based answers to the questions a manager asks every week: what is happening, why, and what should we do now? At a company that sells machinery and services it, those questions look like this:

All of these questions touch the same customer: the quote, the installed unit, the service visit and the invoice are parts of one customer. BI becomes more useful the more of those parts it can show in one picture.

Classic BI works in four steps: it collects data, combines it, defines the measures and shows the result as reports or dashboards.

• Data collection. Data is pulled from the sales tool, service forms, accounting software and spreadsheets. This step is called ETL (extract, transform, load).
• Data warehouse. Records from different programs are combined in one database. Two spellings of the same customer in two programs are matched here.
• Semantic layer. Concepts such as "customer", "won quote" and "overdue maintenance", and the formula behind each measure, are defined once. More in what is a semantic layer?
• Reports and dashboards. Managers see the measures as charts and tables, filter them and compare periods.

Most of the time in a BI project goes into the first two steps. Because the data is born in other programs, it has to be moved and then matched, and every new program adds another link to the chain.

Reporting shows what happened, BI explains why, and an AI agent prepares the work that comes next.

The three stack on top of each other: the agent reads BI's measures, and BI uses the records behind the reports.

AI is moving BI's interface from dashboards to conversation and its role from watching to acting. Instead of building a filter, a manager asks in everyday language: "Which customers have units coming off warranty this quarter?" The agent turns the question into a query on the records, produces the list and drafts a maintenance contract quote for each customer.

The largest vendors are moving the same way. At Tableau Conference 2026, Salesforce repositioned Tableau as an agentic analytics platform: a knowledge engine that combines metrics, relationships, business rules and definitions, and a decision engine that triggers workflows. Analyst firm ISG expects that by 2028 more than half of enterprises will use conversational experiences as their standard BI interface.

This shift has one condition: the AI has to know the company's concepts and how its records connect. However easy the chat window is, the quality of the answer depends on the quality and meaning of the records.

BI is most reliable when it runs in the system where the data is born, because a record's type, owner and links to other records are known the moment it is created.

If the quote lives in one program, the installed unit in a spreadsheet, the service form on paper and the invoice in accounting software, BI has to join these four pieces afterwards. When the same work runs on one platform, the chain gets shorter:

• Accepted quote lines become equipment records with serial numbers, so each unit shows which quote it came from.
• Work orders are opened from the equipment record; the service report links to the work order and the parts used link to stock.
• The invoice is issued from the completed service report, and the payment is posted against the invoice.

Because these links sit in the records themselves, the answer to "were the maintenance quotes we sent for units coming off warranty accepted?" is already in the records, and the measure is read straight from them. The AI agent reads what connects to what from the record too.

PureField is sales and business management software in which teams run the whole customer cycle, from sale to service, on one record and use AI agents to turn those records into decisions and work. Data is born in PureField, and BI runs in the same place.

• Dashboards. CRM statistics show the sales funnel and team performance; service statistics show equipment, service report and cost indicators.
• Copilot. Answers questions you ask in everyday language from your records. When a question needs your records, it opens a management meeting with the sales, finance or service directors; each director looks at the records in their own area, and every number they give rests on a record.
• Installed base revenue. Installed units are scanned every morning for warranty end, upcoming and overdue maintenance, consumables and spare parts, maintenance contract renewal and end of life. Each opportunity is listed with an estimated value and a 0–100 score and can be turned into an opportunity or a draft quote.
• Digital employees. Watch work such as collections, quote follow-up and service planning; by default they prepare a suggestion and wait for your approval.
• Approval rules. Deletes, missing fields and content read from outside sources go to approval, so the decision stays with you.

Copilot and digital employees run on PAIOS (PureField AI Operating System). See which agent works in which module on the AI agents page.