01 / ACCESS
AI is making business analysis available to more people, in plain language.
OpenAI's September 10 Data agent announcement shows the direction of travel: employees can ask questions about company data, investigate changes, and build dashboards without writing database queries. The product is aimed at larger data systems today, but the operating implication reaches smaller companies too. The distance between a business question and an analysis is shrinking.
02 / DEFINITIONS
Faster analysis does not resolve disagreement about what a metric means.
The same announcement emphasizes business terms, metric definitions, custom calculations, and trusted sources. That is the important detail. If sales means booked work in one spreadsheet, sent invoices in another, and collected cash in a third, AI can produce a polished answer to three different questions. The model is not the first problem; the business has not chosen a common definition.
03 / EVIDENCE
Traceability and access control are becoming part of everyday analysis.
OpenAI's new financial-services product highlights citations that let users trace figures and claims to source material, along with administrator-controlled data access. Separately, Anthropic's September 9 assessment of cybersecurity evaluation incidents reinforces why connected systems require carefully bounded authorization. For a small business, the practical standard is modest but firm: know which source produced the number, who may see it, and what must be checked before a decision is made.
THE CJC VIEW
Implementation beats experimentation.
AI can lower the cost of asking a business question. It cannot remove management's responsibility to define the question and maintain the underlying records.
Small businesses do not need an enterprise data program before using AI. They do need a short list of numbers everyone interprets the same way. A clear definition, one primary source, a named owner, and a visible last-updated date will improve both AI-assisted analysis and ordinary management conversations.
PRACTICAL NEXT STEP
Create a one-page definition card for one important metric
- Choose one number your team uses to make a weekly decision, such as open estimates, gross margin, overdue invoices, response time, or scheduled backlog.
- Write one sentence defining exactly what is included and excluded.
- Name the primary system or file, the person responsible for it, and how often it should be updated.
- Add one check that would reveal a bad answer, such as reconciling the total to invoices or reviewing five underlying records.
- Give the definition card and a current export to AI, ask one decision-focused question, and verify every material claim against the source before acting.
The useful breakthrough is not that anyone can generate a dashboard. It is that a well-run business can make its definitions, evidence, and decisions easier to inspect. AI becomes more valuable after that discipline is in place.