AI for small businesses: 15 use cases on data you already have

Everyone wants to do something with AI, but what?

According to Statistics Netherlands (CBS), one in six Dutch businesses used AI in 2025, twice as many as two years earlier. Among businesses with 2 to 10 employees it was 14 percent, and among those with 10 to 50 employees 27 percent. So most small businesses have not started yet, and many owners who do have a ChatGPT subscription mainly use it for writing text and emails.

That is useful, but it is not where the biggest value lies. That lies in the questions you have about your own business every week and currently wait on until someone has made an export. How are we doing? Which projects are overrunning? Who is not paying? AI can answer those questions, provided it has access to your business data.

Below are fifteen practical use cases, grouped by department. Each one lists the data you need. You already have most of it.

Leadership and management

1. Ask questions in plain language. "How does third-quarter revenue compare with last year, by branch?" Instead of waiting for a report, you get an answer within a minute from your connected accounting system. Data: accounting.

2. The monthly figures, summarized. Let AI summarize the main variances against budget and last month in five sentences, with the underlying figures. Data: accounting + budget.

3. A weekly overview in your inbox. An automated report with the five KPIs you care about, every Monday morning. Data: accounting, CRM, project software.

Finance

4. Prioritize receivables. Which ten open invoices have the most impact on cash, and which customers consistently pay late? Data: open items. See also receivables management with data.

5. Flag cost anomalies. Which cost accounts are growing faster than revenue this quarter? Data: general ledger.

6. A first draft of the commentary. AI drafts the commentary for the monthly report; the controller checks and edits it. Read how to set this up safely in AI in finance and accounting.

Projects and operations

7. Projects that are going off track. Which projects have used more than 80 percent of their hours at less than 60 percent progress? Data: project software + time tracking.

8. Job costing by project type. Which types of work consistently miss their margin? See automating job costing.

9. Plan versus actual. Where do planned and worked hours differ most, by team or branch? Data: planning + time tracking.

Sales and marketing

10. Review the pipeline. Which deals have been in the same stage for more than 60 days? Data: CRM, such as HubSpot.

11. Customer value and churn. Which customers bought more last year than this year? Data: sales invoices.

12. Marketing to revenue. Which channels bring visitors who also become customers? Data: Google Analytics 4 + CRM + invoices.

HR and people

13. Absence and staffing. How is absence developing per department, and where does it coincide with overtime? Data: HR system such as Nmbrs.

14. Utilization per employee or team. Who is consistently below target, and is that due to planning or non-billable work? Data: time tracking.

The whole company

15. One help desk in Teams or Slack. Colleagues ask number questions to an AI assistant in Teams or Slack instead of the controller. Data: all connected sources, with permissions per user.

Curious which of these use cases work on your data? Request a 20-minute demo. We connect one source and ask the first questions of your own figures together.

Why your own data makes the difference

A general language model knows nothing about your business. You can paste an export into a chat window, but then you work with a snapshot, risk personal data leaving your organization and have to start over every time. An AI agent that works on synchronized business data calculates with current figures, uses fixed definitions and shows where an answer comes from. Read more about that difference in ChatGPT for business.

How to start

1. Pick one department and one recurring question. Not "something with AI", but "every Monday I want to know which projects are overrunning".
2. Connect the source software. Without access to data, AI can only give generic answers. See the available connectors.
3. Record the definitions. What is margin? When is a project overrunning? That prevents wrong answers.
4. Measure after four weeks. How much time does the question take now? Which decisions were made earlier? Use the AI pilot plan for the approach and the AI business case to calculate the value.
5. Train your team. Since February 2025, the EU AI Act requires organizations that deploy AI to ensure sufficient AI literacy among their staff. A short agreement on what is and is not allowed is part of that.

Looking for inspiration for the first questions? Read 50 questions to ask your business data.