This took no data science team, no BI software, and no six-week consultant project. It took raw Excel exports from ERP, an AI conversation, and a BD manager who knew which questions were worth asking. In one session, a Southeast Asian CNC factory serving multinational corporations went from an unreadable spreadsheet to a growth strategy.
The data was 334,582 rows of production data, 56,835 lots, and RM 54.2 million in 2025 revenue. All of it sat in the ERP as exportable files nobody had analysed beyond monthly totals. One question mattered most: where is the concentration risk? The answer changed the BD team's priorities for the quarter.
The Session
The BD manager exported three raw Excel files: production orders, shipment performance, and customer accounts. He asked the AI to summarise revenue by customer, revenue by part family, delivery performance, and concentration risk. It flagged two data quality issues in four minutes, a job that normally takes a full morning.
The top three accounts represented 71% of total revenue. The largest single account alone was 43%, a number that had never been calculated before. Management knew the account was large, just not that large.
A follow-up question found three part families with rejection rates more than double the factory average, two of them inside that top account, sharing two machines that were already due for maintenance, a link nobody had made before.
What Changed
Right away, the BD team reprioritised its pipeline toward a segment where it could win accounts to dilute the concentration. It moved a pricing review forward for two accounts whose volume had grown without matching revenue. Production also moved scheduled maintenance up six weeks on the high-rejection machines.
What took skill
The technical part, uploading files and running the analysis, is easy. The hard part is knowing which questions to ask. Those questions came from a BD manager who knew what he was worried about and tested it against real data.
The AI doesn't know your factory, or that your top account is mid procurement-review. You know that, and the AI can find the pattern in 334,000 rows in the time it takes to ask.
Before Your First Factory Data Analysis Session
- Export production, delivery, and customer data from ERP, raw is fine
- Write down your five most important business questions before opening the AI tool
- Start broad: let the AI orient on the data before asking for specific analysis
- Ask about data quality issues before drawing conclusions from the numbers
- Sequence questions so each answer informs the next question
- Document the findings immediately
- Turn at least one finding into a concrete change within the same week
