Building-products manufacturer
≈ $400M revenue (approx.) · A dozen brands · plants in three countries
The report that writes itself, across three ERPs
Monthly operating reports assembled by agents from three ERPs that never agreed with each other, with every figure traceable to its source system, so leadership reads one report instead of reconciling three.
- Organization
- Building-products manufacturer, private-equity owned, several plants
- Systems
- Three ERPs from three eras, one corporate spreadsheet
- Runs on
- The manufacturer's cloud tenancy
- Scope
- Monthly operating reports assembled and narrated by agents
Where they started.
Three plants, three ERPs, and a month-end that was a person copying numbers into a corporate workbook and explaining the differences by email. Every figure had a story, and the story lived with whoever assembled it. New ownership wanted a report it could trust without a call to ask what the numbers meant.
What was built.
- Governed connectors into each ERP, read-only, with the mapping between plants' units, cost centers and product families held in one maintained model.
- Agents assemble the operating report on a schedule: pull, normalize, reconcile, and flag mismatches between systems rather than quietly picking one.
- A drafted narrative around the numbers, in the company's own format, for the controller to edit rather than write.
- Every figure carries a reference back to the source transaction set, so a question about a number is a click, not a call.
What changed.
- One monthly report instead of three reconciled by hand.
- Mismatches between systems surface as findings rather than getting averaged away.
- The controller edits a draft instead of building a workbook.
- A foundation for consolidating the ERPs later, without waiting for that project to read the numbers now.
Plant-level margins and pricing are the company's most sensitive numbers. Assembling them on private inference keeps the report inside the company that owns it.
The manufacturer's cloud tenancy
Components involved.
More like this.
Multi-brand industrial manufacturer
≈ $200M revenue across five brands (approx.) · 5 brands · 4 ERPs · 1 CRM
- Use Case:
- Consolidation, forecast and pricing intelligence across four ERPs
- Results:
- One governed lakehouse behind an AI agent: audit-grade consolidated financials that trace to the source transaction, forecast and pipeline by business unit, large-job margin watch, and plain-English inquiry over all of it, inside the company's own Azure tenant.
European logistics group
≈ $2B revenue · 6,000+ employees · 50+ countries
- Use Case:
- Quote control tower over forwarding systems
- Results:
- Quotes that track fuel prices, routing and political risk across the group's forwarding systems, assembled by agents over the company's own data lake and handed to the desk with the reasoning attached.
Sleep-therapy provider
≈ $60M revenue (approx.) · 40 clinics · one clinical reporting app
- Use Case:
- Natural-language clinical reports
- Results:
- Clinicians ask for patient cohorts in plain English and a constrained pipeline turns the question into validated filters, never free-form database queries, on a private GPU that costs a fraction of the hosted model it replaced.
Have one like it?
Every spotlight started as a conversation about a process nobody liked doing, under a data-locality constraint.