When a spreadsheet starts running the business, it is time to design the system behind it.

The short answer

The spreadsheet is rarely the real problem.

The real problem is operational dependence on files, manual reconciliation, knowledge held by a few people, and controls that nobody can see clearly. Replacing the file without understanding that operating system simply moves the fragility into new software.

Signs this is your situation

The workflow has outgrown the file that carries it.

Process discovery

Separate the business rules from the way people currently execute them.

The resulting system

Build the controls people need, not a generic spreadsheet replacement.

The application may combine validated data entry, deterministic calculations, role-based workflow, audit history, enterprise integrations, dashboards, exception handling, scheduled exchange, reports, and APIs. The exact shape follows the process rather than a preset technology stack.

Where AI belongs

Add AI after the workflow has a reliable foundation.

AI can interpret unstructured input, summarize evidence, identify anomalies, explain changes, or support natural-language questions. It should not own exact calculations or approval rules that conventional code can express and test.

How I approach the work

Discover the process, separate the rules, and replace it incrementally.

I begin with the people, files, handoffs, and exceptions. I turn that operating knowledge into a technical design, build the highest-value path in manageable increments, reconcile the output against the current process, and prepare the team to operate it.

Relevant proof

Manufacturing cost, sales, procurement, and operational analytics work

My work in these areas has required capturing business rules, reconciling inputs from multiple systems, making exceptions visible, and turning operational data into software people can inspect and maintain. Client details remain confidential; the relevant evidence is the method, not a technology list or an invented outcome.