Find the repeatable task before choosing the tool
A report may depend on someone downloading files, correcting dates, matching identifiers and emailing the result. If that person is away, the process stops. In another team, the problem may be copying enquiries between an inbox and a tracking sheet. Both are operational tasks before they are software choices.
We identify the input, the rules, the person responsible and the output another person actually uses. If the rules keep changing, the first step may be a simpler template or an agreed procedure. Automation is useful when it makes a known process repeatable, not when it hides an unresolved decision.
An example report workflow, honestly labelled
Excel Bot is a demo project, not a measured client case study. It demonstrates converting an untidy sales export into a structured Excel report: date and amount normalisation, duplicate removal, a revenue chart and breakdowns by region and product. No time saving or revenue outcome is claimed here.
A comparable scoped workflow could read an approved export, check required columns, flag questionable records and produce a report for review. A missing column or conflicting identifier should produce a visible exception rather than a plausible but unreliable total. Your own sample data determines what needs to be tested.
Choose a small script, integration or application
Automation starts from €149; data collection and preparation from €190. These are starting prices for defined tasks, not unlimited data sources or ongoing operation. An AI integration starts from €390 where a model has a useful role, with approved material and human review when needed.
If several people need roles, shared records and status history, a custom web-app MVP starts from €2,490. We compare the options against the workflow instead of assuming a full application is required. Existing subscriptions, APIs, hosting, model usage and maintenance are listed separately; access to external systems must be available.
A useful handover includes the awkward cases
Send an anonymised input file, an example of the finished output, how often the task runs and who relies on it. Explain which errors would matter and where a person should make the final decision. Include the current tools, access limitations and any deadline that affects scope.
We agree acceptance examples before building: correct totals, expected fields, handling of duplicates and clear behaviour when a file or connection is missing. At handover, you receive the agreed files or access, operating instructions and a support arrangement. A scheduled process needs an available computer or server; a powered-off host cannot run a weekly job.
Practical questions
Can we keep using Excel or Google Sheets?
Often, yes. The output can remain in a familiar format. We first check file structure, access and whether simultaneous editing or growth makes a shared application more suitable.
Will AI automatically make the process better?
No. Fixed calculations and structured transformations often need deterministic rules. AI is considered for tasks such as drafting or interpreting text, with explicit review and failure handling.
What happens when the input format changes?
The process should flag a mismatch rather than silently guess. Updating column rules, adding new sources or adapting to a changed external service is maintenance unless already included in the quote.
Discuss your project
Describe the workflow you want to improve. We agree the first version, deliverables and price before development.
Discuss your project