Area 02Process Automation
Recurring process automation.
I automate processes that repeat every week or month: reports, file consolidation, data loads and validations. Inputs, rules and outputs follow the same method every cycle.
- Same rules every cycle
- Checks before continuing
- Traceable from input to output
01The boundary
What gets automated. What stays human.
I automate the repetitive part of the process. Anything that requires judgment is not delegated to the workflow.
Automated
- Data preparationPull and organize the defined inputs.
- Rules & checksApply the same validations and calculations each cycle.
- Output generationPrepare the report, the consolidated file or the data load.
- Run history & recoveryRecord what ran and where to resume.
Human
- InterpretationExplain the results in context.
- Exception reviewResolve cases that need judgment.
- Final approvalReview commentary and authorize delivery.
- DecisionsChoose what changes and what to test next.
02A process I automated
A recurring report that stopped taking days to prepare.
Before≈ 4 days
After≈ 10 minutes
In one reporting workflow I automated, operational report preparation went from about four days to about ten minutes.
The figure measures preparation time, not analysis or approval. It is one case, not a general rule.
03What I build
The output and the flow behind it.
Not just this cycle’s result, but the documented process that prepares the next one.
- 01
Automated workflow
Recurring steps, rules and dependencies, documented and automated.
- 02
Quality checks & recovery
Coverage, freshness and validation rules that identify when a run can continue, stop or resume.
- 03
Review-ready output
A report, consolidated file, data load or presentation with consistent definitions and a visible version.
- 04
Run history & operating guide
Execution records, recovery instructions and documentation for operating the workflow.
04Try it
Run the next reporting cycle.
The example is a monthly process that ends in a report: from source checks to an output ready for the team. See what happens when a required source arrives incomplete, too.
Try it
Run the next reporting cycle.
The example is a monthly process that ends in a report: from source checks to an output ready for the team. See what happens when a required source arrives incomplete, too.
- 01Inputs
- 02Validate
- 03Calculate
- 04Build draft
- 05Review-ready
Paid Media / June 2026
One method. Every reporting cycle.
Choose an example. The workflow checks sources, applies controls and prepares a version for review.
- Check sources
- Apply checks
- Prepare draft
Simulation using synthetic June 2026 data. No live accounts are connected. Demo timing is illustrative.
Simulation using synthetic June 2026 data. No live accounts are connected. Demo timing is illustrative.
05How I build it
From manual process to repeatable flow.
A clear process. No black box.
- 01
Define the cycle
I set the inputs, period, rules and output format.
I document access, time zone, cutoff dates, owners and validation criteria. I automate one process at a time and extend the workflow once the first is stable.
- 02
Automate the process
I build the extraction, validation, calculation and output steps with Python, SQL and BigQuery.
Temporary failures can be retried. Missing or malformed required inputs stop affected outputs and create a visible issue. Credentials are managed outside the workflow.
- 03
Validate & document
I test normal runs and failures, and document how to recover and operate the workflow.
I reconcile against a reference cycle and test incomplete inputs and repeated runs without duplicate delivery. Each workflow has named owners for review, recovery and later changes.
06Labs
Labs in this area
Tools that show this area’s method, with fictional or sample data.
07Technical questions
How the workflow behaves.
The workflow defines which outputs can continue and which must wait. An incomplete required source stops the affected outputs and raises a visible issue; it never becomes an apparently valid output.
Each run is identified by process, period and version. Running the same cycle again preserves the existing output instead of creating another delivery.
From the last valid checkpoint. Completed steps are recorded, so fixing a source does not mean repeating the whole cycle.
Its inputs, rules and output are defined and repeat every cycle. Reports, file consolidations, data loads and validations usually qualify; ad hoc analysis does not.
Definitions are documented once and applied the same way every cycle: cutoff date, time zone, attribution windows and formulas. When a definition changes, the change is explicit and versioned.
08Contact
Interested in the topic?
If you work with processes that repeat every week or month, I’d like to compare how to automate them without delegating judgment.