How the work originally ran
More than one hundred workers reported their output in WeChat groups. Two clerks then counted, checked and consolidated the messages.
The process appeared to work: workers reported production and the clerks produced a spreadsheet. The hidden cost was that the same data had to be rebuilt every day and reshaped again for operations, management, and finance.
The company did not begin by asking for a reporting system. The need became visible only after watching the day-to-day work.
Why it had remained unchanged
WeChat was already familiar to the workers. A new system would introduce accounts, training, devices, and new operating rules. Any of those could make digitization feel like extra work for the people on the shop floor.
Operations, finance, and engineering also saw different problems. Operations cared about valid reports and exceptions. Finance cared about counting rules. Engineering cared about the consistency of the input. No one was responsible for reconciling those different views.
The key decision: keep WeChat as the input
The first version kept the WeChat entry point. After an operator triggered the workflow, it:
- identified the production-report messages;
- organized them by person, product, and date;
- flagged uncertain or unusual messages for human review;
- aggregated the records according to operations and finance rules.
Workers did not have to learn a new entry point before the team could test whether the daily report was useful.
Who had to be involved
I asked each group to confirm a specific set of rules:
- Operations: what counts as a valid report and an exception;
- Finance: which fields matter and how they should be counted;
- Engineering: how inconsistent the messages can be and when human judgment is required;
- Operators: when to trigger the workflow and how to review an error.
Saying “digitization matters” is not enough to win cooperation. Each person needs to see which repeated task will become easier and exactly what they are being asked to confirm.
Where the project stood
The first version produced the daily report. The same dataset was then used for:
- machine capacity;
- daily exceptions;
- reports used by finance.
The workflow still required human review; it was not fully autonomous.
Confirmed / not yet measured
Confirmed: more than one hundred workers kept using WeChat; two clerks had performed the line-by-line consolidation; the first version produced a working daily report; the same dataset was then used for capacity, exception, and finance reports.
Not yet measured: there is no complete public before-and-after study, so I do not publish a labor-saving or efficiency percentage, and I do not describe the two clerks as “two people replaced.”
Related method: what I look for when nobody has asked for a system.