Story 02
The process estimate that took too long to explain.
The answer was not in one file. Neither was the reasoning behind it.
The familiar moment
An engineer preparing a process estimate had to collect plant data, open lab workbooks, and reconstruct why earlier decisions had been made. The sources existed, but they lived in different places and changed on different schedules.
Before the engineer could explain the result, they first had to rebuild its history. Updating the answer later meant doing much of that work again.
What was really wrong
The calculation was only one part of the job. The team also needed to know which data was used, which method produced the result, and why two reasonable methods might disagree.
Why this came first
The team already had a lab calculation it trusted. That gave the new work a clear comparison point and kept engineering judgment in charge.
The tempting shortcut
Why the obvious AI answer was not enough.
Putting every spreadsheet into a chatbot would make the information easier to search, but it would not resolve differences between methods. A fluent answer could hide the exact disagreement an engineer needed to see.
What we built first
A useful workflow, small enough to prove.
We brought the sources together and compared one process estimate with the lab calculation the team already trusted.
Sources
01Name every input
The workflow records where each value came from, who owns the source, and when it was last updated.
Methods
02Keep calculations separate
Different methods remain visible instead of being blended into one unexplained answer.
Review
03Show the disagreement
The responsible engineer sees where estimates differ and decides whether the method is appropriate for the situation.
How we knew it was working
The estimate could be compared with lab results and past operating data. Differences between methods stayed visible for an engineer to review.
Where people stay in control
The lab calculation remains the reference. The engineer responsible for the process approves the method and its limits before anyone relies on it.
What improved
The workflow is designed to let engineers spend less time rebuilding the history and more time reviewing the result. The answer becomes easier to update, question, and explain to a client.
Where the first workflow led
- A clear list of each data source, who owns it, and when it was last updated
- Separate calculation methods so the team can see when they disagree
- Checks on the data and the method before a result is used
- A record showing which data and calculation produced each value
Look at your own business