Say the weekly defect report takes you 3 hours.
Write that number down. Before any tool, before any process map, before the clever AI dream. Without it you can't prove the automation did anything.
Getting there means spending this week being a detective in your own job, not an architect. Don't draw the perfect process map yet. Log it as is first, with all the ugly bits and the waiting in between.
Keep a pain diary for five days
Leave a spreadsheet or a notebook open. Add a row every time you touch a QA/QC task, such as inspections, ITP sign-offs, NCRs, snag lists, photos, test reports, weekly reports or chasing subbies.
For each row, note:
- what triggered it
- how many minutes it took
- how often it happens (daily, weekly, per pour)
- where the data came from and where it went
- who you had to chase
- how annoying it was, from 1 to 5
Do the maths on Friday
Multiply frequency by minutes to get hours per month, then sort.
Say chasing one NCR sign-off takes 10 minutes, four times a day. That's about 13 hours a month, nearly two working days, spent sending reminders.
Hours beat gut feeling. Annoyance level resolves ties.
Circle the hotspots
Flag every row where you retype data, copy between systems or chase a signature. Double entry is where the gold hides.
Split rules from judgment
Ask: "Could I write this as an if-then rule?" If yes, it's a candidate.
If it needs your trained eye on a crack in the slab, keep it human and automate the paperwork around it.
Pick one winner
Choose something frequent, rule-based and already digital. Classic first wins:
- the weekly defect report
- NCR routing
- overdue-defect reminders
Park the clever AI dream for now.
Measure, build small, measure again
Remember the number from the top. Build the smallest version, then measure again after a month. That before/after number buys you the budget for task number two.
Which row would you circle first?