Engineering

The documented process and the real process are different. The gap between them is where nearly all the value hides.

Published
18 March 2026
Reading time
5 min read
Written by
CITS Engineering

Ask a team to describe their process and you get the version from onboarding. Watch the same team for two days and you find the spreadsheet that reconciles two systems, the WhatsApp group where approvals actually happen, and the person who knows which of the three date fields is correct.

Automate the real process

Automating the documented process produces a system nobody uses, because it does not handle the cases the workarounds existed for. Automating the real process produces something people adopt immediately; they have been asking for it, informally, for years.

Start with the highest-volume path

Not the most complex, not the most annoying: the most frequent. It generates the clearest measurement, builds trust fastest, and makes the business case for the next workflow without a slide deck.

Design the exception queue first

Every process has cases that do not fit. The failure mode of bad automation is not stopping; it is proceeding confidently with a wrong result. A clean human queue with full context, and loud alerting when volume rises, is what separates automation people trust from automation people work around.

Automation you cannot observe is automation you cannot trust, and it will be quietly bypassed within a quarter.

Say the quiet part early

People assume automation means redundancy. In our projects it consistently removes the parts of jobs people dislike (rekeying, chasing, reconciling) and the recovered hours get redeployed. Saying that openly in the first workshop does more for adoption than any feature.

CECITS EngineeringAutomation Practice
  • Automation
  • Process
  • Integration
Working On This?

If this article describes a problem sitting on your desk right now, the fastest route is a thirty-minute conversation with the people who wrote it.