Module 09 Summary

What this module established

Filter candidates on volume, time each, and whether the output can be checked. A task whose cases all differ is not a candidate however tedious it is - automate the routine slice and route the exceptions on purpose.

Carry forward

• Model the approval chain as it exists. Long pauses mean durable state, and multiple approvers must be recorded individually.
• Triage over real historical data before building. If 40% are exceptions, the value is much smaller than the proposal assumed.
• Compare including review time and both error rates. Review time usually dominates and the error rate often goes up.

Before moving on

Move on when your comparison states hours saved and errors added, and your automation boundary is written down with the condition for moving it.

Mark complete