Prioritising Automation Work
Last updated:
The loudest complaint is rarely the best candidate
The process people complain about most is frequently the exception handling — unpredictable, varied and hard to automate. The best candidate is usually something routine nobody mentions.
Ask what people do every day without thinking about it. That is where the automatable work is, and it is invisible precisely because it is routine.
Score each candidate
| Factor | Score high when |
|---|---|
| Frequency | Daily or more |
| Duration | Takes real time each occurrence |
| Error cost | Mistakes are expensive |
| Understood | Someone can describe every step |
| Wanted | Whoever does it would be relieved |
| Stable | The process is not about to change |
Anything scoring low on understood or stable should wait, regardless of how much time it consumes.
Do the arithmetic
- Occurrences per year multiplied by minutes each
- Divided by sixty, multiplied by hourly cost
- Plus the annual cost of errors, where countable
- Compared against build plus three years of running and maintenance
Prefer contained work
- One process, not a chain of five
- One source and one destination where possible
- Clear boundaries with what stays manual
- Something that can be delivered in weeks, not months
Contained automations are delivered, measured and trusted. Ambitious ones are delivered late and measured never.
Re-rank quarterly
The best candidate changes as processes change and as earlier automations remove work. Re-running the scoring quarterly keeps the effort aimed at what currently costs most.
It also catches the new manual process somebody invented last month.
Frequently asked questions
What if several score similarly?
Should we automate the biggest time sink?
How often should we re-rank?
What if nothing scores well?
Several candidates and no obvious answer?
Score them on those six factors. Happy to run through the list with you.
Related services
What we build for problems like this one