Strategy6 min read
The processes we tell SMEs not to automate
Automation sells better than restraint, which is exactly why it's worth writing down where it doesn't pay.
Some processes cost more to automate than the problem is worth. Before recommending a build, we would check the volume, variation and cost of the work. Here is how we would assess it.
Low volume with high variation
A process that runs eight times a year and looks different every time is a bad candidate. Automation earns its keep through repetition. If a human has to make a judgement call on most instances anyway, you'll spend the build budget encoding exceptions and still route everything to a person.
The rough test we use: fewer than about fifty instances a year, or more than a third needing human judgement, and we'd rather write the process down properly than automate it.
Processes that exist because of an unresolved disagreement
Some workflows are elaborate because two departments never agreed who owns a decision, and the process is the truce. Automating it makes the truce permanent and much harder to renegotiate. Fix the disagreement first; the process usually shrinks to something not worth automating.
Anything about to be replaced
If the ERP is being replaced in nine months, don't automate around the old one. It is an easy way to spend real money integrating a system that gets decommissioned before the automation has paid back. Wait, or automate the part that survives the migration.
Where the real cost isn't the labour
- The keying takes twenty minutes but the decision before it takes two days. Automate the decision's inputs, not the keying.
- The delay is a supplier waiting to respond, not your team waiting to type.
- The error rate is caused by bad source data, and automation will just distribute the bad data faster.
What we do instead
Usually one of three things: document the process so it survives a person leaving, remove a step nobody can justify, or fix the data quality upstream so the manual work gets easier. None of these are billable in the way a build is, which is exactly why you should be suspicious of anyone who never recommends them.
The cheapest automation is the process you delete.