Skip to content

Strategy & Systems Design

Automation opportunity analysis: automate what pays back, not everything that repeats

Almost every step in a business repeats, so “what can we automate?” has a long answer. The useful question is which steps are worth it: frequent, slow and stable ones first; a person checking where a mistake costs money; and rare, risky or still-changing work left alone until it isn't. Every automation needs looking after, so each one has to earn its upkeep.
Talk through your listSee the four questions ↓
leave with a personautomate, with a checknot worth it yetautomate first1 h5 h10 h20 hhours a month it takes →highlowcost of a mistake, or upkeep →re-typing jobs into accounting · ~22 hreminder calls · ~14 hsorting photos to jobs · ~6 hestimate follow-ups · ~9 hprices change weeklyordering parts · ~4 hpricing unusual jobs · ~3 happroving large invoices · ~2 h

Four questions for every step

Asked of each step on the current-state map, with the people who do the work.

  1. 01

    How often?

    Every job, or twice a year

    A step that happens with every job is worth more attention than one that comes up at year-end.

  2. 02

    How long does it take?

    Including the waiting

    Minutes of typing, plus the time the next person waits for it.

  3. 03

    How stable is it?

    Same steps next month?

    A process that's still changing would have to be rebuilt every time it changes. Settle it first.

  4. 04

    What does a mistake cost?

    A typo, or a wrong price

    Where a wrong result costs money or trust, a person checks before it goes out, or the step stays with them.

Seven candidates, seven verdicts

The same seven steps as the plot, with the reason for each verdict.
Re-typing jobs into accountingevery job, the same fields each timeautomate first
Reminder callsevery day, and the wording never changesautomate first
Sorting photos to jobsevery job; each photo belongs to one jobautomate first, as part of close-out on site
Estimate follow-upsfrequent, but prices change weekly and a wrong price costsautomate the reminder; a person checks the price
Ordering partsa few hours, and the supplier's process changes monthlynot yet; look again when it settles
Pricing unusual jobsrare, and each one is differentleave with a person
Approving large invoicesrare, and it's a judgment callleave with a person, with everything they need ready

Everything repetitive, or what pays back

Automate everything repetitive

  1. 01Every repeated click is a candidate
  2. 02Rare tasks automated because they're annoying
  3. 03Processes that are still changing, frozen into automations
  4. 04Judgment calls handed to rules

The automations need more looking after than the work they replaced.

Automate what pays back

  1. 01Frequent, slow, stable steps first
  2. 02A person checks where mistakes cost
  3. 03Changing processes left until they settle
  4. 04Judgment stays with people, with the information ready

Fewer automations, each one earning its upkeep.

Where the verdicts come from, and where they go

In

Current-state map
every step, every tool
The people doing it
how often, how long
What a mistake costs
money, time, trust
Opportunity list
a verdict for every step

Out

Process automation
what's built first
AI opportunity analysis
steps that read language or photos
Roadmap
what goes in stage 1

Worth doing when the list is long and the budget isn't

Good fit

  • A long list of things you could automate, and no way to choose.
  • An automation you already have is now a chore to keep running.
  • You want the first project to pay back before committing to more.

Not the right fit

  • One obvious, frequent task. Just automate it.
  • The process itself is still being decided. Design it first.

Where this capability is used

Next to automation opportunity analysis

Questions about automation opportunity analysis

Where do the hours come from?

From the people who do the work, checked against a few real days where we can. They're estimates for ranking the steps, not a promise of time saved.

Why not automate everything that repeats?

Every automation has to be watched and updated when the process changes. A rare, risky or changing step can cost more to keep running than it saves.

What happens to steps that need judgment?

They stay with a person. Automation can still gather what they need and put it in front of them, so the decision is quicker.

How is this different from AI opportunity analysis?

This decides whether a step is worth changing at all. AI opportunity analysis looks at the steps worth changing and decides which need a model, and how far to trust it.