What should you automate first?
Which is worth automating first: the thirty-second approval done two hundred times a month, or the hour-long report done once? Most people's gut says the report, because it hurts more each time it comes around. The maths says otherwise, and the maths is right.
Automate first whatever scores highest once you multiply how often a task happens by how long it genuinely takes, not by which one gets complained about most. Past that ranking, a few extra filters narrow it further: the task should be rule-based rather than judgment-based, and it should already sit inside a process that's been mapped and is reasonably well understood, not a process still full of unexplained steps.
Automation vs. outsourcing vs. hiring: how to choose
Automation isn't the only option for a task that's costing you time, and it isn't always the right one.
| Factor | Automate | Outsource | Hire |
|---|---|---|---|
| Best fit | Repetitive, rule-based, high-volume | Needs a skill your team lacks, not often enough to justify a hire | Ongoing, judgment-heavy, core to the business |
| Upfront cost | Setup time and tooling | Usually low, pay-as-you-go | Recruiting, onboarding, salary |
| Ongoing cost | Maintenance as the process changes | Per-project or retainer fees | Ongoing salary, regardless of volume |
| Risk if wrong | Locks in a bad process, or removes a needed judgment call | Dependency on an external party's quality and availability | Cost persists even if the volume of work drops |
Most growing businesses end up using all three, automation for the rule-based volume work, outsourcing for occasional specialist needs, hiring for judgment-heavy, ongoing responsibility. The mistake is defaulting to one option for everything.
AI shifts this comparison in one specific way. A task that used to sit in the "outsource, because the input's too messy for simple rules" column sometimes moves into "automate" once AI can handle the messy part. Re-check anything you've previously written off as unautomatable. That verdict isn't permanent.
When automation isn't worth the investment
A few honest signals that automation is the wrong call, at least for now:
- The task is low-volume. If it happens a handful of times a year, the setup and maintenance cost rarely earns itself back.
- The rules keep changing. A process that's redesigned or reprioritised every few weeks is expensive to keep automated, since every change means updating the automation too.
- The judgment call is the point. If the value of the task is a person weighing up context a rule can't capture, automating it removes the thing that made it worth doing well.
- It's a control, not a chore. A deliberate check that exists to catch a costly mistake is protecting something. Automating it away because it "feels slow" can quietly remove the protection along with the friction.
How much automation does a small business actually need?
Enough to remove the specific tasks that are genuinely costing time, money, or accuracy, and not more than that. It's tempting to treat automation as a general modernisation project once the first one or two wins land, but automation for its own sake tends to produce a pile of disconnected tools rather than a meaningfully more efficient business. A useful check: can you name the specific cost each automation is removing? If not, it's worth pausing before adding another one.
A simple decision sequence
If you want one path to walk through rather than four separate criteria, this is the order that matters:
- Does it happen often enough to matter? If it's a handful of times a year, stop here, leave it alone regardless of what else is true.
- Can the decision be written down as a rule? If not, it needs a person, whether that's someone already on your team, a new hire, or an outsourced specialist.
- Is it a control, not a chore? If the step exists to catch a costly mistake, keep the judgment with a person, even if the paperwork around it gets automated.
- Do you have the skill in-house, and is it frequent enough to justify a hire? If you have it, automate it or keep doing it manually. If you don't and it's frequent, hire. If you don't and it's occasional, outsource.
Most tasks get filtered out well before step 4, which is the point: the sequence is designed to stop you early, rather than talk you into automating everything that's left standing.