How to put a dollar figure on the busywork that nobody's measuring — and which ones are actually worth automating.
Manual work is the most expensive thing in your business that never shows up on a P&L. It doesn't get a line item. It doesn't get a budget review. It just quietly eats hours, attention, and morale until somebody finally says 'this is ridiculous' — usually right before they quit.
The simple formula
You don't need a consultant to put a number on it. For any recurring task, the annual cost is:
“(minutes per run) × (runs per week) × 52 ÷ 60 × (loaded hourly cost) = annual $ cost”
Loaded cost is salary plus benefits and overhead — usually 1.3× to 1.4× the base hourly rate. For a $70K coordinator, that's roughly $46/hour loaded.
A worked example
Take a real one we see constantly: an ops coordinator manually pulls a weekly status report by copying numbers from three dashboards into a slide deck.
- Time per run: 75 minutes
- Runs per week: 1
- Loaded cost: $46/hour
- Annual cost: 75 × 1 × 52 ÷ 60 × $46 = $2,990
Three thousand dollars a year for one slide deck. Multiply that by the 8–15 recurring 'small' tasks every SMB has, and you're easily into six figures of invisible cost.
But the dollar cost isn't the worst part
The real cost shows up in three places that don't make it into spreadsheets:
- 1Error rate. Manual data entry runs about 1% error in the best conditions. A 1% error on invoices, inventory, or payroll is not a rounding problem — it's a customer service incident.
- 2Switching cost. Every interruption to 'just pull that number real quick' costs about 23 minutes of refocus time. Most knowledge workers lose more time to switching than to the task itself.
- 3Talent drag. The people you most want to keep are the ones who hate doing the busywork the most. Manual processes are a quiet resignation factory.
Which ones are actually worth automating?
Not all of them. Use a simple two-by-two: frequency on one axis, decision complexity on the other.
- High frequency, low complexity: automate first. This is the highest ROI quadrant — recurring reports, data syncs, status updates, simple approvals.
- High frequency, high complexity: automate the parts that are mechanical and leave a human in the loop for the judgment call. This is where AI assist actually earns its keep.
- Low frequency, low complexity: write it down, leave it alone. Automating something that happens twice a year is a hobby, not a return.
- Low frequency, high complexity: don't touch it. These are the cases where institutional knowledge matters most — automate the inputs to them instead.
Once you have those five numbers, the conversation changes. It's no longer 'should we look into automation?' — it's 'we're spending $X a year on this; what's the cheapest way to stop?'
