How Much Time Can Automation Save a 20-Person Team? (Worked Example)

How much does a three-minute task actually cost per year?

About £3,000 a year, per person, for a single task repeated twelve times a day. That is not a rounding error — it is a meaningful fraction of a junior salary, and it does not appear anywhere in your accounts.

Here is the arithmetic in full, because the number only lands when you can check it:

• 3 minutes per instance, 12 instances a day = 36 minutes daily
• 36 minutes × 5 days = 3 hours a week
• 3 hours × 46 working weeks (after holiday and bank holidays) = 138 hours a year
• 138 hours is roughly 3.5 working weeks — from one person, on one task
• At a £32,000 salary plus ~20% employer NI and overhead, the loaded cost is about £22 an hour
• 138 × £22 = £3,036 a year

Now scale it. If three people in a 20-person business each carry that same task, you are spending roughly £9,100 a year and losing 414 hours — about a quarter of a full-time role — on work no one chose to do.

Why doesn't this cost show up in your P&L?

Because it is already inside a salary you have decided to pay. Accounting recognises the wage; it does not recognise what the wage was spent doing.

This is the structural reason small repetitive work survives so long. A £9,000 software subscription gets a procurement conversation, a renewal review and a line item someone defends every year. A £9,100 pattern of re-typing addresses between two systems gets nothing, because no invoice ever arrives for it. It is paid in headcount, and headcount is a fixed cost you have stopped questioning.

The consequence is predictable. Volume grows, the three-minute tasks multiply, and the business hires — not because the work became more valuable, but because the friction consumed the capacity you already had. That is the hire you did not plan.

Which small tasks are worth automating first?

Rank them by annual minutes, not by how irritating they are. The tasks people complain about loudest are rarely the most expensive ones.

Run this over one week — it takes about twenty minutes of attention, spread out:

1. Monday, 5 minutes. Open a spreadsheet with four columns: Task, Minutes per instance, Instances per day, People who do it.
2. Tuesday to Thursday, 2 minutes a day. Each person adds any task they did more than twice that day which involved moving information between two places. That last clause matters — it filters out judgement work and leaves the mechanical work.
3. Friday, 10 minutes. Multiply: minutes × instances × people × 230 working days ÷ 60 = annual hours. Sort descending.

Look only at the top three. Everything below them is noise until the top three are gone.

The pattern that almost always tops the list is re-entry: the same fact typed into a second system because the first system cannot talk to it. Tenancy details from an enquiry form into the property manager. Candidate details from a job board into the ATS and then into payroll. A new client from the CRM into the accounting tool. Nobody designed this work. It accumulated.

What does it cost to fix?

For a single well-defined re-entry task, usually far less than the annual cost it removes — which is the only ratio worth arguing about.

Two honest caveats, because the ratio is not always favourable:

• Automation does not recover 100% of the time. Exceptions still need a human. A realistic target is 70–85% of the mechanical portion, and you should plan around the remainder rather than pretend it does not exist.
• If the underlying process is wrong, automating it makes the wrong thing faster. Sometimes the correct answer is to remove a step entirely, or fix the upstream form that is producing bad data — no automation required. I have recommended exactly that more than once, and it is a cheaper outcome for the business.

The test I apply before building anything: does this permanently remove the task, or does it just make a person's version of the task quicker? Only the first one compounds.

Frequently asked questions

How much time can automation realistically save a 20-person team?

In UK SMBs with typical tooling, mechanical data-transfer work commonly runs at 2–5 hours per person per week. Removing 70–85% of that across a 20-person team recovers the equivalent of one to two full-time roles a year. The variance depends almost entirely on how many systems your data crosses.

Is it worth automating a task that only takes three minutes?

Yes, if it repeats. Frequency, not duration, determines cost. A three-minute task done twelve times daily costs more per year than a two-hour task done monthly.

How do I calculate the cost of a manual task?

Minutes per instance × instances per day × people doing it × 230 working days ÷ 60 = annual hours. Multiply by loaded hourly cost (salary + roughly 20% for employer NI, pension and overhead, divided by about 1,700 productive hours).