Calculating automation ROI before you build

Automation ROI is a simple sum done honestly: (hours saved × loaded hourly cost, plus any revenue or error-reduction gains) minus (build cost, subscriptions, and — the part everyone omits — ongoing maintenance). Run that sum before building and insist on payback within about six months for a small firm. Most bad automation projects would never have started if this arithmetic had been done on one sheet of paper first.

Why calculate at all — isn't automation obviously worth it?

No, and the assumption that it is drives most of the waste. Automation is an investment with a cost curve like any other: money and attention up front, a monthly subscription tail, and a maintenance obligation that lasts as long as the workflow does. Some automations return their cost in weeks; plenty never return it at all, because they automated a task that was cheap, rare, or about to change anyway.

The reason to calculate is not precision — the inputs are estimates — it is discipline. The sum forces you to name the task, count its frequency, and price the upkeep, which is exactly the specificity that hype avoids. This is the working method behind everything in AI Automation for B2B: what actually works: treat AI spend like any other capital allocation, not like a lottery ticket.

What belongs on the cost side?

Four lines, and the last two are the ones that get missed:

• Build. Your time or a builder's fee. Internal time counts at loaded cost — a founder's week is not free.
• Tools. The workflow platform, API credits, enrichment subscriptions — the monthly tail, annualised.
• Maintenance. Workflows rot: APIs change, fields get renamed, connections expire. Budget a few hours a month across a serious workflow set, priced at whoever fixes them.
• Attention. Someone must monitor error logs and own failures. Small per workflow, real in aggregate.

A useful rough rule: over three years, expect the running and maintenance cost of an automation to rival its build cost. If the case only works with maintenance at zero, there is no case.

How do you value the time saved honestly?

By frequency, duration, and — crucially — whether the freed hours actually go anywhere. The mechanism, worked through: when a task takes 20 minutes and runs five times a week, then it consumes roughly 85 hours a year. When the person doing it costs £40 an hour loaded, then the gross saving is about £3,400 a year. When the build costs £1,500 with £50 a month in tooling and a couple of maintenance hours a quarter, then year-one cost is roughly £2,300 — and payback lands around eight months. Marginal, honestly: worth doing if the freed hours become selling or delivery time, not worth doing if they evaporate into the inbox.

That last clause is the audit point most calculations skip. Time saved only has value if it is redeployed. "The team saves an hour a day" is a cost line erased only when that hour turns into billable work, pipeline activity, or a hire deferred.

What gets missed on the benefit side?

Three things that often outweigh the hours. Speed: an enquiry answered in two minutes converts at a different rate from one answered in four hours — automations that compress response time carry revenue, not just savings. Consistency: a follow-up sequence that always runs beats a human who mostly remembers; the deals rescued are a benefit line. Error reduction: mis-keyed data and missed handoffs have real costs, and deterministic workflows do not have bad Fridays.

Counterintuitively, this is why the dull automations score best. Content pipelines of the kind I describe in AI-assisted content without losing your voice save real hours; but enquiry routing and follow-up usually beat them on ROI because they touch revenue directly. The showy projects — the ones fashionable on the conference circuit critiqued in The hype curve and the boring middle where money is — tend to sit at the bottom of the table.

What threshold should trigger a build?

Pre-decide it, once: payback inside six months, build; six to twelve, build only if it removes a bottleneck or a compliance risk; beyond twelve, decline and revisit in a year when tooling is cheaper. Writing the rule down matters as much as the rule itself — it converts every future "should we automate this?" from a debate into a lookup, which is precisely the overwhelm-reduction argument of Decision rules: pre-deciding your way out of overwhelm.

One sheet of paper, five lines of arithmetic, one pre-agreed threshold. It is not sophisticated. It is merely the difference between firms whose automation portfolio compounds and firms with forty workflows nobody can explain.

---

Next step: the Growth System Audit — £450, seven days, credited against any build — maps where your growth system leaks and what to build first.