Automate this first: the frequency × time × error rule

Ask a business owner what they would automate and you get the task they hate most. Ask their bookkeeper and you get a different task. Ask the person who answers the phone and you get a third. All three are real, and none of them is necessarily the one to do first. The one to do first is the one with the best payback, and payback is arithmetic.

The rule

Frequency, times time, times error cost. How many times a week does the task happen? How long does each one take? And what does it cost when it is done wrong — a credit note, a lost customer, an afternoon unpicking it? Multiply the three and you have a number for every candidate. Rank by the number.

What the rule tends to find

• Enquiry to CRM to first reply: high frequency, short time each, high error cost when one is missed. Nearly always first.
• Quote to invoice to payment chase: medium frequency, long time each, and the errors cost cash. Usually second.
• Document generation — quotes, packs, reports from data you already hold: frequency varies, time per item is large. Often third.
• The task the owner hates: frequently low frequency and low error cost. It goes on the list; it rarely goes first.

What the rule leaves out

Two things. First, the cost of building the automation, which varies a lot: a two-system connection over supported APIs is days; a workflow that needs judgement in the middle may need a human step or a model, and costs more. Second, whether the process is worth keeping at all — the best automation is sometimes deleting the step. The ranked list is where those two conversations start, not where they end.

Where AI fits

Most of the top of the list is deterministic plumbing: this system tells that system, this event triggers that message. Plumbing is cheap and reliable and should be used wherever a rule does the job. AI earns its place where the input is unstructured — an email to triage, a document to read, a call to summarise — and where judgement has to be drafted for a person to approve. Put it there and nowhere else, and the economics hold.