An operations manager is asked to justify automating routine enquiries. She calculates two minutes per question, forty questions a day, and arrives at eighty minutes.
The number looks modest and the request is declined. Nothing about her arithmetic was wrong and it was measuring the wrong thing.
Those forty questions land on people doing work that requires concentration, and the reply is the smallest part of what each one costs.
Component 1: the reply itself
The visible cost. Two minutes to read, decide and answer. This is the only component anybody ever counts and it is frequently the smallest of the four.
Component 2: the switch out
Stopping, saving state, holding the current task in mind. Short, but it happens every time and it is where errors in the original task are introduced.
Component 3: the recovery
Getting back to where you were. For anything requiring real concentration this is far longer than the interruption, and it is entirely invisible on any timesheet.
Component 4: the residue
Part of the attention remains on the interrupted item, particularly where it was unresolved. The next twenty minutes of work happen at reduced quality rather than stopped.
A worked example
Forty interruptions a day across a team, at two minutes each, is eighty minutes of visible cost. Add ten minutes of recovery to each and it becomes seven hours and twenty minutes, which is very nearly a full person.
The same example at a cost
At a loaded cost of twenty five pounds an hour, eighty minutes is roughly thirty three pounds a day. Seven hours and twenty minutes is about a hundred and eighty three. Across a working year those are seven and a half thousand and forty two thousand.
What that comparison is actually for
To show that the decision about automating routine questions was being made on a fifth of the real number. Nothing about the automation changed, only the arithmetic used to judge it.
Not every interruption is equal
An interruption during routine administration costs little. One during quoting, diagnosis, writing or anything requiring held context costs the full amount, which is where the measurement should focus.
Measure it for a fortnight
Log interruptions with the task being done and an honest recovery estimate. Two weeks produces a number nobody can argue with, which is the method in mapping where repetitive enquiries actually come from.
Separate the ones requiring judgement
Something only that person can decide is a legitimate interruption. A question about opening hours is not, and the split is usually heavily towards the second.
Batch what remains
Answering in blocks rather than continuously removes most of the recovery cost, and it works only where somebody else is covering the immediate response.
Count the quality cost too
Estimates written while distracted, quotes with errors and the finish on interrupted work. These cost money in ways no time log captures, as in flooring quotes that stall at measuring.
Remember the same loss happens at handovers
Context lost between people is the same cost in a different form, which is the ground in handover between shifts and what gets lost.
Use the real number to decide
An assistant answering the routine questions removes the interruptions that carry no judgement, which the full arithmetic in the time cost of context switching and how chatbots sets out. See pricing, run your own figures on the chatbot ROI calculator, and the sector view sits under small business.