AI Automation

A practical framework for calculating AI automation ROI

Most automation business cases collapse under scrutiny because they count time saved rather than cost removed. Here is the model we use with clients.

8 July 2026 · 8 min read

Time saved is not money saved

Most automation business cases open with hours reclaimed. Hours only become money when a role is not backfilled, a shift is not added, or the same headcount absorbs materially more volume. Start the model with the cost line you expect to change, then work backwards to the process.

Model four numbers, not forty

Volume of the task per month, fully-loaded cost per unit today, expected automation containment rate, and cost per automated unit including model, telephony and maintenance. Those four numbers produce a defensible payback period. Everything else is commentary.

Price the failure path

Every automated workflow escalates some share of cases to a human, and escalated cases cost more than they did before because they arrive later and warmer. Include that penalty explicitly at the containment rate you actually expect in month one, not the rate you hope for in month six.

Instrument before you deploy

If the baseline is not measured, the improvement cannot be proven and the programme loses its budget at the first review. We insist on two to four weeks of baseline instrumentation before a single agent goes live.

Let's map your highest-value opportunity

Book a 30-minute consultation. We will review your current systems and funnels and tell you honestly where the fastest return sits.