Field guide
Measure what changed in the business.
AI usage is easy to count. Business value takes more care. These conversations suggest starting with one workflow, comparing it with the old way of working, and separating faster output from a result the organization actually values.
Measure AI value with a clear baseline, actual workflow outcomes, review effort, implementation costs, and business impact.
Set a baseline before the experiment
Jim Spignardo connects adoption to high-value use cases and real ROI. Eddie Irvin puts process clarity before tool choice. For one workflow, record the work volume, current time and cost, error or rework rate, and the outcome it supports before changing the process.
Hear the conversations: Jim Spignardo · Eddie Irvin
Count the whole new workflow
An AI tool may draft quickly while a person spends more time checking, correcting, or moving the output. Ronnie Kwesi Coleman warns that productivity by itself can miss the business outcome. Include human review, integration, maintenance, and exceptions in the after-state.
Hear the conversations: Ronnie Kwesi Coleman
Separate capacity from cash
Time returned to a team is valuable capacity, but it is not automatically a cash saving. Report separately whether the change releases capacity, lowers actual cost, improves quality or speed, protects revenue, or creates new revenue. Denis Romanovskiy's focus on economic effect and Markus Schmidberger's focus on agent outcomes both point beyond usage metrics.
Hear the conversations: Denis Romanovskiy · Markus Schmidberger
Make the decision explicit
Compare the observed change against the cost of getting there, then decide to expand, redesign, or stop. Darren Ward's adoption work ties projects to business priorities; that link is what makes the measurement worth acting on.
Hear the conversations: Darren Ward
Questions answered
What leaders ask next.
What should an AI ROI calculation include?
Include the baseline and observed business outcome, plus software, implementation, human review, integration, and maintenance costs. Report qualitative benefits separately when they cannot be measured credibly yet.
Hear the conversations: Ronnie Kwesi Coleman · Denis Romanovskiy
Is time saved the same as money saved?
No. Time saved may create capacity or faster service. A cash saving requires an actual reduction in spending or a clearly attributable financial effect.
Hear the conversations: Denis Romanovskiy · Markus Schmidberger
Are tool logins a useful success metric?
They can show activity, but they do not show that a workflow improved or that the business gained value. Pair adoption activity with a real outcome.
Hear the conversations: Jim Spignardo · Markus Schmidberger
When should leaders stop an AI use case?
Stop or redesign when the measured benefit remains smaller than the full operating cost, the review burden is too high, or the workflow does not serve a priority.
Hear the conversations: Ronnie Kwesi Coleman · Eddie Irvin
Keep reading
Explore the related topics.
Hear the next conversation.
Get new episodes and occasional useful notes from Using AI at Work.
Follow the show →