ROI
Whether the thing paid for itself. Usually measured badly, in both directions.
Return on investment is the ratio of what a thing produced to what it cost. For AI projects it is notoriously hard to measure, in both directions: teams overstate savings by counting the demo, and sceptics overstate cost by ignoring the work the tool removes.
A workable approach is to measure one workflow, before and after, on time spent and on error rate. Not headcount saved, which almost never materialises, but minutes per task and rework rate. Those two numbers are observable and they move.
The most honest ROI number also includes the cost people forget: evaluation, maintenance, and the review time. A tool that saves an hour a week but needs an hour of review is not a saving, and finding that out early changes the decision.
Related terms
AI Adoption
The organisational work of getting people to actually use a tool that works.
Cost per Token
What you pay for each unit of text the model reads or writes.
Evaluation
Measuring whether the system does the job, on a fixed set of cases, repeatedly.
Agentic Workflow
A process where an agent does a meaningful part of the work, not just answers a question about it.
Training
The expensive process that sets a model's parameters by exposing it to data.
Want this working inside your team?
It starts with one conversation about your workflow.