The three hard metrics that actually matter
AI ROI doesn’t live on marketing multipliers. It lives on three numbers any finance lead understands:
- Time saved per unit of work. How long a human took to draft the email, classify the ticket or summarise the case, and how long it takes now, including human review when there is one. The sample must represent the process volume, variety and risk; there is no universal number of cases that fits every project.
- Cost per unit of work. Fully loaded cost of the process before and after: prorated salary, AI licences, infrastructure, maintenance and review hours. If cost per unit drops while volume stays the same or grows, the business case holds.
- Error rate. Share of outputs that need correction, retraction or rework. A faster AI with a higher error rate than the manual process isn’t savings: it’s hidden debt that surfaces later.
These three are measured before you start, the baseline, and measured again at agreed checkpoints. Weeks four, eight and twelve would be an illustrative pilot schedule, not a universal timeline. No baseline, no defensible ROI.
The soft metrics that don’t show up in the spreadsheet
Some improvements don’t fit in an Excel column but move the business anyway. Worth listing them explicitly so the case isn’t lopsided:
- Consistency. Compare quality by time, channel and season to test whether the agent reduces variance; do not assume it.
- Customer response speed. As an illustrative calculation, going from twenty-four hours to thirty seconds would be a large change, but value must be validated with the channel’s actual timings and satisfaction data.
- Freed-up team capacity. Hours only count as a benefit when measured and tied to an outcome: less overtime, more volume or higher-value work. You have to decide explicitly where they go.
- Team satisfaction. Removing repetitive work may improve the team experience. Treat that as a hypothesis and test it through surveys and retention data without attributing the whole change to AI.
How to build the business case without inflating ROI
An honest business case survives production and the first finance review. Three rules we recommend:
- Count the full cost, not just implementation. Model, integrations, observability, monthly maintenance, updates when the provider changes the model and internal governance hours. Counting only the up-front phase inflates ROI and hides later costs.
- Use ranges, not points. In a purely illustrative example, replacing “€120,000 savings/year” with “€80,000 to €140,000/year depending on adoption” makes uncertainty visible. The real amounts must come from the project baseline.
- Attribute carefully. If the process was already going to improve for other reasons, a new CRM, recent training, a shift change, don’t attribute the whole delta to AI. The clean portion is the one that holds up under scrutiny.
Common mistakes when measuring AI ROI
Four common risks, with no universal severity order:
- Not counting maintenance. The model changes, the API is updated, integrations break. If maintenance cost isn’t in the case, ROI in later periods will be inflated.
- Ignoring HITL cost. If a person reviews every output, that time is real cost. Add it in and only reduce the forecast when measurements justify doing so.
- Crediting all savings to AI. When the agent ships alongside a redesigned process, part of the saving comes from the redesign, not the model. Splitting the two avoids disappointment.
- Measuring only at the start. An early reading may capture the learning curve but not structural cost. Define a longer checkpoint suited to the process cycle as well.
The goal isn’t to present the highest number possible. It’s to present a number that still holds when someone reopens the spreadsheet over a horizon long enough to include operations and maintenance.
