Imagine a Monday board meeting. The IT director reports that a pilot has helped the team prepare proposals faster, recovering 150 hours a month. The CEO looks at the figures: “Why do we have no more money at the end of the month?”
Both can be right. The tool has improved the work. Payroll is unchanged, and salespeople cannot send more proposals because they are waiting for production estimates. One team has spare capacity; another still has a queue.
Before the next budget is approved, explain how faster proposals will improve the business result. Otherwise IT will keep reporting recovered hours while the board keeps waiting for more money.
Decide where the result should show up
Start with a specific change: more proposals, shorter customer waits, fewer corrections, or the ability to handle growing demand without another hire. The number of people using a tool tells you little about those outcomes.
For proposals, track whether earlier delivery improves acceptance. For service work, track completed cases and their full handling cost. Measure quality alongside speed: an incorrect proposal sent in five minutes can be more expensive than a correct one prepared in an hour.
Establish a baseline using representative work, including difficult cases. Compare similar tasks after launch and account for seasonality, team changes and marketing. Otherwise a new campaign can make the AI pilot look more successful than it was.
The process owner should explain how the team will use its additional capacity. In a small company, one accountable person and a shared page recording the baseline, target and review date can be enough.
- 01Faster task
- 02Available team capacity
- 03Additional work or avoided cost
- 04Economic benefit
A bottleneck in the next stage can stop the benefit reaching the business.
Count the whole task
Subscription costs are visible. Data preparation, integration, training, reviewing answers, handling exceptions and maintenance are easier to miss. A draft written in seconds still needs someone to read it. With usage-based pricing, also check what happens to costs when case volume doubles.
Calculate the cost of a correctly completed case. In customer service, that includes cases handed from a bot to a person. The price of one model response leaves out much of the work.
Recovered time creates economic value when it avoids overtime, delays a necessary hire or supports additional work with real demand. For incremental sales, count contribution margin rather than total revenue. Do not count the same hours again as a separate labour saving.
150 hours is a pilot result, not proof of return
Consider a hypothetical example. After review and corrections, a team recovers 150 hours each month. Every hour put to productive use delivers PLN 80 of economic benefit. Setup costs PLN 24,000 and ongoing operation costs PLN 4,000 per month. Assume stable volume and twelve full operating months from launch.
Potential benefit is PLN 12,000 per month. At 50% utilisation, it falls to PLN 6,000, or PLN 72,000 over the first year. That equals total first-year costs: ROI is zero and setup breaks even after twelve months.
At 25% utilisation, benefits do not even cover monthly operation. At full utilisation, the simplified setup payback is three months. The tool and technical result are identical; the way the business uses the time changes the outcome.
| Utilisation | Annual benefit | ROI | Setup payback |
|---|---|---|---|
| 25% | 36,000 PLN | -50% | No payback |
| 50% | 72,000 PLN | 0% | 12 months |
| 100% | 144,000 PLN | 100% | 3 months |
Show quality benefits separately
Shorter queues, less repetitive work and better documentation can justify a project before they appear in the income statement. Show them alongside ROI, with measures such as correction rates, waiting times or escalations.
Do not assign a convenient monetary value to “better customer experience” just to make the business case work. Establish its relationship to financial results first. Until then, it remains a valuable qualitative outcome.
A pilot should end with a decision
Agree expansion criteria before starting: lower case costs, no deterioration in quality and confirmed use of recovered capacity. Also decide what would trigger a redesign or closure.
In an SME, begin with one frequent process causing a tangible problem. Across a larger portfolio, operational improvements and experiments with new services may need different review dates. Each still needs a spending limit and a decision point.
At the next board meeting, show full cost, realised benefit and the gap against plan. If AI has freed up time, ask who used it and for what. That answer should shape the next funding decision.
← All posts