The most interesting AI question in a management meeting may now be what happened to the time saved. A faster first draft is useful. A measurable improvement in delivery, capacity or profitability requires a further operating decision.
McKinsey's August 2026 report, The state of AI in 2026: On the road to ROI, illustrates the gap. Eighty percent of respondents said AI had improved their individual productivity, while 37% attributed at least some impact on their organisation's earnings before interest and taxes to AI use. The latter share was broadly unchanged from the previous year. These are self-reported findings from a global survey, rather than a measured return for a particular Canadian business. Source: McKinsey's August 2026 report.
Separate three kinds of business benefit
CREDIUM recommends separating capacity, cash savings and incremental contribution when evaluating an AI initiative. Combining them into one headline number makes it difficult to understand what the business has actually gained.
- Released capacity: People have time available for other useful work. The payroll cost may remain the same.
- Realised cost reduction: A specific expenditure is reduced or avoided, with an accountable owner confirming the change.
- Incremental contribution: The business earns additional contribution after the associated delivery costs, supported by evidence that the initiative helped create it.
Quality and reliability also matter. Some benefits are strategically valuable without being immediately convertible into a dollar amount. Record them explicitly rather than assigning an unsupported financial value.
Measure from request to accepted output
Establish a baseline for the complete task: preparation, production, review, corrections and handoff. After introducing AI, measure the same boundaries. Comparing manual completion time with AI drafting time alone exaggerates the improvement whenever review remains necessary.
Use a representative mix of ordinary and difficult cases. Record the acceptance criteria before the pilot starts so that speed is not rewarded at the expense of accuracy. Where the volume of work varies, compare results per accepted output as well as total hours.
An illustrative calculation: time saved is capacity first
Consider a fictional team producing 400 research summaries each month. Suppose each summary previously required 24 minutes of total work. In a pilot, AI-assisted preparation and drafting take nine minutes, followed by seven minutes of review. The accepted output now takes 16 minutes, releasing eight minutes per summary.
Across 400 summaries, that equals 3,200 minutes, or approximately 53.3 hours of monthly capacity. At an illustrative loaded labour cost of CAD 60 per hour, the capacity has a cost-equivalent value of CAD 3,200. It is not automatically CAD 3,200 in cash savings.
The next question is operational: can the team use those hours for additional client work, a research backlog or another defined priority? If the hours are fragmented and cannot be redeployed, the practical benefit may be smaller. These assumptions are an example for planning, not a forecast or a reported CREDIUM client result.
Include the costs that sit outside the subscription
Record setup, integration, training, ongoing usage, monitoring, review and exception handling. Separate one-time costs from recurring costs and use the same evaluation period for costs and benefits.
A simple project ROI calculation is net realised financial benefit divided by total project cost for that period. For example, confirmed incremental contribution plus confirmed cost savings, less project costs, forms the numerator. Report released capacity separately until the business has a defensible basis for treating it as a realised financial benefit.
Avoid counting the same hour twice. If released capacity supports additional revenue, the model should not also treat the full labour cost of that hour as a cash saving without evidence of a separate cost change.
Assign an owner to the benefit
The technology owner can keep a system running. A business owner needs to decide how improved capacity changes scheduling, service levels or delivery commitments. Without that decision, a pilot may produce enthusiastic feedback and little visible change in business performance.
Set a review date and a small number of measures: accepted output, total effort, recurring cost and the chosen business outcome. Compare those measures with the baseline and record what else changed during the period. A stronger sales month, for example, may have several causes.
Make the next investment decision explicit
Expand when the workflow is reliable and the benefit has an owner and a credible path to realisation. Revise when the task is valuable but the operating design is incomplete. Stop when the evidence does not justify the ongoing cost.
For selecting the workflow, see AI adoption in Canada. For integrating measurement into management reporting, use the principles in strategic KPIs. A useful AI business case makes the connection between a better task and a better business outcome visible.