For many CFOs, AI has moved beyond experimentation and into the budget. Copilot licences, ChatGPT and Claude subscriptions, training and governance are all adding up. As adoption increases, attention is shifting towards whether those costs are producing an adequate return.
Not all AI returns are created equal
Some returns are relatively straightforward to quantify. If AI enables a new hire, reduces outsourcing, rework, or delays, the financial benefit can be measured directly against the investment.
Much of the value currently attributed to generative AI is less direct. Employees report saving time drafting documents, analysing information, or completing administrative work. Those gains matter, but an hour of salaried time saved does not automatically reduce cost or create revenue. It creates capacity, and the economic return depends on what the organisation can do with that capacity.
Capacity is only valuable when it is used well
If a team can process more work without adding headcount, avoid a planned hire, increase customer-facing activity or remove a constraint on throughput, the productivity gain has a clear route to financial value. If the released time is simply absorbed back into the working week, multiplying hours saved by an hourly labour rate risks overstating the return.
Where the time is saved also matters. Ten hours released from a process with spare capacity may have little financial impact. The same saving at a bottleneck that limits revenue, throughput or access to scarce expertise can be considerably more valuable.
This is why AI investment needs to be considered at the level of the business and its workflows, rather than simply as a collection of useful tasks.
Begin with the economics of the business
A recent AI Discovery session with a professional services firm in Hastings reinforced this. We began by examining the business’s growth aspirations to 2030 and its current product mix and relative margins. We then looked at where capacity was constrained, where scarce expertise was being underutilised, and which external industry shifts were impacting profitability. Only then did we explore where AI might make a material difference. A key observation was where AI, used by an intermediary, was steadily eroding margins of a core channel.
The most valuable AI opportunities change economics, not just effort
The strongest AI opportunities are those that alter what the business can economically achieve. That may mean increasing throughput without adding headcount, extending scarce expertise across more high-margin work, reducing the cost or time of serving customers, or making a previously uneconomic service viable. These opportunities matter because they change the relationship between cost, capacity and output, rather than simply making an existing task faster.
From productivity gains to business outcomes
The first phase of generative AI adoption has understandably concentrated on helping individuals work faster. The next phase will require more discipline around which productivity gains are worth pursuing and how they translate into business outcomes.
For CFOs, that means moving beyond measures such as licences deployed, users active or hours reportedly saved. Those are useful indicators of adoption and productivity, but they are not sufficient measures of return.
AI is not only an internal efficiency tool; it is also reshaping the economics of the industries in which businesses operate. It should therefore be assessed in terms of what it enables the organisation to do differently: reduce cost or risk, increase throughput, extend capability, remove constraints or make new forms of value creation economically viable.
Matt Crowther Co-founder, Centaur AI
matt@centaurai.nz
021 793 363
