The 2026 AI Cost Reckoning: What Greek CFOs Are Now Asking
For three years, Greek enterprises bought AI on faith. Pilot budgets were approved with phrases like "this is where the puck is going" and "we cannot afford to be late." Few CFOs pushed hard, because everyone agreed that AI was important and the cost of being late felt greater than the cost of being slightly wasteful.
2026 is when that posture ends. Three things happened roughly simultaneously: AI subscription costs are now visible on every department's expense line, the first wave of pilots reached year three without a clear ROI story, and the board started asking pointed questions about productivity numbers that the CIO struggles to answer. The honeymoon is over, and the CFO is back at the table.
This is healthy. But it is also the moment most Greek AI strategies are going to wobble, because the questions a serious CFO asks are not the questions most CIOs have been preparing for.
The honeymoon ended for a specific reason
McKinsey's 2025 State of AI report contains a finding that surprised almost nobody and changed almost everything. Workflow redesign is the lever, not the model and not the platform. About one in five organisations have actually done it.
The other four-fifths spent two to three years buying AI and bolting it onto processes that were designed before AI existed. The result is roughly what you would expect: lots of AI in use, modest productivity gains, no obvious EBIT signal in the financial statements. The technology works. The operating model never changed to take advantage of it.
CFOs see this in the numbers. Greek enterprises that committed budget to AI in 2023 and 2024 are now in the year where the cumulative spend is significant enough to be worth scrutinising and the productivity story is too vague to defend. Hence the four questions.
Question 1: What did we buy, and what is it producing?
The first question is the inventory question. CFOs want a list of every AI tool the company is paying for, every AI agent that is running, and the measurable output of each. Most CIOs cannot produce this list inside an hour. Some cannot produce it inside a week, because shadow AI subscriptions have been quietly accumulating across departments for the past two years.
The output side is harder. For each tool, the CFO wants to know: what specific business metric did this move, and by how much? "Productivity improved" does not survive scrutiny. "Customer service agents handled 18% more tickets per week with the same headcount, freeing 0.7 FTE in the support team" does.
Greek enterprises that cannot answer this question for the majority of their AI spend are about to face budget cuts not because the technology was wrong, but because the measurement was missing.
Question 2: What would happen if we cut the bottom 30%?
The second question is the rationalisation question. Once the inventory exists, the CFO will ask which of these tools is delivering measurably less than its cost. The answer is uncomfortable, because in most enterprises the answer is between 25% and 40% of the AI portfolio.
Some of this is genuine waste, like a tool that was bought for a use case that never materialised, or a duplicate capability across two vendors. Some of it is more painful: a tool that the team likes but cannot tie to a specific output.
The right CFO conversation here is not "cut everything we cannot prove." It is "let us run the cut, and reinvest the savings into the use cases that are working." Without that frame, the AI budget gets cut by the percentage of waste plus the political margin the CFO needs to feel safe, which is usually too much.
Question 3: How do we know the productivity gains are real?
The third question is the measurement-rigour question. CFOs are appropriately suspicious of self-reported productivity gains, because every department head has a structural incentive to overstate them.
The credible answer requires actual measurement. It needs time-and-motion data on the workflow before and after deployment, ideally with a control group, and a documented baseline so the comparison is honest. Greek enterprises that did this work upfront have credible ROI stories. The ones that did not are now being asked to retrofit measurement onto deployments that have been running for a year, which is harder than it sounds.
The mistake to avoid here is letting the measurement question become a reason to stop investing. The right response is to invest in measurement infrastructure as a first-class capability, not to defund the AI projects until measurement is perfect.
Question 4: What is the workflow redesign plan?
The fourth question is the strategic one. Once a CFO understands that workflow redesign is the actual EBIT lever, the question becomes: what is the company's plan to do that redesign, and what is the timeline?
This is the question where most enterprises do not yet have a credible answer. Workflow redesign is harder than tool deployment. It involves changing approval chains, retraining people, sometimes restructuring teams. It is the work that the past three years of AI investment was supposed to set up, and it is mostly still ahead.
The CFOs we work with are not asking this question to embarrass the CIO. They are asking it because they need to allocate capital to the next phase of AI investment, and the workflow-redesign phase will be more expensive than the tool-acquisition phase. They want to know that the company has a plan to spend that money well.
How to structure a board-ready ROI case
The format that holds up under board scrutiny is structurally simple. For each AI investment, document four things:
The hypothesis: what business outcome were we trying to achieve, and how much did we expect it to move?
The intervention: what did we deploy, what did we change in the workflow, and what did it cost (full cost, including internal time)?
The result: what actually moved, measured against the baseline, with the methodology disclosed?
The next step: based on what we learned, what are we changing, sustaining, or stopping?
This is not novel. It is the same format any serious capital allocation decision uses. The reason most AI investment cases struggle is that they were never structured this way. The deployments happened, the budget was spent, but the learning never got organised into a form a CFO can review.
Greek enterprises that retrofit this structure onto their existing AI portfolio find two things. First, the picture is not as bad as the absence of clear reporting suggested. Many tools are doing real work, just unmeasured work. Second, the retrofit work itself sharpens the next year of investment, because the team learns what to track from the start.
The CFO and CTO partnership Greek enterprises need
The conversation we keep having with Greek leadership teams is about restructuring the CFO and CTO relationship around AI. The historical pattern was: the CTO proposed, the CFO challenged, a budget was negotiated, and the project ran. That pattern does not work for AI investment because the productivity outcomes depend on operating-model changes that neither role owns alone.
The pattern that does work treats AI investment as a joint capital and operations decision. The CTO brings the technical reality and capability roadmap. The CFO brings the measurement discipline and capital allocation framework. The COO or business unit head brings the workflow redesign mandate. Decisions get made together, and accountability for outcomes is shared.
For Greek enterprises that have not yet restructured this relationship, 2026 is the year to do it. The CFO scrutiny is healthy. The risk is letting the scrutiny become a reason to retrench rather than a forcing function for better AI strategy.
What we tell our clients
If you are about to walk into a board meeting where the AI ROI question is going to come up, three things to bring:
The honest inventory. Every tool, every cost, every measurable output. If a tool has no measurable output, that is itself the answer: either build the measurement or kill the tool.
The workflow plan. Not a generic transformation slide, but specific processes you intend to redesign in the next four quarters and the EBIT thesis for each.
The measurement infrastructure proposal. Treat the ability to measure AI productivity gains as a first-class investment, with its own budget and its own owner.
If you can walk into the room with those three artefacts, the AI conversation goes from defensive to strategic. If you cannot, the budget gets cut by reflex.
We help Greek enterprises build the measurement and workflow-redesign capability that turns AI spend into EBIT signal. If your CFO is starting to ask the harder questions, that is the moment to talk. Get in touch at inbusiness.gr.