The Roland Berger x AGOS GBS readiness data tells one story clearly. Seventy-four percent of GBS organizations across the region have an AI vision. Twenty percent have a way to evaluate whether it is paying off. Wolters Kluwer’s 2026 APAC CFO survey found the same pattern, with eighty-three percent of regional finance leaders now calling AI a force reshaping finance. The gap between vision and measurement has just become a boardroom demand for measurable AI return within twelve months.
The mandate has hardened. AI investment approved in 2025 is being asked to deliver an operating-budget return by the end of 2026. Finance functions are first in line to prove it, because their processes are the most repeatable, rule-based, and voluminous in the enterprise. That puts a different demand on the CFO too, to test agentic AI deliberately, accept some pilots will fail fast, and treat those failures as data rather than a verdict on the technology. If AI works in Finance, the executive committee expects it to work everywhere else, and if it does not, next year’s AI budget line is at risk.
The standard for proof has also moved. A year ago, time saved was acceptable, and a thirty percent drop-in invoice hours counted as a genuine win. That does not survive a 2026 board meeting. Boards now want Finance to show cost reduction the operating P&L reflects, working capital gains that Treasury team can harness, lower logistics cost from narrower forecast variability, analyst guidance accurate enough to move the share price, and real-time fraud detection that pre-empts the audit cycle.
These outcomes are achievable in twelve to eighteen months. What blocks them is rarely the technology. Grant Thornton’s 2026 AI Impact Survey found the real blocker at scale, with nearly three in four organizations giving agentic AI data access while only one in five has a tested incident response plan. That gap shows up as the same pattern everywhere. A process gets identified, designed, and proven, then sits because data access has not been signed off or security review is mid-cycle. The build is solid while the IT clearance stays unresolved, and a twelve-month timeline does not survive that stall.
The CFO can prevent it by putting three questions to IT before any AI spend is approved, and by bringing the management board into agreement on the answers. First, what is the data access protocol for an agentic AI agent operating inside this process, and who owns the sign-off? Second, when an agent makes a judgement call the human reviewer cannot easily explain, what audit trail does the organization require? Third, if a model’s output cannot be reconciled to the underlying transaction data, what is the incident response, and who escalates it?
These are governance questions rather than technology questions, and they need answers before the contract gets signed. Finance leaders who raise them early, and secure board agreement on them, avoid the holding pattern that has stalled too many good builds.
There is a parallel workstream that decides whether any of these holds, which is the redesign of Finance roles once the tool is live. It is a discipline of its own, and we have written separately about the talent shift that has to happen before the tool ever lands. The one point worth carrying here is that the redesign has to be planned alongside the build, because a deployment that reshapes the operating model without repositioning the people inside it stalls just as surely as one held up by IT.
The mandate has moved past whether AI is coming for Finance. The real question is whether you have started. Pilots will stall and some will fail, and that is the cost of moving rather than a verdict on the technology. The functions that begin now, and keep progressing through the failures, become the proof points. The ones still waiting for certainty before they start will watch the decision get made for them.
Joon Teoh is Chief Executive Officer of AGOS Asia, an AI-first GBS transformation partner working with CFOs and GBS leaders across ASEAN.
TNGlobal INSIDER publishes contributions relevant to entrepreneurship and innovation. You may submit your own original or published contributions subject to editorial discretion.
Featured image: Steve A Johnson on Unsplash