Most AI initiatives in finance and accounting fail before the rollout begins. The last piece in this series ended with three priorities: build the right coalition, make a credible business case, and balance pressure for fast returns with the groundwork real change requires.

Those priorities align in theory. In practice, coalitions and credible business cases take time, while pressure for fast returns leaves little of it. Something gives, and it is rarely the pressure.

John Kotter mapped eight steps for driving change. This piece focuses on the first three, where finance and accounting AI initiatives often stall before rollout: creating real urgency, building the right coalition, and making a credible business case despite pressure for fast returns.

Urgency without urgency theater

Kotter's first step is creating urgency. In finance and accounting, the real question is what kind leaders create. "AI is the future, we have to move now" is an announcement, not an argument. Teams that have lived through ERP rollouts and reporting changes know the difference. Manufactured urgency asks people to commit to a feeling instead of solving a problem.

Real urgency names the problem: this workflow, this deadline pressure, this audit risk. Compare "AI is transformative" with "the close takes four extra days because reconciliation is manual, and here's what that costs us." A team can rally around the second. The first is just a slogan.

Fast-ROI pressure turns urgency into a trap. Leadership wants returns; the team wants relief from a heavy workload. Those goals may support the same initiative, but treating them as interchangeable creates excitement at the top and skepticism below.

Building the coalition: who should be in the room

Kotter's second step, building a guiding coalition, is where AI initiatives can go wrong. Finance leaders and data and technology owners may be able to approve the project without understanding what it will ask of the staff accountant closing the books.

Resistance comes from more than one direction. Leaders may worry that AI will not deliver returns quickly enough or that delaying adoption will leave the business behind; staff may fear being replaced or not knowing where to start. A coalition that aligns leaders and assumes everyone else will follow solves only half the problem. Without the people who will feel the friction in the room, a business case that looks complete can fall apart in the first week.

A credible AI vision has to be specific, not aspirational

Kotter's third step is defining a clear vision of where the change leads. That is harder with AI because the range of possible use cases can make almost any outcome sound plausible. When leaders cannot narrow that potential into a few concrete changes, the vision stays broad enough to inspire but too vague to guide decisions or help employees understand what it means for their work.

"This will save you time" rings hollow when teams have heard the same promise about systems that failed to deliver. A credible vision makes the change tangible: which task gets lighter, which frustration goes away, and which role benefits. People should be able to picture the difference in their own work, not just agree that AI has potential.

A business case has to work for two audiences

AI initiatives often get squeezed between ambitious return expectations and the time required to change how work gets done.

That is why an AI business case has to work for two audiences. Investment committees need the ROI case: cost savings, speed, and a return timeline. Teams need the adoption case: less rework, less manual reconciliation, and a clearer audit trail. Build only the first, and the initiative may win approval without winning adoption. The strongest business case answers both questions: Why is this worth funding, and why is it worth changing how we work?

What comes next

A credible case and the right coalition get the team to the starting line. Next comes building broader support, creating early wins, and preparing people to change how they work. That is a question of individual readiness, and where this series turns next.

Where has your team seen an AI case get approved but never actually take hold? Was it a coalition problem, a business case problem, or something that only showed up once the rollout started?