“There’s going to be two types of companies: those who are great at [artificial intelligence (AI)], and everybody else. And the ‘everybody else’ is going to fail because AI is such a transformative tool.”
This insight came from the famous American businessman and television personality Mark Cuban in a recent conversation with Clipbook founder Adam Joseph. As Cuban stresses in their discussion, ignoring AI will be the fastest way for companies to fall behind.
While most finance leaders know that AI matters, many are stuck chasing big, risky transformations that stall or disappoint. Too many organizations have treated AI as a single, transformative event: buy the right platform, flip a switch, and watch the function modernize.
But finance rarely evolves through quantum leaps. It evolves through trials and evaluations, gradual adoption, and earned trust. Today, AI is better understood not as the next great differentiator, but as a foundational capability, one that must be adopted the same way earlier generations adopted automation and analytics— gradually and with a clear eye on measurable value.
Finance leaders today are operating somewhere between oversimplification and overwhelm. On one side is the promise of AI being “plug-and-play,” where an organization can purchase a solution, connect a few data sources, and unlock increased efficiency. On the other end is the assumption that AI can only deliver value after years of data cleansing, process redesign, and wholesale system replacement. Both sides are a bit misleading, and both approaches can stall adoption before any benefits are realized.
For many organizations, the first hurdle is the assumption that generative AI can simply be layered on top of existing solutions. Gianne James, senior vice president of governance and assurance at FIRST Insurance Funding, says that’s one of the biggest misconceptions she sees organizations make about generative AI: “The reality is that generative AI is only as strong as the data and governance structures underneath it.”
Before even looking at vendors, James says finance leaders should be asking whether their data management and data governance foundations are solid: “If those aren’t in place, AI will only amplify existing issues.”
At the same time, some finance teams assume that adopting AI requires a massive, high-stakes overhaul of existing systems and structures.
“Meaningful results usually come from targeted, well-scoped improvements to existing finance workflows, not from replacing entire platforms or restructuring teams,” says Ray Beste, a principal AI strategist at Sikich.
This controlled, staged approach is also recommended by Kirstie Tiernan, national advisory AI leader and board member at BDO USA.
“Finance leaders often picture AI like a wrecking ball that’ll upend their core operations, but I see it more like a power tool that you can pick up when you need it,” Tiernan explains. “It makes the job faster and cleaner.”
Tiernan believes that most early AI wins happen around very targeted, unglamorous workflows. She agrees with James that you need a solid data foundation, secure environment, and focus on people—but she suggests that companies think big and start small.
“AI initiatives often fail because they try to ‘boil the ocean’ so to speak. Finance teams launch these huge, visionary programs without grounding them in day‑to‑day operational realities. Teams also skip over data readiness or underestimate the change‑management lift,” Tiernan stresses. “If AI doesn’t make someone’s Tuesday easier, it’s not going to get adopted.”
AI certainly isn’t a magic bullet that’ll solve every finance challenge. In fact, in his discussion with Joseph, Cuban stresses that treating AI like a shortcut or a magic fix often leads to bad decisions—he says companies that do well will be the ones that use it thoughtfully and with clear intent.
Therefore, the first step for finance leaders is to understand what AI can and can’t do.
James reminds finance leaders that generative AI can go beyond productivity tools; it can support creation, automation, pattern detection, and trend analysis: “That’s why leaders need to pause and ask themselves: Do we truly need AI here, or would something simpler like robotic process automation suffice? Are we eliminating repetitive steps, or are we seeking more sophisticated insight?”
To illustrate, Beste says the tax team at Sikich uses software to manage who’s working on which tasks across multiple teams in multiple locations. Certain specialists aren’t needed until a specific point in the process and notifying them at the right time required emailing multiple people on a regular basis.
“The initial thought was that AI could help solve this problem,” Beste explains. “In the end, workflow automation turned out to be the right solution, with AI assisting in the development and testing.”
Beste adds that using AI to help develop and test the process reduced the build time by about 90%, and the final solution reduced a weekly task from hours to roughly a five-minute process.
James explains that a common misstep organizations make is having AI decisions reside only under IT or technology purview, when instead they should be cross functional: “The strongest AI use cases are thoughtful, enterprise-level decisions that consider strategy, capacity, training, and how people will actually use the outputs.”
Understanding how the end user will actually interact with an AI solution is key to successful implementation. As AI gets more deeply integrated into business processes, it’s critical for end users to understand the role that technology plays, and what levers influence any outcomes. Their knowledge of business priorities, situational context, and acceptable risk enables them to recognize flawed or non-optimal outputs.
“The person interacting with the AI is often the largest source of risk but also the first line of defense,” James emphasizes. “If users aren’t equipped to question what the model produces, issues will surface downstream.”
The interaction between end user and AI means that education and governance are just as essential to success as the implementation of the technology itself. Equally important is transparency into the data that forms the foundation for AI, especially in financial contexts, where accuracy, compliance, and trust are critical.
The risks of poorly governed integrations are already playing out in the courts. In Mobley v. Workday Inc., Derek Mobley alleged that he was immediately rejected from over 100 jobs by companies that utilized Workday’s AI-powered platform to screen candidate applications. These rejections often occurred within minutes of submission, including outside normal business hours. Mobley claims that the AI screening process unlawfully discriminated against him based on age, race, and disability. At the time of this writing, the case remains ongoing.
“Strong data governance is nonnegotiable,” James reiterates. “Building AI on an unstable foundation can be costly, harmful to users, and damaging to an organization’s reputation.”
Recognizing that AI doesn’t exist outside the regulatory landscape, especially within finance, James recommends bringing risk and compliance teams to the table early: “The most effective leaders don’t wait to build governance in response to AI deployment— they establish an AI governance framework upfront, grounded in accountability, oversight, and clarity of roles.”
James adds that embedding governance into the system from the ground up doesn’t slow innovation, rather it creates the conditions for sustainable, trusted adoption.
When beginning AI integration, another best practice is to start with a well-defined problem statement and a clear understanding of what the AI solution is intended to address. Beste explains that if a problem is framed too simply, such as “use AI to close,” rather than with more detail, such as “cut variance commentary prep from eight hours to three, with review sign-off,” it becomes difficult to determine whether the solution is actually working.
Another challenge Beste sees frequently is organizations underestimating what it takes to move from concept to enterprise deployment. He says that teams often overlook questions such as, “How will the solution fit into existing workflows?” or “Who maintains the knowledge sources the AI relies on?”
While Beste says that organizations underestimate the initial investment, Randy Johnston, executive vice president of K2 Enterprises, warns that organizations might overestimate the impact.
“Expectations are often faulty or too grandiose,” he says. “Starting small on very discrete projects will allow for some early wins and provide actionable results.”
Low-friction uses tied to daily work tend to deliver the fastest return on investment (ROI), Beste says. This includes activities like variance-analysis summaries, account reconciliation explanations, contract and invoice review, and policy interpretation.
“AI embedded directly into existing finance team tools, such as in their enterprise resource planning (ERP) systems and Microsoft tools (Excel, Teams, and Outlook), delivers value quickly without having to retrain teams on entirely new systems, thus accelerating adoption,” Beste advises.
Tiernan observes that the fastest ROI tends to appear in workflows that are high-volume, rules-based, and in areas where no one on the team is eager to spend their time. One of the most surprising implementations she witnessed was when a finance team she worked with invested $105,000 in accounts payable.
“After implementation, operating costs dropped by roughly $21,000 per year. Efficiency and risk-control improvements also delivered about $61,000 in net annual value,” Tiernan recounts. “The payback period was only about 1.7 years, and the forecasted five-year ROI landed around 190%. The team’s control maturity also improved, meaning their accounts payable processes also became more easily auditable and dramatically more reliable.”
She emphasizes that a sweeping AI transformation isn’t always necessary to achieve impact. Sometimes just modernizing a single pain point can yield significant cost and time efficiencies that create the momentum needed for future investments.
Looking forward, Johnston believes the next major leap will come from how AI connects to core finance systems: “The development and delivery of model context protocol (MCP) interfaces from various vendors in 2026 will provide interaction with AI into ERP data and other systems, and implementing MCP models with pre-defined prompts will be a small change that’ll have significant impact.”
Of course, the central question for finance leaders isn’t what AI can do but how its impact can be measured. Hours saved was once considered the gold-standard metric, but today, it’s just the standard baseline starting point—efficiency gains are now a given.
“What leaders really care about now are the key performance indicators (KPIs) that change the financial profile of the business,” Tiernan explains.
To unbox these KPIs, Tiernan suggests that leaders ask themselves:
Tiernan adds that the metrics that matter the most are tied to working capital improvements, real operating cost reductions, cycle-time compression, stronger control maturity, fewer exceptions, and revenue-adjacent gains.
Beste adds that another critical metric to track is adoption rate: “If people aren’t using the solution available, none of the potential gains are realized.”
Overall, Beste says that the three most important metrics to track are: hours saved per user per week, percentage of the team actively using the tool, and user satisfaction scores. He also advises providing an easy mechanism for end-users to supply feedback for improvement.
This leads into the next step, which is designing adoption strategies that build momentum without overwhelming teams. To do so, Beste says the first step is ensuring that humans are kept “in the loop.”
“AI should provide recommendations, but people need to review and approve outcomes,” Beste emphasizes.
Tiernan advises establishing a clear path so that everyone understands the purpose behind the effort. She recommends connecting AI solutions to real business problems and tangible outcomes: “Introduce AI in a way that real humans can absorb without feeling like they just got handed another job.”
BDO’s approach is formalized through a standardized process that begins with general education about AI and scales to strategy, implementation, and expansion. “When teams see how the work ties to measurable results, adoption becomes a shared mission instead of a leadership directive,” she says.
While the future role of AI in finance is undeniable, the magnitude of its impact will depend on governance and adoption. Organizations that treat AI governance as foundational and take a people-centric approach to implementation are most likely to see practical wins that extend beyond the hype.
