Everyone has shifted from talking about artificial intelligence to figuring out how to effectively adopt AI, and in Finance it’s no different. AI is dominating boardroom conversations, and for CFOs and finance leaders, the pressure to separate signal from noise has never been higher. 

Leaders are trying to discern: What’s real and what’s hype?  How should finance organizations think about an overall AI strategy, and how do AI modules in finance systems fit in? How do you evaluate and determine what is the right AI tool for you and your company? All of these questions and more were explored during a podcast discussion in which Chief Technology and Innovation Officer Vikram Bhandari and Riveron colleague Glenn Snyder joined podcast host Paul Barnhurst to discuss AI within finance functions and broader enterprise strategies.

How should companies look at and evaluate AI solutions?

As Vikram observes, many well-intentioned company leaders evaluate AI solutions the wrong way. They compare features, look at flashy demos, fancy dashboards, and perhaps mobilize a pilot project, and then wonder why some solutions fail. Often, AI initiatives fail because of a shaky foundation. For CFOs, it’s important to determine whether your finance architecture is mature enough to absorb AI without breaking. The winners will be the companies that are structured, consistent, and decision-focused when it comes to AI adoption and innovation, and this will be true in contexts where data, processes, and ownership are strong enough to reliably support the business.

To guide a successful approach to AI, Vikram recommends aligning around solving problems rather than evaluating exciting tech features, and equally important is understanding how AI integrates into existing data infrastructure. In any successful technology transformation, it becomes critical to connect AI into a unified data and systems architecture across ERP, EPM, and CRM platforms.

Another key area is understanding where AI should reside. Namely, who should own the AI platform: Finance or IT? And do finance teams need data scientists? In Vikram’s experience leading strategic technology enablement initiatives, as soon as Finance becomes overly dependent on IT or a third-party vendor, adoption is slowed or, worse, the system fails. Glenn has observed similar situations where companies struggle to fully adopt and integrate their technology solutions because key stakeholders might be more focused on developing a report rather than solving underlying business problems. For CFOs, an effective approach to harnessing AI is ultimately about strengthening the office of the CFO, ensuring finance is not just consuming AI outputs, but owning the decisions, governance, and processes that make those outputs actionable.

Before leaders jump into an enterprise AI solution, here are a few things that Vikram recommends to guide success:

• Identify high-value, decision-oriented use cases
• Conduct data-readiness assessments to ensure accuracy and consistency
• Map the workflows where AI will live
• Establish clear governance and ownership structures upfront
• Consider the human impacts of AI and related needs for change management 

Vikram points out: “If you can’t define the decision you’re trying to improve, AI won’t help you.”

Done right, an enterprise AI strategy strengthens companies and their finance functions

In many companies, different areas are leveraging AI modules in their core systems without any enterprise strategy, risk management, or governance model. For CFOs, this creates real risk, ranging from inconsistent outputs to lack of auditability. In these cases, the disconnected AI solutions could produce different answers to the same questions, as the AI agents are stuck in siloed data sets. An enterprise AI strategy helps unify these efforts—connecting data, models, and decision-making across the organization. 

An enterprise AI strategy can help to connect AI agents to create a conversational AI response that can cut across multiple data sets to produce a more accurate and valuable solution. This conversational AI approach allows finance leaders to leverage AI beyond  retrospective reporting (“What has happened?”) and predictive analysis ( “What will happen?”) to prescriptive insights (“What should we do if this happens?”).

1	2	3
Moving away from a legacy finance focus…	…to improved capabilities…	…toward an ideal AI-enabled state
Retrospective Reporting	Predictive Analytics	Prescriptive Insights
Asks: What has happened?	Asks: What will happen?	Asks: What should we do if this happens?

As Vikram describes in the podcast discussion, companies should not be thinking about AI in specific category, rather it should include predictive models, signals generation, optimization models, and recommend actions where AI can communicate insights. Ultimately, the enterprise AI strategy will result in a decision system that cannot be achieved if leaders are only looking at platform AI strategies.

AI forecasting considerations for finance leaders

Many AI modules require multiple iterations to learn and improve prior to delivering accurate forecasts. For instance, the VP of FP&A at a large tech firm shared with Glenn that, although the firm uses AI for forecasting, the team still maintains their manual forecast process. The leader’s rationale for continuing with the manual process was that, although the AI forecast is “75% wrong,” it remains valuable as a diagnostic tool to improve the company’s existing process.

In fact, one EPM vendor recommends running AI for six quarters before truly leveraging it as a forecast tool. To improve or shorten the AI adoption process, consider:

• Focusing relentlessly on data and foundational aspects, as AI does not learn in abstract, it learns in the data quality and model design
• Mobilizing a phased approach that includes Phase 1, Calibration and Validation, with two to four forecast cycles to benchmark output; and Phase 2 – Trust Building: two to three additional cycles for explainability and consistency
• Understanding the realities, as AI forecasting is not plug-and-play but requires iteration, structured data, and feedback loops
• Accelerating learning by running historical back-testing based on a prior forecast (for example six quarters of forecasts, not actuals) to better simulate the conditions of real decisions and leverage AI forecasting tools sooner

Vikram closed the podcast with the advice, “Think of AI as a capability multiplier, not a shortcut.” AI will only deliver value if your data, processes, and governance are strong. Organizations that focus on strengthening the office of the CFO and transforming their technology foundation will be far better positioned to turn AI into a true driver of performance and enterprise value.

“If you can’t define the decision you’re trying to improve, AI won’t help you.” - Riveron CTIO Vikram Bhandari notes in a podcast discussion on AI within finance functions and broader enterprise strategies.

https://www.thefpandaguy.com/fpa-unlocked-podcast/ai-strategy-finance-leaders-must-use-to-avoid-siloed-tools-and-failed-adoption-with-vikram-bhandari