There’s a quiet irony in today’s AI conversation. As technology continues to advance at breakneck speed, with generative tools and machine learning models becoming more accessible and powerful, many finance teams feel an unspoken pressure: to do something (read anything) with AI.

It’s no longer just a matter of curiosity or innovation. In some organisations, the perception has taken hold that if a project doesn't involve artificial intelligence, it lacks ambition. This mindset, while understandable in the current climate, is also problematic.

Because in finance, as in many disciplines, the most strategic decision is often not whether you can use AI, but whether you should.

This article explores how finance leaders can approach AI selection with intent, rigour, and realism. It considers where the technology has shown measurable value, where traditional methods continue to excel, and how organisations can establish a thoughtful framework for adoption. As the Executive Finance Summit in Zurich draws near, it’s a timely moment for reflection – not just on what’s possible, but what’s truly valuable.

The myth of AI as a universal solution

AI, especially in the context of FP&A, is best understood as a capability, not a panacea. It can enhance accuracy, reduce cycle times, and spot patterns across large data sets. But it is not automatically better than the models and methods finance teams have refined over decades. Nor is it equally suited to every task.

Yet the narrative around AI often implies otherwise. Technology vendors and media alike tend to frame AI as a universal upgrade, capable of improving any process it touches. 

This creates two risks:

• First, it can lead to misallocated investment in the form of time, talent, and funding directed toward AI use cases that do not warrant the complexity.

• Second, it risks undervaluing the deep domain expertise and process maturity that finance teams already possess. In many scenarios, what’s needed is not a transformation, but a refinement.

Strategic AI use begins by stepping back from the hype and asking a harder question:

What problem are we trying to solve?

A framework for selecting the right use cases

Not every problem is an AI problem. But some clearly are. So how can finance leaders distinguish between them? 

A useful starting point is to consider four core dimensions:

1. Data quality and availability. AI models require large volumes of clean, structured data. If a process draws on inconsistent, fragmented, or highly qualitative inputs, then AI may struggle to deliver meaningful results. In these cases, process redesign or data standardisation may be more valuable than advanced modelling.

2. Pattern complexity. AI shines in areas where patterns are too subtle or dynamic for rules-based models to capture. This includes demand forecasting in volatile markets, fraud detection, or identifying driver relationships in cost structures. If the relationships in your data are well understood and relatively stable, traditional methods may be more transparent and efficient.

3. Business criticality and tolerance for error. High-stakes decisions, such as board-level reporting or statutory planning, often demand explainability and auditability. AI models, especially black-box ones, can struggle to meet these standards. Where regulatory scrutiny or reputational risk is high, clarity may outweigh complexity.

4. Value-to-effort ratio. AI models can be resource-intensive to develop and maintain. If the expected performance uplift over current methods is marginal, then the investment may not be justified. Quick wins with a high return on effort are better candidates for early adoption.

This framework is not about discouraging innovation, but about directing it where it matters most. By evaluating potential use cases through the lens of data readiness, complexity, risk tolerance, and value, finance leaders can ensure that AI adoption supports the company’s strategic priorities.

Where AI has already proven itself

Despite the need for selectivity, there are several areas where AI has demonstrated consistent value within FP&A workflows. Predictive forecasting stands out as a clear example. Machine learning models trained on historical financial and operational data can generate rolling forecasts with impressive speed and accuracy. Particularly in industries with large volumes of granular transactions, such as retail or logistics, AI can uncover demand shifts, seasonality effects, or cost trends far earlier than human modellers.

Another promising use case is in driver-based planning. AI can help identify which variables most significantly impact outcomes, even across complex and non-linear relationships. This can improve the calibration of assumptions and support more dynamic scenario modelling.

Additionally, anomaly detection is increasingly being used to enhance variance analysis. Rather than scanning thousands of lines manually, AI models can surface unusual activity across entities or accounts, helping finance teams focus their attention more effectively.

These are not theoretical possibilities. They are operational use cases already being embedded in finance teams with the right data and governance in place.

When traditional methods still outperform

It is equally important to acknowledge the enduring power of simpler methods. In areas where inputs are limited, relationships are well understood, and outputs require transparency, conventional financial modelling remains more practical.

Consider a standard 12-month cash flow projection for a stable business unit. A structured spreadsheet, built with business rules and validated by experience, may offer faster turnaround, greater stakeholder confidence, and easier scenario adjustment than a machine learning model trained on years of historical data.

Similarly, when building plans in areas with significant human judgment such as R&D investment or M&A evaluation, AI may not offer a material advantage. These decisions rely less on historic trends and more on strategic intent, competitive insight, and qualitative reasoning.

Overengineering a solution with AI in these contexts can slow down delivery, increase technical debt, and frustrate users. The goal, then, is not to eliminate traditional approaches but to augment them – knowing when to bring in advanced analytics, and when to trust in established processes.

The road ahead: deliberate, not default

As finance teams continue to explore the potential of AI, the most strategic move may not always be adoption, but discernment. Knowing when to invest (and when to decline) is fast becoming a defining capability of forward-looking FP&A leaders. Saying no to AI in the wrong places is not a failure of ambition, but a marker of focus and maturity.

At Apliqo, we believe the future of finance will be shaped not just by new tools, but by smarter choices. These are the very conversations we’ll be advancing at the Executive Finance Summit in Zurich on the 25th of June, where thoughtful leaders are gathering to chart a path forward built on both intelligence and intention.

If you’d like to join us, be sure to apply here and come contribute to this incredibly important conversation about how AI will disrupt and transform FP&A and the CFO office more generally.