At the 31st AI FP&A committee meeting, on the 3rd June 2026, Fernanda Noronha (Director of Finance, Process Optimisation, Data & Analytics at Grainger) challenged one of the most common assumptions in finance transformation that AI itself will transform FP&A.

Her message was clear: AI may accelerate reporting, automate analysis, and support forecasting, but it does not create transformation by itself. The differentiator is the operating model around it, including how finance is structured, how data is governed, how processes are standardised, and how FP&A teams use technology to support better business decisions.

Many organisations now have access to powerful AI tools, large language models, automated workflows, and advanced analytics platforms. Technological access is no longer the main constraint; instead, the question is whether FP&A has the structure, skills, and governance required to turn tools into better decisions.

From Faster Reporting to Better Decisions

A recurring theme of the session was that AI placed on fragmented data, manual processes, and siloed teams will only deliver limited value. It may make some activities faster, but it will not necessarily improve decision-making quality.

For AI to have a meaningful impact, it must be combined with integrated data, standardised processes, and dedicated analytical capabilities. This moves FP&A away from producing more reports and towards shaping better decisions.

Fernanda described this as an AI-enabled finance operating model that is built around four core capabilities:

• Data engineering

• Data science

• Process design

• Visualisation or self-service analytics

Together, these capabilities create the foundation allowing FP&A to focus less on building data pipelines, cleaning data and creating dashboards, and more on interpreting insights, challenging assumptions, and influencing decisions.

The future of FP&A is not about replacing finance judgment with automation; it is about industrialising the repetitive analytical work so that finance professionals can spend more time on business interpretation.

In the discussion, Fernanda also emphasised that the chatbot was first tested within finance before being considered for wider business use. The team spent significant time training the model on company language, refining the semantic model, and collecting feedback from finance users. This showed that the development of AI tools can be relatively fast, but the testing, adjustment, and trust-building process takes much longer.

Four Practical Use Cases

This operating model was presented through four use cases:

The first was fully automated daily sales reporting, where data is processed overnight so business leaders receive relevant information early enough to act on it.

The second was KPI governance and standardisation, which ensures that finance teams use shared definitions before dashboards, reports, or AI models are built.

The third was machine learning forecasting, where machine learning supports the statistical baseline while FP&A remains responsible for interpreting outputs and applying business judgment.

The fourth was the finance chatbot, which supports conversational analytics by answering routine questions and producing simple explanations from governed data.
 

The Hard Part is Not Building the Tool

The Q&A discussion highlighted that the biggest implementation challenges are not always technical. Committee members raised practical questions about hallucinations, accuracy, prompts, semantic models, token costs, accountability, and adoption.

One important takeaway was that development can be fast, but testing and refinement take time. In the chatbot example, the proof of concept began in finance before broader use was considered, and the team had to train the model on company language, test outputs, collect user feedback, and improve the semantic model. Accuracy was treated as an ongoing process of monitoring and adjustment. Super-users also played an important role because they understood how the business asked questions and could help improve the language and logic behind the model.

The discussion also showed the importance of human involvement. Some low-risk tasks may be suitable for AI support, whereas larger decisions and client-facing judgment require a human in the loop.

The discussion also highlighted that accountability does not move to the model. For automated reporting, finance leaders still need review points, process owners, and clear escalation routes before outputs are shared more broadly. Automation can accelerate the process, but FP&A remains responsible for validation, interpretation, and judgment.

Change management was another important theme. Adoption was not automatic, even with an analytically mature finance team. Leadership support and finance “super-users” helped demonstrate when the chatbot could answer routine questions faster, while still allowing users to check outputs against existing reports. The broader lesson is that AI adoption depends not only on providing people with access to tools but also on building trust, confidence, and new working habits.

The Emerging Role of the Finance Architect

Before implementing a dashboard, chatbot, forecast model or AI agent, the Finance Architect asks a more fundamental question: What decision are we trying to improve? This stops organisations from starting with the technology and then searching for use cases. Instead, the business problem comes first, with technology to solve it.

This role reflects a broader change in FP&A that, as routine reporting and data preparation become more automated, the value of FP&A shifts towards scenario planning, strategic questioning, storytelling, and business partnering.

Conclusion

The conclusion from the meeting was that if implemented properly, AI raises the standard for FP&A. It frees finance talent from repetitive work and creates an opportunity to focus on higher-value activities.

The organisations that redesign their operating model will not just produce faster forecasts or more dashboards, but they will be better positioned to create decision intelligence, which is the ability to combine data, technology, process, and human judgement to make better business decisions.