Imagine asking your finance system, in plain English, to update your fourth-quarter forecast based on the latest sales pipeline and market data and receiving a fully modeled, risk-adjusted answer in minutes. No spreadsheets. No chasing teams for inputs. No delays. That’s not a vision for the future. It’s emerging now.

In June 2025, the AI4Finance Foundation quietly launched FinRobot, an open-source platform designed specifically for finance. What sets it apart is its focus on building AI-native agents for enterprise resource planning (ERP) systems—not generic copilots but agents that understand enterprise financial structures, workflows, and decision rights.

This could be a major breakthrough: ERP has long been the least agile part of the tech stack.

FinRobot uses embedded agents that can analyze and act on real-time data, automate planning cycles, and trigger cross-functional workflows—all within the ERP system itself. The result: Forecasts update continuously, budgets adjust automatically, and finance delivers strategic information in real time to the leadership team.

The ultimate payoff? More responsive planning, less firefighting, and faster decisions aligned with business goals.

AI agents mark the beginning of a new era of intelligent, autonomous finance.

The inadequacy of traditional planning

Legacy budgeting systems are ill-suited for the turbulence of today’s economy. Inflation shocks, supply chain disruptions, and changing customer preferences highlight the risks of relying on rigid, calendar-based planning. Yet many financial planning and analysis (FP&A) teams remain trapped in long planning cycles that can’t adjust to the pace of change. Reports are outdated by the time they reach decision makers. Reforecasting is often just as cumbersome as initial budgeting. That explains why global CFOs rank FP&A as their top transformation priority (see Figure 1).

To navigate volatility successfully, finance teams need five key performance attributes: accuracy, timeliness, flexibility, innovation, and value/cost. Only 13% of the CFOs in our 2022 survey said they consistently achieve that goal (see Figure 2). But technology is poised to close the gap. By the end of 2024, 35% of companies had adopted generative AI in finance or were considering it. And over the last six months of 2024, the pace of adoption in finance was among the fastest for all business functions.

Generative and agentic AI

Generative AI helps humans interpret and interact with data; agentic AI makes decisions based on data and manages systems autonomously.

Many companies are deploying machine learning (ML) in financial forecasting, with 28% of finance teams now using some form of ML in quarterly planning, according to the Association of Financial Professionals. These tools are improving prediction accuracy by identifying patterns across large data sets. But ML has its limits. It often requires significant up-front tuning, structured input data, and technical oversight.

That’s where generative AI and agentic AI enter the picture, each bringing complementary capabilities to forecasting.

Generative AI is often deployed to generate and summarize content. But in forecasting, it’s proving to be a game changer. One of its most valuable uses is synthesizing signals from reviews, news, and internal messages, turning messy text into forecasting-ready variables in minutes.

Generative AI also enhances forecast interpretability through tools such as retrieval-augmented generation. Imagine a financial analyst wondering why third-quarter revenue is projected to fall. Instead of digging through spreadsheets and models, the analyst can ask an AI interface and receive a plain-language answer that references the model’s assumptions, input changes, and historical trends. This level of explainability builds trust and enables self-service planning across business units.

Perhaps most important, generative AI allows interactive, scenario-based forecasting. Users can input “what if” queries such as “What happens if we reduce marketing expenses by 10%?”—and receive real-time modeled outcomes. This interactivity enables a new kind of strategic agility in which planning becomes a continuous, collaborative process, not a quarterly exercise.

Agentic AI, by contrast, refers to autonomous systems that manage entire forecasting workflows. These agents do not merely respond to queries; they act. One agent may clean and ingest data, another might select the appropriate forecasting model, a third can generate outputs, and a fourth might trigger alerts or even propose budget reallocations.

Microsoft is a good example. Within its finance organization, AI agents are starting to power core FP&A functions such as forecasting, variance analysis, reconciliation, and reporting. Forecasting agents have replaced Excel-based modeling with a no-code ML platform. Reconciliation agents automatically match financial records for each account, reducing cycle time from hours to minutes. Analyst agents interpret causes, build visual dashboards, and draft executive narratives.

These systems are tightly integrated into Microsoft 365, enabling analysts to access forecasts, insights, and actions within Excel, Teams, and Outlook. For example, an analyst can summarize inbox trends, simulate scenarios, and generate reporting content via Copilot-enabled agents. Microsoft’s AI agents highlight a deeply embedded and scalable model of agentic finance.

Together, generative and agentic AI aren’t just improving forecasts; they’re redefining the entire realm of forecasting.

Of course, autonomy doesn’t mean human abdication. As organizations scale AI adoption, governance remains essential, particularly around data provenance, model oversight, and decision accountability. AI agents must be auditable, bias-tested, and aligned with the enterprise’s risk posture. The goal is augmented intelligence, not unchecked automation.

Three ways to modernize planning

So how do companies put these new tools and ideas into practice? The path forward isn’t one-size-fits-all. Organizations can modernize FP&A in three ways: streamlining existing processes and data, enhancing them with AI, or reinventing planning from the ground up.

Streamlining

Streamlining focuses on simplification and speed. Many companies spend months building annual plans that are obsolete by the time they’re complete. By trimming unnecessary layers of detail, sequencing tasks more logically, and introducing automation for tasks such as reconciliations, organizations can dramatically compress the planning timeline.

One critical element is getting the data foundation right. While many AI forecasting efforts begin in operations, their impact often reverberates throughout finance. A transformation at global power management company Eaton illustrates how integrating real-time data across the supply chain enables faster, more accurate decision making in both production and financial planning. 

To address fragmented data systems that hindered accurate forecasting and supply chain efficiency, Eaton implemented Palantir's Artificial Intelligence Platform. The platform integrated data from more than 72 ERP systems, encompassing more than 300 plants and 32 million unique parts. That comprehensive approach provided real-time insight into supply chain operations, allowing Eaton to quickly identify and resolve material shortages, prevent downtime, and unblock the assembly of finished goods. The example underscores a vital lesson for finance and planning leaders: Before layering on intelligence, organizations must ensure that their data is unified, structured, and trustworthy, laying the groundwork for effective AI-powered forecasting.

Enhancing

Enhancing planning with AI (especially generative AI) enables richer insights and faster feedback. A global consumer products company, for example, used traditional ML to cut the time required to prepare a revenue forecast from two weeks to two hours. Accuracy rose to greater than 97%. Analysts no longer need to compile decks or interpret variances manually. AI does that groundwork, empowering teams to focus on strategic choices. Now the company is considering integrating generative AI to simulate scenarios, flag deviations and generate narrative summaries, including weekly reports and reallocation suggestions.

Reinventing

Reinvention requires rethinking the operating model altogether—well beyond technology. Some companies are moving away from fixed annual budgets in favor of rolling forecasts and event-triggered planning. Take Hilti, for example: It replaced static budgets with three rolling forecasts per year and relative performance metrics. Its bonus system is tied to external benchmarks, not internal targets. While Hilti’s planning system remains human-led, it is built for responsiveness, allowing the organization to adapt quickly to macroeconomic shifts, industry disruption, or internal innovation.

Hilti’s reinvention began in 2006, proving that bold process redesign can unlock agility even before the emergence of today’s AI technologies. By designing financial systems around real events (not the calendar), Hilti exemplifies what agile FP&A can look like.

Reshaping finance

Dynamic planning is rapidly becoming best practice. The gap is widening between companies stuck in rigid, calendar-based cycles and those moving fast with intelligent, always-on planning. The challenge is how quickly companies can rewire their planning approach to harness the full power of AI.

Leaders who experiment boldly with AI and design their systems for adaptability will be the ones defining finance’s next decade.