Stephen Hambling talks on Kybos Diary of a CFO about the three types of Artificial Intelligence (AI) in Enterprise Performance Management (EPM) systems that are changing Financial Planning & Analysis (FP&A) forever.
If you’re drowning in AI buzzwords but still not clear what it really means for FP&A in your EPM system, you’re not alone. Most finance teams are being told that AI will “transform the function”, but there are very few explanations that actually cut through to what that means in day‑to‑day planning, forecasting and reporting.
That’s why I’ve put together a straightforward, no‑nonsense article and video that strips out the marketing spin and focuses on what really matters for finance leaders that provides practical, real‑world use cases that you can understand, challenge and ultimately apply in your own environment. No data‑science jargon, no over‑hyped promises, just clear explanations, framed from one finance professional to another.
The focus is on how AI can support the work you already do in your EPM platform: improving forecast accuracy, speeding up variance analysis, simplifying scenario modelling and helping you communicate the story behind the numbers to your board and operational stakeholders.
Watch to see where AI can genuinely add value in planning, forecasting and analysis, and where, for now, it’s just background noise. The aim is to give you enough clarity to ask the right questions of your vendors and internal teams, to separate credible capability from the hype and to build a realistic view of what you can achieve over the next 6–24 months.
AI capabilities broadly fall into three key areas:
1. Predictive analytics / machine learning
This is the most established use of AI in EPM. Traditionally referred to as machine learning, it uses historical data to predict future outcomes. These models can process very large data sets with multiple variables and dimensions, identifying patterns and relationships that are hard to detect with traditional human analysis or standard Excel work.
You might, for example, feed a model with three to six variables and it will identify correlations that aren’t immediately obvious from conventional reporting. This can be extremely powerful, but it does depend on having the right environment: a business with sufficient data volume, a degree of predictability, a robust history and, critically, clean, well-structured data. Core machine learning is already a proven, practical application of AI for forecasting and analysis, and remains the most mature and predictable area.
2. Large language models (LLMs)
The second use case has accelerated rapidly over the last six months, in EPM and beyond. Large language models are advancing at pace, and their application in EPM systems is now very real.
In practice, this means you can interact with your EPM system in natural language—typing or speaking—to ask the AI to perform tasks and return answers in plain English. For example, you might have a standard cost centre report showing actuals vs budget and variances across multiple cost centres or a full P&L. Instead of manually analysing the data, you can ask the AI to:
– “Summarise this data”
– “Provide an analysis of the variances”
– “Highlight anomalies or unusual movements”
The experience is similar to using ChatGPT: you receive a narrative summary in natural language that you can reuse for board packs, variance commentary, or management reporting, typically produced today in Excel and PowerPoint.
These capabilities are improving quickly, driven by rapid development in the underlying LLM platforms (such as those from OpenAI and others). EPM vendors typically connect securely to these large foundational models rather than building their own, as that is more economical and effective.
We’re already seeing functionality such as “provide a summary” evolving into “provide a detailed analysis”, going deeper into the data and offering contextual hypotheses about why certain numbers are up or down. This is important because analysis ultimately exists to support decisions, identify trends and understand drivers. LLMs are increasingly helping to suggest possible reasons for changes and to spot emerging patterns.
As more business and industry context is fed into the models, they will become more capable of explaining the “why”, not just the “what”. Knowing that revenue has moved is useful; understanding why it has moved is essential. Expect this area to develop quickly over the coming months, not years, and become more deeply embedded in day-to-day EPM workflows.
3. Agentic AI
The third area is agentic AI. This is where AI is connected directly to systems or processes so that it can not only analyse data but also take actions, effectively doing what a human user would do inside the process.
This is already gaining momentum. There are strong examples in marketing technology, for instance. We use HubSpot, which has had effective agentic AI features for some time. In software development, AI can already write and modify code on your behalf with minimal human input, fundamentally changing how that industry operates.
The same pattern is now emerging in EPM. Instead of simply asking, “Give me a summary of this data”, you might say:
– “Take my existing forecast and increase this cost line by 3%”
– “Improve that revenue stream by 5%”
– “Adjust the forecast so that profit reaches £X million by changing these specific drivers”
– “Create three alternative scenarios based on these assumptions”
The AI will then go away and build those scenarios for you. While this is still relatively early in EPM, it is progressing quickly and is likely to advance significantly over the next year.
Crucially, this opens up new use cases for non-finance stakeholders. Operational managers, for example, could adjust their forecasts using natural language inputs instead of manually changing fields and drivers in the system. That makes sophisticated planning and scenario modelling far more accessible across the business.
What you need in place to benefit
To take advantage of any of this, you first need an appropriate platform. If you are still heavily reliant on Excel alone, your ability to leverage AI for forecasting and planning will be extremely limited. You need a modern EPM system, with AI capabilities built in, to realise the benefits.
Just as importantly, your data must be properly structured. AI performs significantly better with well-organised, consistent data and a solid historical record. If your data is messy, incomplete or poorly contextualised, the quality of AI-driven insights will suffer.
If you’re not already on this journey, you need to move quickly. In this case, the risk of being left behind is real, and we will see a widening gap between organisations that embrace modern EPM with AI and those that do not.
Where the real value lies in FP&A AI
AI will automate parts of the build process including report creation, layout changes and standard analyses, and will allow managers to generate and explore scenarios more quickly. That will save some time. But building reports is not usually the biggest time sink in finance.
The real value comes from freeing up time to understand the underlying business story. A good EPM system already enables fast production of reports and analysis. The time you gain should be used to investigate why numbers are moving.
For example, suppose sales are up in Singapore. You can show that on a variance report or a chart almost instantly. The crucial question is: why? You only get to the real answer by speaking with the local team, understanding whether this is a one-off event (e.g. a weather-driven spike) or a genuine new market development. That real-world context is what decision-makers actually need.
A modern EPM platform enables that by giving you the time and structure to focus on understanding the business, not just producing numbers. AI then enhances this further, but without a robust core EPM solution, AI on its own will not deliver meaningful value.
In summary
– AI is evolving rapidly, and EPM is no exception.
– There are now clear, practical AI use cases in EPM: predictive analytics, natural language analysis via large language models, and emerging agentic AI that can actually execute planning and forecasting actions.
– To benefit, you must be on a modern EPM system with AI on its roadmap and embedded into its architecture, and you must have your data and core models set up correctly.
– When selecting a system, ensure it has a clear, credible AI roadmap. For leading vendors, AI now represents a substantial part of their product strategy. If a provider cannot show you a roadmap that is heavily AI-focused, that should raise concerns.
– The leading EPM platforms are moving quickly in this space, Jedox included. This landscape will continue to change; in six months’ time, we could easily revisit this topic and see significant advances again.
The practical order of priorities is:
1. Get onto a modern EPM system like Jedox.
2. Ensure it has robust, inbuilt AI capabilities and a strong AI roadmap.
3. Structure and clean your core data and models properly.
4. Then, progressively take advantage of the AI features as they mature.
What you cannot do is wait passively. The pace of change is simply too fast.
About Stephen Hambling,
CEO and Founder of Kybos, a UK based Jedox partner.
I provide simple, clear, commercial forward looking financial support to founders, entrepreneurs and CFO’s of Small and Medium sized companies. Typically I’m the special projects guy, commercial troubleshooting, systems help, M&A or cashflow & turnarounds. I’ve implemented several hundred finance systems. Implementations are not a core skill set for the average finance team, so it’s easy for a project to go off track. Clear simple guidance before, during and after an implementation can make the world of difference.
Other content you may like
Aout Kybos
Kybos is a dedicated UK Jedox gold partner. We build planning and analysis solutions that deliver value fast using accountancy qualified consultants. Whether you want a fully customised application or to build upon an existing solution, Kybos consultants are here to help.