If Claude Can Talk to My ERP, Why Do I Still Need EPM?
A Pivot2 perspective on Enterprise Resource Planning (ERP), Enterprise Performance Management (EPM) and AI
There’s a question we need to get better at answering
If Claude or ChatGPT can connect directly to my ERP, interrogate my financial data and answer questions in seconds, why do I still need EPM?
It’s a fair question. And the answer can’t simply be “because EPM has AI too.”
The ERP just became much easier to talk to
NetSuite is a good example of where this is heading. With NetSuite’s AI Connector Service and MCP, AI clients can increasingly interact directly with ERP data using natural language.
Instead of running a report (or asking a junior analyst to run a report!), exporting it to Excel, manipulating it and then explaining it, we’re moving rapidly towards a world where a finance leader can simply ask:
“Show me revenue by customer for the last six months, identify where margin is deteriorating and tell me what I should investigate.”
That is genuinely powerful. And it raises an uncomfortable question for the EPM industry.
Does every organisation still need a separate EPM platform? Probably not.
Not every company needs EPM
If your business has a relatively simple structure, a small finance team, straightforward budgeting and forecasting, and most importantly, only a handful of people contributing to the process, ERP + Excel + AI might be all you need.
AI is making that combination significantly more capable.
In many ways it’s already past the point where asking questions about the numbers isn’t actually the hard problem anymore.
Creating the numbers is.
Imagine asking:
“What happens to our forecast cash position if revenue finishes 8% below forecast, we delay six hires and our major project slips by three months?”
Any AI model released today can reason about that scenario.
But first it needs to know: Which forecast are we using? Which workforce assumptions have been approved and how many staff have already been onboarded? How does delaying those hires affect salary and on-costs? How does the project delay affect revenue recognition and cash? What happens to capex and depreciation? What exchange rates should we use?
And once those assumptions change, how is that answer reflected in the P&L, balance sheet and cash flow?
That’s not really an AI problem. It’s a financial modelling and process problem. And that’s where EPM starts to shine.
ERP records what happened. EPM records what happens next.
This is probably the simplest way I’ve come to think about it.
An ERP is fundamentally very good at recording what has happened.
EPM is designed to help the organisation record and model what happens next.
Budgeting. Forecasting. Workforce. Revenue. Projects. Capex. Cash. Allocations. Consolidation. Profitability. Scenarios.
AI can make both worlds dramatically easier to interact with, but it doesn’t remove the distinction.
If I ask Claude why revenue declined last quarter, it can retrieve the transactions and help analyse them.
But if I ask:
“What happens if APAC revenue is 10% below plan for the next six months?”
The starting point of that question is the current forecast, the current assumptions and importantly, an organisational view on how that should be modelled.
EPM gives AI something incredibly important: context
This is where I think the value proposition for EPM is changing.
For years every CFO has talked about the want to reduce the number of load bearing spreadsheets in the planning process and that has been the key goal of every EPM project. But I’m not convinced that’s the most compelling story anymore.
Excel isn’t disappearing. ERP reporting is improving rapidly. And AI makes all of those tools more powerful.
The bigger opportunity is to create a governed financial model of the business that AI can reason over.
A good EPM environment contains far more than numbers. It contains business logic.
It knows how accounts roll into KPIs. It knows how employees and vacancies affect workforce costs. It knows how revenue is calculated. It knows how project timings affect revenue, cost and cash. It knows how entities consolidate. It knows which assumptions belong to which scenario. It knows the organisational hierarchy. It knows which forecast is being worked on and which has been approved. And it knows how changing an operational assumption flows through to a financial outcome.
That’s business context.
And context is what makes AI substantially more useful. And without it, even the best AI models are simply guessing.
AI doesn’t make the finance foundation less important
AI makes asking sophisticated questions incredibly easy.
But a beautifully articulated answer based on the wrong forecast, an outdated hierarchy or inconsistent assumptions is still the wrong answer, even confidentially delivered.
Finance has always cared about trust.
Where did this number come from? Which assumptions were used? Can I reconcile it? Can I explain it to the Board?
AI doesn’t make those questions disappear.
So who actually needs EPM?
I don’t think the answer should necessarily be based on company size.
It’s about complexity.
A $20 million business with one entity, a straightforward revenue model and five people involved in budgeting might cope perfectly well without a sophisticated EPM environment.
An organisation with multiple entities, complex workforce planning, projects, capex, allocations, rolling forecasts and 50 budget contributors still needs a structured model.
The warning signs are familiar: multiple versions of the forecast; finance spending days consolidating spreadsheets; budget owners working from different assumptions; workforce planning disconnected from the financial forecast; difficulty producing an integrated P&L, balance sheet and cash flow; constant reconciliation between management reporting and the ERP; and scenario modelling that requires someone to disappear into Excel for three days.
That’s not fundamentally a reporting problem.
It’s a finance operating model problem.
Maybe we need to stop selling EPM as a better spreadsheet
This is where I think the EPM industry including us at Pivot2 needs to challenge some of our own thinking.
For years, one of the central messages has been:
“Get your planning out of Excel.”
But perhaps the more relevant question now is:
“Do you have a financial model of your business that is good enough for AI to reason over?”
Because we’re rapidly heading towards a world where access to AI itself won’t be particularly differentiated.
Most organisations will have it. Most finance teams will be able to ask questions of their ERP. Most executives will have an AI assistant capable of analysing information.
The differentiator will be what sits underneath it.
Is the data trusted? Are the assumptions governed? Is the financial model connected? Can operational changes flow through to financial outcomes? Can finance explain where the answer came from?
ERP + EPM + AI
I don’t think the future is EPM versus AI.
They solve different parts of the problem. ERP tells us what happened. EPM provides a governed model of what happens next.
AI makes it dramatically easier to understand both and decide what to do about it.
And that is where I think the opportunity gets really interesting.
AI isn’t necessarily the thing that replaces EPM. But it might finally be the thing that makes all the work sitting underneath EPM the models, metadata, assumptions, integrations and business rules far more valuable.
Because once asking the question becomes easy, the quality of what sits underneath the answer becomes everything.
If you’re thinking about what AI means for your EPM roadmap, or how to make the most of the foundations you already have in place, get in touch with the Pivot2 team. We’d love to talk it through.
McKinsey & Company (2023). Rewiring planning for a more agile finance function
Oracle (2024). The future of planning: Trends in cloud EPM adoption
FSN Modern Finance Forum (2022). The Future of Planning, Budgeting and Forecasting
Deloitte (2023). Finance 2025: Digital transformation in the finance function
Gartner (2024). Market Guide for Cloud Financial Planning and Analysis Solutions