EPM 3.0: model comprehension
"Hidden knowledge" and long-forgotten artefacts are said to be the main obstacle to AI in finance. Paradoxically, AI might be the remedy. Its overlooked value isn't automation. It's the accumulation of knowledge. EPM 3.0 is when AI doesn't just run inside the model, but understands it.
I recently saw a post on how important a precise description of FP&A processes is for an FP&A agent to operate. Then another, saying that "hidden knowledge" and long-forgotten artefacts are the main obstacle to using AI in finance workflows.
OK, but that's not really the point.
Paradoxically, AI might be the remedy for this.
Accumulation, not automation
What does today's AI actually do? It collects knowledge and acts on it, faster than a human can. It's why AI is so good at coding, where patterns matter. An often-overlooked characteristic.
Everyone strives for "intelligent automation." But the real value of AI in financial modelling, planning, reporting and analysis isn't automation. Automation is, in the end, only a surface effect.
The real value, the one we all overlook, is the accumulation of knowledge.
Model comprehension
So think it through. If my agent knows all my models, rules, hierarchies, allocations, dimensionality, flows and dependencies, and it gets the right tools, it can analyse the whole model faster than humans. The same way it produces code faster than coders, who also happen to be humans.
Model comprehension. Including all those dark legacy corners that can finally be understood, and cleared.
EPM 1.0: the IT era
Stored the numbers.
EPM 2.0: the cloud era
Let finance shape the numbers. Sold as agility, delivered as bureaucracy.
EPM 3.0: different
The model itself becomes legible.
EPM 3.0 is when AI doesn't just run inside the model, but understands it: what every driver does, how every formula chains, why this quarter's plan disagrees with last quarter's actuals, at the structural level.
A perfect mix
Add to this an external knowledge layer, like the recent Anthropic financial agents, and we get a perfect mix, don't we?
The next planning platforms won't compete on features. They'll compete on whether their model can be read.
Knowledge is the real value
An agent that comprehends every driver, formula and dependency in your model, dark legacy corners included, paired with the right tools to act. That's the layer we're building with Novi. Join early access to follow the work.