The profitability model reconciled to the cent, yet five of six reporting views showed zero. How hypothesis-led problem solving found the cause in one run, and how to apply the method to your next decision.

Having a model and being able to defend it are two different things. Nine questions, one per governance dimension, to check in an afternoon whether your FTP, profitability and allocation models would hold up. Part 3 of the model governance series.

AI agents are already running finance models. The platform governs access; someone has to govern the logic they execute. Why that matters for the Agentic Office of the CFO, where SR 26-2 leaves banks, and five practices to start with.

One FTP curve feeds profitability, pricing, incentives, RAROC and the budget. Each link has an owner. The chain usually has none. How to map it in a week, and the signs it needs attention.

SR 26-2 pulls your FTP curve into the model inventory and pushes many cost allocation engines out. The risk stays. A Finance reading of the new guidance, model by model, with five questions to ask this quarter.

IPM Insights tells you where a margin moved. A CFO wants to know why. How to build a conversational layer on Oracle EPCM where the model computes, the agent explains and every answer shows its source.

AI in Oracle EPCM reads the structure and numbers of a cost model. It doesn’t judge whether the logic is sound. Five design flaws AI makes worse, and a five-question check to run before you switch it on.

IPM Insights flags margins that moved more than expected. On a profitability model, many of those alerts are model mechanics. Here is how to configure Insights so the alerts that reach the business are worth reading.

The PCM Agent turns structured English into EPCM modeling actions. Here is what it does well, where it stops, and how to structure a waterfall so your requests produce the rules you meant.

Oracle EPCM now ships a PCM Agent, IPM Insights and generative summaries. Here is what each one does, what is still missing, and why every AI feature is only as good as the model underneath.

A CFO showed me his AI dashboard — models, anomaly detection, the works. Then I asked how many committee decisions it had actually changed. None. That's the trap: measuring AI value in finance by hours saved instead of decision quality. Here's what to track instead — and why regulators will soon demand it.

Twelve executives approved the same profitability model. They had each signed a different one. EPM implementations don't create the misalignment between FP&A and BP&A, they expose it during the build, at the worst possible cost. Why the dimensional model is the forcing function that surfaces the disagreement your deck hid for months, and how to settle the definitions before the cube does it for you.

How Banco de los Trabajadores (BANTRAB) built advanced profitability and cost management on Oracle EPCM with Asher & Company — cutting model build cycles by 37%, decision cycles by 75%, and manual data prep by 64%, with Sáasil (powered by Anthropic) closing the gap between financial experts and the architecture.

67 essential Oracle EPCM terms across 8 families — every definition you need plus the practical context we apply on real implementations. From cubes and allocation rules to genealogy, traceability, and methodology under D2V™. Click any element to see how it shows up in production.

In 2025, AI was cited as a factor in 55,000 layoffs. In 2026, hyperscalers will spend $700 billion on AI capex. And 95% of enterprise AI pilots delivered no measurable P&L impact. These three numbers tell three stories — and bundling them as cause and effect is the most expensive narrative error of the year.