AI in ERP: What SAP, Oracle and Microsoft Are Building for Finance Teams

You don’t have to go looking for AI. It’s coming to your ERP whether you’re ready or not. SAP, Oracle, and Microsoft are all moving fast — and the changes to how finance teams work are already arriving.

Here’s what each vendor is building, what’s actually live today, and what it means for finance professionals who use these systems.

SAP: Joule

SAP’s AI assistant is called Joule. It’s built into SAP S/4HANA Cloud, SuccessFactors, Ariba, and other SAP products, and SAP’s stated goal is to make it the AI layer across their entire portfolio.

For finance users, Joule can answer questions about financial data in natural language (“What were our top 5 cost overruns last quarter and which cost centres drove them?”), surface anomalies in transaction data, assist with period-end tasks, and help navigate complex SAP processes that historically required specialist knowledge to operate.

The Q4 2025 release added bidirectional integration with Microsoft 365 Copilot — meaning Joule data can surface inside Teams, Outlook, and Excel without needing to switch into the SAP UI. For finance teams who live in Office 365, that’s a meaningful change to the workflow.

SAP has also embedded AI into their financial planning tools, with predictive forecasting that learns from historical patterns and flags where your forecast assumptions look inconsistent with what the data suggests.

Oracle: Fusion Cloud AI

Oracle has been embedding AI into Oracle Fusion Cloud for longer than the current wave of generalist AI tools. Their approach is to build AI into specific workflows — accounts payable automation, expense classification, anomaly detection in the general ledger — rather than a single chatbot interface.

For project accounting specifically, Oracle’s AI features include automated project cost allocation, predictive project completion forecasting, and intelligent matching of supplier invoices to purchase orders. If you’re managing a project portfolio, these aren’t theoretical features — Oracle has been deploying them with customers for several years.

The more recent development is Oracle’s Digital Assistant for Finance — a conversational AI layer on top of the finance modules that lets users query data and complete tasks using natural language. It connects to Oracle Analytics Cloud, so financial analysis that previously required report-building or a data team can be done through a conversation.

Oracle’s $50 billion infrastructure raise in early 2026 signals they’re positioning themselves as a primary AI cloud provider — which matters because it means heavy continued investment in the AI capabilities being embedded in Fusion Cloud.

Microsoft: Copilot for Finance

Microsoft Copilot for Finance is the most immediately relevant for most UK finance teams, because it lives inside the Microsoft 365 products most organisations already use — Excel, Outlook, Teams, and SharePoint.

The finance-specific features include account reconciliation assistance (it can surface discrepancies and suggest matches), variance analysis in Excel (highlight the significant movements and draft commentary), and financial data extraction from documents and emails.

Copilot for Finance connects to ERP systems including SAP and Oracle via connectors, so it can pull financial data into the Office workflow without manual extraction. The integration with SAP Joule is a good example — you can ask a finance question in Teams and get an answer drawn from SAP data without leaving the Microsoft environment.

The caveat worth noting: Copilot for Finance requires Microsoft 365 Copilot licensing (currently £25-30 per user per month in the UK on top of existing M365 costs), and the ERP connectors vary in depth. The Excel variance analysis features are genuinely useful. The more complex cross-system queries are still maturing.

What this means for finance teams

Three things worth taking from this:

You don’t have to implement anything new. These capabilities are arriving in the tools your organisation already pays for. The question isn’t whether to adopt AI — it’s whether you’re paying attention when it arrives and whether you’re positioned to get value from it.

The value is in the workflow, not the feature. A natural language query feature in your ERP is only useful if you actually use it, know the right questions to ask, and trust the output enough to act on it. Finance teams that invest in understanding these tools will get disproportionately more value from them than those who treat them as something IT manages.

Vendor AI is not the same as generalist AI. SAP Joule and Oracle’s AI features are trained on financial and ERP-specific data. They’ll answer system-specific questions well. For broader analysis, interpretation, and communication tasks, you’ll still get better results from Claude, ChatGPT, or similar generalist tools. The smart approach is using both.