AI agents in ERPs: what SAP, Microsoft, and Oracle are changing in 2026

In six months, the narrative from ERP vendors has shifted. In January 2026, SAP announced the integration of Joule, its AI assistant, with Microsoft Copilot. Oracle responded in March with 22 agent-based applications delivered all at once. Microsoft transformed Copilot Studio into an autonomous agent platform in April. Then, at Sapphire in May 2026 in Orlando, SAP unveiled its vision for the "autonomous enterprise," featuring over 200 AI agents orchestrated by Joule. And on September 23, 2026, Microsoft gave the category a name by formalizing its vision for "agentic ERP" for Dynamics 365. Five announcements, one consistent message: AI will no longer be a product alongside your ERP; it will live inside it.

"40% of enterprise applications will integrate task-specific AI agents by the end of 2026, up from less than 5% in 2025." Source: Gartner, August 2025

Why are AI agents in ERPs arriving so quickly?

Because the three major vendors have just delivered their agentic layers almost simultaneously, and Gartner predicts that 40% of enterprise applications will integrate specialized AI agents by the end of 2026, compared to less than 5% in 2025. The pace over the last few months is unprecedented: the merger of SAP Joule with Microsoft Copilot in January 2026, 22 agent applications from Oracle in March, and the transformation of Copilot Studio into an autonomous agent platform in April, according to a review by My Business Future.

And the pace isn't slowing down. At Sapphire 2026, SAP announced over 200 AI agents in its portfolio, with Joule positioned as the central orchestration layer and a new interface, Joule Work, designed to replace screen-based navigation, according to coverage by SAPinsider (May 2026).

The most recent volley came on September 23, 2026: Microsoft formalized the term "agentic ERP" for Dynamics 365, featuring an account reconciliation agent that resolves exceptions in batches and a finance agent whose skills—reconciliation, variance analysis, and collections—extend into Copilot, Excel, Outlook, and Teams, according to Microsoft's announcement (September 2026).

The delivery method matters as much as the technology. Joule agents are arriving within S/4HANA Cloud licenses. Microsoft agents are being deployed through Dynamics 365 and Copilot Studio. Oracle’s Fusion Agentic Applications are being activated for existing Fusion Cloud customers. In other words, many organizations will receive AI agents without having asked for them. The issue is landing on the desks of IT and finance departments before a budget has even been approved.

What exactly is an AI agent in an ERP?

A copilot answers your questions; an agent takes action: it monitors transactional data, makes decisions within defined rules, and executes tasks within the system. To understand the two technology families involved, see our article on the difference between traditional AI and generative AI.

Examples from the first releases: a procurement agent that identifies an imminent stockout, checks alternative suppliers, and prepares a purchase order proposal; a finance agent that reconciles invoices with purchase orders and only escalates discrepancies; an HR agent that applies replacement rules before approving time off. Nothing spectacular. That is precisely the point: agents target the repetitive work that currently occupies skilled staff.

Are the AI agents being delivered actually used in production?

Very little for now: according to a survey by DSAG, the German-speaking SAP user group (February 2026), 43% of SAP customers have AI use cases in production, but 77% of those scenarios run on non-SAP solutions, compared to 3% on SAP AI. Agents are included in licenses; their actual usage, however, remains marginal.

The explanation can be summed up in one sentence. Trying Copilot or ChatGPT on a text requires nothing; connecting an agent to the transactional core of an ERP requires clean master data, documented processes, and governance rules. The first step is easy; the second is steep.

The North American trend is following the same path. Our analysis of AI adoption in Canadian businesses shows that adoption has tripled in two years, driven primarily by peripheral uses like text analysis, long before transactional processes.

Key benchmarks to keep in mind:

• 40% of enterprise applications will integrate specialized AI agents by the end of 2026, up from less than 5% in 2025 (Gartner, August 2025);
• more than 40% of agentic AI projects will be abandoned by the end of 2027 due to a lack of clear value or sufficient risk controls, according to Gartner (June 2025);
• over 200 AI agents announced by SAP at Sapphire 2026 (SAPinsider, May 2026);
• 43% of SAP customers have AI use cases in production; 77% of these scenarios run on non-SAP solutions, 3% on SAP AI (DSAG, February 2026);
• 19.2% of Canadian businesses use AI, and the proportion reaches 40.4% in finance and insurance (Statistics Canada, Q2 2026).

Methodological note: The DSAG survey covers 198 SAP client companies in Germany, Austria, and Switzerland, surveyed from December 2025 to January 2026. Canadian figures are sourced from the Canadian Survey on Business Conditions. Neither of these surveys describes your specific situation.

What are the risks of activating agents on poorly managed processes?

An autonomous agent connected to outdated master data or an unclear process accelerates errors instead of fixing them. Three questions to resolve before activating anything: what decisions can the agent make on its own, at what dollar amount does a human need to confirm, and who is responsible when the agent makes a mistake? Approval hierarchies and audit trails already exist in your ERP; agents must inherit them, not bypass them. We have detailed this framework in our four governance controls for AI agents in finance.

Caution is not a sign of timidity. Gartner predicts that more than 40% of agentic AI projects will be abandoned by the end of 2027 due to rising costs, unclear business value, or insufficient risk controls, according to the forecast published by Gartner (June 2025). The same firm describes a phenomenon of "agent washing": out of the thousands of vendors claiming to be agentic, only about 130 actually are.

The 2026 bottleneck is no longer the delivery of agents. It is the number of processes clean enough to receive them.

There is also a commercial angle to keep in mind. Joule agents require S/4HANA Cloud: ECC and on-premise clients do not have access, which adds very real migration pressure. The vendor sells the agent, and the certified integrator sells the migration. To distinguish what is useful from what is forced, an independent perspective helps; we have already explained why a vendor-certified integrator cannot provide neutral advice.

How can you prepare your ERP for AI agents without a complete overhaul?

By treating the arrival of agents as a process project: a defined scope, clean master data, written governance rules, and a measured pilot before any scaling. The approach consists of five points:

• inventory candidate processes and choose a bounded use case, such as invoice reconciliation or inventory monitoring;
• validate the quality of master data: duplicates, inactive vendors, and up-to-date approval rules;
• write governance rules before deployment, not after;
• measure before and after: exception rates, turnaround times, and hours saved;
• train teams: managers become pilots of agents and their exceptions.

At PlanAxion, this is exactly the purpose of a scoping workshop: aligning stakeholders, inventorying use cases, prioritizing them based on value, complexity, and data maturity, and then leaving with a defensible roadmap.

Which process should you start with?

AI agents in ERPs are no longer just a conference promise: they are in the licenses you are already paying for, or in those you will be offered at your next renewal. However, the DSAG figure reminds us that delivery does not equal adoption. The choice that remains in your hands is which process you entrust to them first. Pick one that causes pain every month, measure it, put guardrails in place, and then decide on the next steps based on the numbers.

Frequently Asked Questions

What is an agentic ERP?

An agentic ERP is an integrated management system where AI agents execute portions of processes, such as account reconciliation or invoice matching, under human supervision. The term, formalized by Microsoft in September 2026 for Dynamics 365, describes a direction shared by major vendors, not a mature category.

What is the difference between a copilot and an AI agent in an ERP?

A copilot answers questions and suggests actions: it waits for a request. An AI agent acts on its own within defined rules: it detects a situation, makes a decision, executes the action in the ERP, and then escalates uncertain cases to a human.

Why do so few companies use their ERP vendor's AI in production?

Because AI at the heart of an ERP requires more than office-based AI: master data quality, governance, and process integration. The 2026 DSAG survey measures the gap: 43% of SAP customers have AI use cases in production, but 77% of those scenarios run on non-SAP solutions.

Do you need to migrate to the cloud to get AI agents in your ERP?

Often, yes. SAP Joule agents require S/4HANA Cloud, Microsoft agents run through Dynamics 365, and Oracle delivers them in Fusion Cloud. On-premises systems are generally left behind. However, migration should be decided based on your business case, not just pressure from licensing.

Will AI agents replace finance or procurement teams?

No. Initial use cases target repetitive work: invoice reconciliation, inventory monitoring, and routine validations. Teams will shift their focus to handling exceptions, managing agents, and analysis. Human judgment remains necessary, especially for decisions above approval thresholds.

Where should you start with AI agents in your ERP?

Start with a limited, measurable pilot on a single process, using validated master data and written governance rules. Measure the before-and-after, then scale if the numbers hold up. A scoping workshop is often enough to prioritize use cases.

Qualifying your first AI use case in your ERP

Before adding an AI agent or feature to your ERP, you must define the process, the accessible data, the controls, and human accountability. Our enterprise AI workshop helps prioritize use cases; our artificial intelligence consulting services then cover the pilot and its integration. Describe the process you want to improve and your current ERP to determine where to begin.