SAP now ships dozens of Joule agents that can post transactions on their own, and Microsoft wired Copilot into the mix. Readiness research says most companies are about to hand power tools to master data that cannot carry the weight.
At Sapphire in May, SAP put the phrase "autonomous enterprise" on its main stage, per the SAP News Center. The pitch has hardware behind it: Joule has grown into an agent platform with more than 40 specialized agents and over 2,400 skills across finance, procurement, HR, and supply chain, and Joule Studio went generally available this year as a low-code agent builder. These agents do more than answer questions. They create purchase orders, post goods receipts, and run financial simulations.
Microsoft used the same conference to announce agent-to-agent integration between Microsoft 365 Copilot and Joule, per the Microsoft Azure blog, so agents on both sides of the fence can hand work to each other. Oracle ships its own agents, concentrated in finance workflows. Every major ERP vendor now sells some version of the same sentence: the system will act on its own.
Here is the part of the keynote that applies to you specifically: the agent will act on your data. Everything else in the demo was theirs.
1. The readiness gap, in numbers
E3 Magazine's Agentic AI Readiness Index 2026 found that only 15 percent of companies are fully prepared to run agent-based AI in production, while roughly 60 percent report investing tens to hundreds of millions in the category. The same body of research notes that 41 percent are running agentic AI in production anyway, on top of known gaps in data reliability and governance.
Read those numbers together. The money is committed, the deployment is happening, and the preparation is missing in most of the buildings where both are true.
Gartner added a sharper warning in a May 26 press release: applying uniform governance across all AI agents leads to failure. The distinction that matters is between an agent's ability to act and the scope of access it is granted. An agent that can post journal entries and an agent that summarizes reports should live under different rules, and most rollouts give them the same badge.
2. An agent is speed, and speed amplifies what is underneath
The demo works because the demo data is curated. Production is where your actual vendor master lives, with its duplicates, its stale banking details, and its three spellings of the same supplier. An agent posting transactions against that master does what your team used to do, at machine speed, without the veteran clerk's pause that has quietly saved you a dozen times a year.
| What the demo shows | What production will show |
| A clean, curated vendor master | Duplicates, stale bank details, three spellings of one supplier |
| The happy-path purchase flow | The exceptions your team handles from memory |
| An agent with tidy demo permissions | The role model you have not redesigned since go-live |
| Instant, reversible actions | Month-end close, where reversals have consequences |
3. Five audits before you switch anything on
Master data ownership. Vendor, customer, item. Every domain gets a named owner and a dedup pass before any agent touches it. If nobody owns the vendor master, the agent does now, and it did not ask for the job.
Category structures a human can read. Your chart of accounts and product hierarchies encode decisions. If people cannot explain the structure, an agent cannot respect it, and it will file things with total confidence in the wrong place.
Agent identity and least privilege. Give each agent its own identity, scope access to the minimum, and apply Gartner's split: what it can do and what it can reach are separate decisions, made separately.
Exception paths on paper. Agents automate the happy path. Your business runs on exceptions. Document them before go-live, because the undocumented ones will be handled by a machine that does not know they exist.
Logging and reversibility. Every agent action attributable, every action undoable, tested before launch. When the auditor asks who posted the entry, "the ERP did" needs a better answer behind it.
4. Then earn autonomy in stages
Run one process in shadow mode with a human approving every action. Widen to supervised autonomy on low-risk transactions. Grant real autonomy only where the logs prove the agent has earned it. Crawl, walk, run. The vendors solved capability this year. Readiness was never theirs to solve. It is yours.
Final thought
The autonomous enterprise is a real direction, and the tooling arriving this year is legitimately impressive. But an agent inherits the data estate you actually have, and no keynote can audit your vendor master for you. Do the unglamorous work first, and the agents get to be the productivity story instead of the incident report.
AI, Properly is the Ayvede field report: AI news read plainly, with a takeaway you can use. Ayvede assesses AI and data readiness before your firm switches anything on. Advise. Vet. Deliver. Ask a question in writing.