Half of finance teams still take more than five business days to close — and everyone is now being sold a finance-close agent. The value and the safety both hang on one boundary: the LLM orchestrates and drafts; the numbers run in deterministic code with human approval.
SAP went from announcing a copilot in September 2023 to launching an “Autonomous Enterprise” in May 2026. Here is a dated, sourced map of what is generally available, what is announced, and the questions to ask before you buy the narrative.
Everyone’s asking which SAP process to automate next. In 2026 that’s the easy question. The hard one — the one almost nobody is budgeting for — is: who is the boss of this agent, what does it spend, and can someone stop it in the next thirty seconds?
GCC SAP shops that survived ZATCA Phase 2 are tempted to reuse the playbook wholesale for the UAE. That instinct will cost them — the UAE’s architecture is not Saudi’s, and teams that treat this as a copy-paste exercise will spend 2027 firefighting instead of collecting cash.
The AI just closed forty tickets before your coffee cooled and your SLA dashboard is green. That tells you almost nothing about whether the problems stayed fixed — and in an SAP landscape, the gap between “closed” and “solved” is where your next P1 is hiding.
If your go-live plan ends at “train the users on Fiori,” you haven’t decided anything. You’ve deferred the decision to the users — and they’ll decide to keep using SAP GUI.
Steering committees interrogate timelines, budgets, and org-readiness for months. Almost nobody interrogates whether the mock cutover is testing anything real — and that gap is what actually delays go-live.
Everyone is demoing agents that post journal entries, approve POs, and reconcile accounts in seconds. Almost none of them can tell an auditor, six months later, exactly why the agent did what it did — and that gap, not the model’s intelligence, is what decides which agents survive contact with production.
Behind every thirty-second agent demo sit three separate meters — model tokens, SAP AI Units, and the human who has to approve what the agent posts. The one nobody prices is the one that decides the business case.