Every major utility software vendor now leads its marketing with AI. That makes comparison harder, not easier, because “AI” covers everything from a help-text chatbot to an autonomous agent that posts transactions. This post cuts through the announcements to show what SAP, Oracle, and Cayenta have actually shipped, what is still on a roadmap, and which capabilities map to which utility sizes.
Why the AI comparison matters now
The 2026 procurement cycle is the first where AI capabilities are a genuine differentiator rather than a checkbox. Utilities replacing aging CIS platforms are evaluating ten-plus-year commitments, and the AI architecture a vendor chooses today shapes what the platform can do in 2030. That is reason enough to compare on AI specifically, separate from the deeper platform reviews in our Oracle vs SAP utilities comparison and the Cayenta CIS review.
SAP: Joule, Industry AI, and the autonomous enterprise push
SAP’s AI story runs through Joule, its AI copilot and agent layer embedded in S/4HANA and the SAP Business Technology Platform. The scale of the buildout is significant: as of Q1 2026, SAP reported more than 30 Joule agents and 2,500-plus skills. By the Sapphire conference in May 2026 that count had grown to 200-plus agents and 50-plus assistants, spanning finance, supply chain, and industry verticals.
For utilities specifically, SAP announced the Utilities Customer Self-Service Agent as generally available in Q4 2025. That agent handles inbound customer interactions through natural-language channels, pulling data from the underlying SAP CIS and FI-CA contract accounts.
Model breadth also expanded in May 2026 when SAP and Anthropic announced that Claude is available through the SAP Business AI platform via the Model Context Protocol, giving customers a choice of foundation model alongside SAP’s own and other provider options.
At Sapphire, SAP named RWE as a utility customer in the context of its autonomous enterprise showcase, signaling that at least one large investor-owned utility is engaged with the agentic automation roadmap. What RWE has in full production versus pilot is worth clarifying in any vendor conversation.
SAP’s AI is strongest when the utility already runs a deep SAP estate (S/4HANA, FI-CA, SAP MDM). Joule agents read and act within that estate; they are not designed to orchestrate outside it.
Oracle: Opower household scale and targeted CIS AI
Oracle’s AI approach is more surgical. The largest deployed footprint is in customer engagement through Opower, which Oracle reported at approximately 44.6 million households enrolled as of April 2026. That figure, which Oracle reported and should be validated against current disclosure, represents the enrolled base for personalized energy insights, behavioral efficiency nudges, and demand-response communications. It is the most scaled AI-driven customer engagement asset any of the three vendors can point to.
On the CIS operations side, Oracle shipped AI-powered call summarization for utility customer service in May 2025, reducing after-call work for agents. Separately, Oracle introduced AI anomaly detection in its Meter Data Management product in June 2025, targeting loss detection and faulty-meter identification in AMI interval data.
These are bounded, production-shipped capabilities. Oracle is not yet leading on the broad agentic automation narrative that SAP is pushing, but it is shipping specific AI features into parts of the stack that matter to utility operations.
Cayenta: Cayla agent in production at mid-market utilities
Cayenta’s AI layer is called Cayla, which covers both a voice and digital customer-facing agent and an in-app assistant for staff. The scope is narrower than Oracle’s or SAP’s, but Cayenta has a documented production deployment. The City of Lakeland, Florida went live with the Cayla agent in approximately May 2026, implemented with partner Dial AI. A municipal utility running Cayla in production is meaningful evidence of real-world readiness at the mid-market scale Cayenta targets.
Comparison by capability and utility size
| Capability | SAP (Joule / Industry AI) | Oracle (Opower + CC&B AI) | Cayenta (Cayla) |
|---|---|---|---|
| Customer self-service agent | GA, Q4 2025 (utility-specific) | Opower engagement scale (44.6M households, Oracle-reported) | GA, production at Lakeland FL ~May 2026 |
| Call / interaction summarization | Joule CX layer | Shipped May 2025 | In-app assistant |
| Meter / MDM anomaly detection | SAP MDUS + Joule | Shipped June 2025 (Oracle MDM) | SmartWorks MDM (AI features not separately dated) |
| Agent / agentic automation breadth | 200+ agents, 50+ assistants (Sapphire May 2026) | Targeted; not a broad agent platform | Narrow; Cayla scope |
| Best fit by utility size | Large IOU on SAP estate | Large IOU on Oracle estate, Opower cross-stack | Municipal and mid-market |
| Model choice | SAP AI + Claude (Anthropic, May 2026) | Oracle LLMs | Not publicly disclosed |
An honest verdict by utility size
Large investor-owned utilities on SAP get the broadest AI surface area in 2026, but the value is proportional to how deeply the SAP estate is already deployed. Joule does not add much value to a utility that runs IS-U standalone without S/4HANA Finance, BTP, and the surrounding stack.
Large investor-owned utilities on Oracle get a mature, scaled customer engagement platform in Opower and targeted AI in billing and MDM. The agentic automation story is less developed than SAP’s, but the shipped capabilities are real and production-proven.
Municipal and mid-market utilities evaluating Cayenta get a narrower AI layer that is already in production at a comparable peer, without the implementation complexity of the enterprise platforms. For many municipals, Cayla doing a focused job reliably is worth more than access to 200 agents that require significant configuration and integration.
The honest framing is this: AI is not yet a reason to change CIS platforms. It is a reason to ask harder questions about your current vendor’s roadmap, the integration architecture AI will require, and how human-in-the-loop controls will work for regulated billing decisions. Those questions connect to the broader platform evaluation covered in our guide to choosing the right utilities software.