The utilities industry has long operated on a simple mandate: generate power, move it through the grid, and bill the customer. But SAP is betting that AI will push utilities beyond this transactional identity toward something more ambitious. Bruno Pincolini, SAP’s Global Lead of Business AI for Utilities, describes the shift as transforming utilities “from providers of essential services to platforms for innovation and resilience.”
Joule agents are central to that transformation. SAP is currently positioning Joule not as a standalone chatbot but as a “process-aware copilot,” per Pincolini, that spans the full utilities value chain, from smart metering and market operations through bill-to-cash workflows, customer experience, and service and asset management.
SAP is now putting specific product commitments behind that vision. As part of its 2026 roadmap, the company plans to roll out AI-driven capabilities targeting customer self-service, consumption analysis, and billing exception handling. In an ASUG interview, Pincolini previewed the new and forthcoming features and governance architecture geared toward making AI viable in a sector defined by compliance constraints.
SAP's approach is as deliberate in what it prioritizes as what it avoids. “The aim is not to replace proven transactional controls in SAP for Utilities,” Pincolini explains, “but to speed insight, decision-making, and guided execution on top of them, without compromising governance.”
Process-Native Versus Generic AI
General-purpose AI copilots struggle in enterprise contexts because they lack access to the data structures and business logic that define how a company actually operates. When a customer calls about an unexplained spike in their bill, a generic chatbot can deliver vague answers or search public knowledge bases, but it cannot pull the customer’s actual meter reads, compare them to their rate schedule, or check whether a recent service order might explain the discrepancy.
Pincolini noted that general-purpose AI copilots “risk hallucinating or giving imprecise answers because they lack real data, context, or integration.” SAP’s answer is what it calls “grounding.” Joule connects directly to the SAP Knowledge Graph and SAP Business Data Cloud, anchoring its responses in a customer’s actual enterprise data rather than statistical inference about what a plausible answer might be.
The depth of that integration appears in Joule’s industry-specific vocabulary. The agent is programmed to understand commercial structures like business partners and contract accounts alongside technical assets like devices, registers, and meter reads. Joule can also navigate process statuses such as dunning levels and service orders. When a billing inquiry arrives, the system pulls invoices, payment history, and service records to construct a response grounded in fact rather than inference.
Joule can also coordinate complex tasks that span multiple SAP applications—S/4HANA Utilities, Service Cloud, Field Service Management, and Enterprise Asset Management—drawing on approved APIs and SAP Integration Suite to execute workflows end-to-end. Pincolini cites this orchestration capability as the definition of “process-native.” Beyond the vocabulary of utilities operations, the AI understands the actual mechanics of how work gets done in SAP systems.
Inside the Contact Center
The most concrete implementation of Joule today lives in SAP Service Cloud 2.0, where several AI features target the realities of utility contact center operations. A Case Classification Agent analyzes incoming service tickets and categorizes them according to company-specific business requirements. A Case Summary function generates condensed histories from the full thread of customer communications, sparing agents from scrolling through dozens of exchanges to understand the situation.
For email-originated cases, a Case Type Determination model assesses the subject line and description to establish the appropriate case type, then inputs that data to automatically initiate case creation.
The Similar Case Recommendation feature scans the previous twelve months of categorized cases, using semantic sentence-level analysis rather than simple keyword matching to surface the three most relevant precedents. Semantic analysis captures meaning and context that keyword matching misses, which improves relevance when agents need to see how the organization handled comparable situations.
New and Upcoming Features
SAP’s current roadmap for Joule in utilities extends the assistant’s reach into customer self-service and consumption analysis, while finally targeting one of billing’s persistent pain points: exception handling.
Since Q4 2025, a Utilities Customer Self-Service Agent has been equipped to handle customer interactions with visibility into the customer’s full profile: rate structures and active products, historical usage, and potential upgrade paths. SAP’s stated goal is performance “equal to handling by human agents,” which, if achieved, can cut operating costs while potentially improving customer loyalty and identifying revenue opportunities that purely transactional self-service misses.
Also live since Q4 2025, an AI Summary for Billed Consumption capability compiles a rolling year of billing history for each metered service, surfacing usage trends and flagging what drove any fluctuations from one cycle to the next. This gives utility administrators the context they need for informed decision-making and proactive service management.
In the second quarter of 2026, SAP plans to introduce a Joule Agent for resolving outsorted billing documents. These exceptions are a manual bottleneck in an otherwise automated billing system and require human judgment. The agent will review the full context behind each exception—the Business Process Exception Management (BPEM) case file, the customer’s master data and consumption history, and the billing document itself—then propose a resolution with supporting rationale. The aim is to support billing specialists by providing analysis and recommendations, while leaving final decisions to humans.
The Governance Question
Utilities operate under regulatory constraints that make governance non-negotiable. Billing regulations, shutoff protections, and complaint-handling requirements create a compliance landscape where autonomous AI recommendations could easily cause problems.
SAP addresses this through what Pincolini describes as multiple layers of guardrails:
- Technical security controls sit alongside role-based access that ensures agents only see and do what they’re authorized to see and do.
- Agent orchestration runs through a controlled hub.
- Audit logging enables accountability.
- Regulatory-change management capabilities help organizations adapt as rules evolve.
The architecture is designed for regulated, risk-sensitive domains where an AI’s recommendations must remain within the bounds of what is legally and procedurally permissible.
Calculating The Cost of AI
SAP’s AI commercial model is divided into two tiers. Many embedded AI capabilities come included with cloud application subscriptions and inherit the data residency, role-based access, and audit controls of the host application. Basic AI features and a starter allocation of Joule messages typically require no additional licensing.
Premium capabilities follow a consumption-based model metered through SAP Business Technology Platform (BTP). This tier includes platform services like the Generative AI Hub and AI Core, specialized tools like Document Information Extraction and Joule Studio for building custom agents, and usage-based charges for high-volume Joule interactions and agent orchestration. All require AI Units, billed against SAP BTP entitlements under CPEA or pay-as-you-go terms.
Joule operates across both models. The in-application experience is included with eligible SAP cloud subscriptions. However, when Joule uses premium SAP BTP AI services or third-party large language models via the Generative AI Hub, that usage is billed against SAP BTP consumption.
On the privacy front, SAP commits to keeping customer data out of foundation model training unless the SAP user explicitly authorizes it. For an industry where regulatory trust is foundational, that commitment to data boundaries is reinforcing the guardrails approach Pincolini emphasizes overall for the sector.
Stay ahead with First Five, ASUG's exclusive newsletter. Subscribe today for a curated email featuring top insights tailored for SAP users.