The meter-to-cash process is one of the most exception-heavy workflows in enterprise software. Tariff configurations that span sprawling billing schemas, meter reads that arrive implausible and need specialist investigation, energy products that vary by household type, installation, and regulatory jurisdiction. Utilities have lived with this complexity for decades, and the tools they use to manage it have demanded a corresponding level of manual effort.

During a recent webcast, Holger Schweinfurth, Senior Director Utilities Solution Roadmap and Transformation Area Energy at SAP, outlined an AI agent strategy built specifically around the meter-to-cash process. The agents target the specialist roles that live inside this workflow: tariff experts configuring energy products, meter data specialists managing reads and resolving quality issues, and billing specialists investigating exceptions. 

Schweinfurth framed the underlying challenge as business process exception management (BPEM). Across meter-to-cash, problems surface constantly, and resolving them is manual, time-consuming work that pulls specialists deep into system data. The agents are meant to take on that load.

The billing document agent, planned for Q3 2026, is the furthest along of four agents SAP has mapped across the meter-to-cash process. The others include an Implausible Meter Reading Resolution Agent, an Outsorted Utilities Invoicing Document Resolution Agent, and the Utilities Tariff Agent.

From Machine Learning to Agents

The billing document agent isn’t starting from scratch. Its foundation is the machine learning scenario SAP introduced with S/4HANA 1909, which enabled automated classification of implausible meter reads. SAP is now extending that into an agent that can analyze a flagged billing document, surface key findings, compare it against previous billing periods, and pull in related meter reading data.

In the demo, shown via video on a standard Fiori worklist interface, the interaction was conversational. A specialist selects a flagged document, opens the agent, and the system automatically populates an initial analysis with findings and recommendations. From there, the specialist can ask the agent to compare the current document to a prior period—the demo surfaced an increase in billing amount and offered possible explanations—or retrieve meter reading history. 

The agent draws from multiple data sources to assemble its responses, working within the same interface where specialists already review implausible documents and ML confidence scores. SAP is projecting roughly 50% higher dispatcher productivity and 50% fewer incorrect field service resource allocations.

Comparison, retrieval, and summarization are table stakes for AI tooling in 2026. The more interesting move is that SAP is anchoring the work in an existing ML scenario that utilities already use, extending it incrementally rather than asking customers to adopt an entirely new paradigm.

Tariff Work: Two Steps, Different Timelines

Schweinfurth described a two-phase approach to tariff intelligence under the umbrella of a Utilities Tariff Agent, currently in an ongoing proof of concept. The first phase, a Tariff Analyzer, works through existing billing schemas to surface how they’re configured and flag where they may need attention across the mass of tariffs in a given system. 

SAP is projecting an 80% reduction in system training effort for tariff experts and billing specialists, and a 50% time reduction for tariff creation and adjustment. Early customer results have been promising enough that SAP has invited additional utilities to participate.

The second phase is further out. A Tariff Designer would allow users to enter parameters—household type, wall box installation, other customer-specific attributes—and have the system generate a first-pass energy product and tariff configuration. 

He clarified that this is still in specification, not approaching delivery. But the ambition is notable: tariff design is one of the most labor-intensive and expertise-dependent tasks in utilities billing, and meaningfully automating even part of it would represent a significant shift in how energy products get to market.

SAP Business Data Cloud Finds Its Utilities Use Case

Business Data Cloud (BDC) got less airtime than the AI agents during the session, but it may matter just as much. Schweinfurth pointed to correlating asset data with customer data: connecting what’s happening at a device or meter level to the account and billing structures downstream. Utilities have spent years building workarounds to answer questions that span those boundaries, which helps explain why SAP is investing here even though the utilities-specific buildout is still early.

SAP has been shipping utilities-specific BDC data products since Q1 2026, starting with time series and energy settlement master data, and the scope widens considerably through the rest of the year. By Q4, the coverage extends to meter reading and billing—which matters because those are the same domains the AI agents need to draw from. 

Tariff, pricing, and BPEM data products are still further out on the roadmap. Schweinfurth explained the connection between the capabilities: BDC is the data integration layer he sees underpinning the agent strategy at scale.

The Bigger Picture

For utilities organizations running S/4HANA or planning their migration path, the session offered a useful read on where SAP’s industry investment is headed. The billing document agent is the nearest concrete deliverable. The tariff work is in active proof of concept with a longer horizon. And SAP BDC is building the data integration layer that the broader agent strategy will eventually depend on.

SAP wants to make the meter-to-cash process less dependent on specialists manually navigating system complexity, and it’s building the AI tooling and data infrastructure to support that shift.