Smart grid, legacy back office: The technical debt utilities are carrying into their next transformation phase

Article

J.D. Power’s 2025 United States Electric Utility Residential Customer Satisfaction Study recorded the lowest satisfaction score ever measured — 499 out of 1,000 — with billing reliability and outage communication identified as the primary levers available to recover it. The same dynamic is playing out across markets. In the United Kingdom, energy providers paid out approximately £20 million in compensation for billing mistakes over five years, with complaints up 141% over that period and billing disputes accounting for 58% of all Energy Ombudsman cases in 2024.

Those aren’t customer experience problems in isolation. They’re symptoms of whether back-office processes run reliably, in sequence and at scale. For most large utilities, the orchestration layer responsible for that is still catching up to the grid it’s supposed to support.

Utilities have spent the better part of a decade winning the case for grid modernization, deploying Advanced Metering Infrastructure (AMI), integrating renewable energy and managing Distributed Energy Resources (DER). Grid edge intelligence is generating more operational data than most organizations know what to do with. What hasn’t moved at the same pace is the software layer connecting all of that to the business processes that serve customers and satisfy regulators.

In most large utility environments, these application and data pipeline workflows are still being automated by legacy schedulers and scripts designed before hybrid cloud existed, before AMI was standard and before a single weather event could simultaneously spike customer calls, outage jobs and data ingestion volumes across a dozen interconnected systems.

The seam running through every utility’s automation environment

Utilities deal with a version of the problem with an extra dimension: the systems that run the grid and the systems that run the business weren’t built to work together — and for most of their history, they didn’t need to.

Operational technology (OT) and information technology (IT) evolved in parallel in most large utilities, governed by different teams and built to different standards. AMI put metering data in both worlds simultaneously. DER integration put real-time grid signals into billing and settlement workflows. Outage management systems started feeding customer communication platforms. What most utilities built to bridge that divide was a collection of schedulers, scripts and point-to-point integrations, each solving a specific problem without anyone owning the full picture.

As a result, a meter-to-cash workflow can touch a dozen systems with their own scheduling layers and their own definitions of “done.” When one step slips, the cascade is difficult to see and harder to interrupt because no single platform spans the full picture. For years, keeping those worlds loosely coupled was an acceptable tradeoff. It isn’t anymore — not when time-of-use tariffs, demand response activations, electric vehicle charging programs and AI-driven forecasting all require them to move in sync.

Stress reveals weakness

In the utility sector, spikes aren’t rare, and they’re considerably less forgiving of fragmented automation than normal conditions. During a major weather event, for instance, utility operations teams are running outage management workflows, restoration sequencing and customer communication jobs simultaneously, while billing cycles, AMI data ingestion and regulatory reporting continue in the background. Each process depends on others completing in the right order, at the right time, against dependencies that no single monitoring console can see across.

The response ends up being human coordination: multiple dashboards, manual judgment calls, compliance trails that go dark and a customer experience that degrades in ways that are difficult to trace after the fact. Regulators and customers don’t distinguish between a transmission failure and a process failure. When automation silos prevent outage workflows and billing systems from executing in sync, the damage is the same.

The operational case for addressing this is clear. The organizational dynamics around doing so are considerably less straightforward. Legacy scheduler renewal contracts route through procurement, extensions get signed and the harder conversation about whether the solution still fits the strategy gets deferred. Another cycle of expensive upgrades with painful agent patching gets funded, and the war-room model gets staffed again during the next storm season.

AI ambition and the foundation it needs

The IFS Global Utility Survey 2024 found that 82% of utility executives consider AI essential to their digital transformation strategy, yet only 20% have completed that journey. That gap isn’t primarily a technology problem, as every utility surveyed had initiated transformation. The ones stalled are the ones that haven’t addressed the execution layer underneath their AI ambitions.

Predictive outage management, demand forecasting, dynamic pricing — none of these perform against fragmented, inconsistently sequenced data workflows, and no amount of hiring fixes an orchestration layer that can’t support the models sitting on top of it. Closing that gap requires consolidating what sits underneath the models, and the utilities moving fastest on AI are the ones that addressed the orchestration layer first.

Consolidating onto a strategic orchestration platform

That’s the shift utility companies are making with RunMyJobs by Redwood. They’re replacing legacy tools, open-source schedulers and bridging scripts with a single orchestration and execution control plane that governs end-to-end workflows across the full span of utility operations.

Legacy, self-hosted workload automation (WLA) schedulers require agents installed across every server and environment, each tied to OS updates, security patches and version dependencies. In a hybrid OT/IT utility environment, that maintenance load is constant, consuming engineering time, increasing technical debt and operational costs that boards are now expecting to fund AI initiatives and digital transformation instead.

RunMyJobs SaaS is cloud-native and agentless. Updates arrive as part of the SaaS service, so you maintain your cybersecurity posture without dedicated patching cycles. The engineering capacity previously absorbed by platform maintenance shifts toward grid digitalization and the AI programs already on the roadmap.

For utility operations teams, that translates directly:

• Billing cycles, regulatory submissions and outage workflows no longer depend on manual monitoring to catch failures — built-in dependency management and real-time triggering respond automatically when a step doesn’t complete on time
• Audit-readiness is built into execution by default, with every step logged and traceable, so compliance reporting doesn’t require post-hoc reconstruction
• Grid resilience improves measurably, as 99.95% uptime means mission-critical outage and restoration workflows execute under peak load without adding infrastructure or having lots of people on call
• SAP and non-SAP systems orchestrate from a single platform, eliminating the parallel scheduling layers and maintenance-heavy custom scripts most utilities are currently running

Consolidation also changes the economics. Restrictive, self-hosted legacy licensing, maintenance fees and unpredictable cost increases are replaced by transparent, predictable SaaS pricing — and total cost of ownership (TCO) drops meaningfully when agent patching, infrastructure management and upgrade projects that delivered no new capability are no longer on the bill. That spend gets redirected toward AI initiatives and digital transformation programs that boards are already asking about. For IT leaders, that’s legacy costs converting into investment fuel. For business leaders, it’s the operational headroom to bring new tariffs, demand response programs and customer-facing digital services to market faster.

Renewal as the decision point

Most large utilities are closer to this decision than they realize. Legacy WLA contracts come up for renewal and get routed through procurement and extended for another cycle without the strategic question ever being asked.

If a renewal is on the horizon for your organization, the question worth asking isn’t whether your current environment is stable enough to extend, but whether “stable enough” is still sufficient for what the business is being asked to deliver.

Explore Redwood Software’s proven approach to migration and talk to an expert about what consolidation could look like in your environment.

Your AI tools now speak RunMyJobs: MCP brings agentic AI to your enterprise applications

Article

According to a recent Redwood Software customer survey, 68% of RunMyJobs by Redwood users work with AI tools multiple times per week. They ask ChatGPT to troubleshoot errors, use Copilot to draft scripts and paste job logs into Claude and ask what went wrong.

None of those AI models can reach into RunMyJobs and take action. They answer questions about your workflows but can’t run them, check on them or build new ones. Your AI assistants and your automation platform operate in separate worlds, and that gap is exactly what makes it hard to get real value out of either.

Model Context Protocol (MCP) support in RunMyJobs changes that.

Why every major AI platform adopted the same protocol

MCP is an open standard that gives AI systems a shared way to connect with external tools. Think of it as the USB port of agentic AI. MCP lets you plug any MCP-compatible AI agent into any MCP server without custom integrations. Drop an MCP server in front of a product, and AI systems can immediately interact with it, reading context, calling functions and taking action.

Anthropic released MCP in late 2024 and donated it to the Agentic AI Foundation under the Linux Foundation in December 2025. Since then, OpenAI, Google DeepMind, Microsoft, Salesforce and ServiceNow have adopted it. The protocol has moved past experimentation, as its interoperability across AI platforms is proven in production today.

For RunMyJobs users, MCP means something specific: the business logic, connectors and workflows you’ve spent years building are now accessible to AI agents through a standardized protocol that every major AI platform already supports. Any MCP-compatible AI tool or large language model can now trigger your workflows, check job status and build new workflows through the RunMyJobs MCP server with no custom API work and no rearchitecting. The workflows and connectors you’ve built over years of production use become tools that AI agents call on demand.

This is how your backend processes become agentic.

What can AI tools do through RunMyJobs’ MCP?

• Trigger workflows and jobs: Any MCP-compatible agent can kick off your existing RunMyJobs workflows to make your current processes agent-ready without migration.
• Check job status: AI tools can query whether critical jobs are running, finished or failed and surface that information inside whatever platform your team uses.
• Manage workflows: Coding agents can validate and deploy RunMyJobs workflows through MCP, cutting development time.

These capabilities work with Claude, Microsoft Copilot Studio, ServiceNow Agent Builder, n8n, ChatGPT and Salesforce Agentforce. Authentication, access controls and permissions all flow through RunMyJobs — agents can only do what their associated role is allowed to do.

Here’s what that looks like in three real-world use cases.

SAP’s Joule: Submitting and monitoring jobs from your ERP

Your SAP basis administrator needs to trigger a nightly data extraction early because a report deadline moved up. Instead of switching to the RunMyJobs console, logging in, finding the workflow and submitting it by hand, they stay in SAP and tell Joule: 

“Submit the GL account extraction workflow for company code 1000.”

Joule calls RunMyJobs through MCP. The workflow starts. Joule confirms it.

An hour later, the same admin asks Joule about the financial data load. Joule checks RunMyJobs and finds that the extraction finished, but the transformation step is still running. Estimated completion: 45 minutes.

No dedicated SAP integration project made this possible. MCP standardized the connection. RunMyJobs partitions and roles still control who can trigger what, so your governance model is intact, but your admin gets a faster, context-aware path to the same workflow they’ve run hundreds of times before.

This is what it means to agentify your existing SAP processes: The workflows don’t move, and the business logic stays where it is. Joule just gets a direct line to it.

ServiceNow: Remediating failed batch jobs without the 2 AM phone call

Your nightly accounts receivable batch job fails at 2:14 AM. Today, that triggers a ServiceNow incident. An on-call operator picks it up, logs into RunMyJobs, reads the error log, figures out the cause, restarts the job with corrected parameters and closes the ticket. That process takes 30 to 90 minutes, depending on who’s on call and how fast they diagnose the issue.

With ServiceNow Agent Builder and MCP, the ServiceNow agent handles most of that loop. It detects the failed job alert, queries RunMyJobs through MCP for real-time error details and job history and matches the failure pattern against known remediation steps. If the fix is a known restart with corrected parameters — wrong file path, stale credentials, a transient connection timeout — the agent resubmits the job in RunMyJobs and updates the ServiceNow incident with what it found and what it did.

If the failure falls outside the agent’s confidence threshold, it escalates to the on-call operator with a pre-built diagnostic summary: the error, job chain context and last three successful runs for comparison. Your operator starts the investigation 15 minutes ahead of where they’d be without it.

RunMyJobs still controls execution. Partitions and roles still govern who — or what — can restart which jobs. ServiceNow still owns the incident lifecycle. MCP connects the two external systems without custom middleware in between.

Microsoft Copilot Studio: Finance teams running month-end close

Month-end close involves dozens of batch processes across ERP, consolidation and reporting systems. They run in strict sequence, often at night, and someone watches a console to catch failures.

Your finance controller builds a Copilot agent in Microsoft Copilot Studio. The agent submits the intercompany elimination workflow through MCP to RunMyJobs. When that job finishes, the agent triggers the consolidation jobs. If reconciliation fails, the Copilot agent sends the controller a Microsoft Teams message — a plain-language summary of the failure, plus the remediation workflow they can approve with one click.

The controller doesn’t need RunMyJobs training. They tell the Copilot agent what outcome they need, and RunMyJobs handles execution. Your finance team stays in Teams and focuses on the close, not the tooling.

What this means for your RunMyJobs investment

The good news for teams feeling pressure to adopt agentic AI: you don’t have to rewrite your enterprise workflows. You don’t have to move your batch processing into a new tool. MCP exposes what you’ve already built to the agents your developers want to build, through a standardized protocol they already know.

The automation fabric you’ve built in RunMyJobs is your real AI asset; MCP is how you unlock it.

The RunMyJobs governance model — partitions, roles, access controls — still applies. Scalable agentic orchestration doesn’t require trading away the enterprise-grade controls you rely on. Your AI-powered workflows run under the same oversight as everything else in your automation environment. But your teams get a new way to interact with automation that fits inside the AI tools they’ve already adopted.

Redwood is building toward a model where AI agents and workload automation run side by side, supporting open standards such as agent-to-agent (A2A) and MCP to unlock existing business logic and make it accessible in a governed and observable platform operating at enterprise scale. MCP support in RunMyJobs is where that starts, and the foundation is the automation you’ve already built.

See how RunMyJobs works with MCP to unlock your investment in enterprise applications and expose them to agents. Get a demo today.