A cross-functional AI initiative that survives one executive sponsor does not survive a reorg. The utilities that build durable AI capability do not depend on a single champion at the top; they build a coalition of champions at the working level, across planning, operations, and compliance, each of whom owns the AI program's credibility in their own function and can defend it when the organizational chart changes.

Why Single-Champion Programs Fail

The most common failure mode for utility AI programs is not technical. It is not a model that underperforms or a vendor that over-promised. The most common failure mode is an organizational one: a program that was built around a single executive champion who leaves, retires, moves to a different role, or loses organizational authority in a restructuring. When that champion exits, the program loses its funding rationale, its access to operations, and its ability to navigate the compliance review that keeps AI tools inside NERC accountability boundaries. Programs that look like utility-wide transformations turn into abandoned pilots within 18 months of a leadership change.

The pattern is predictable because the programs are built on a predictable organizational foundation: a champion at the C-suite or VP level who creates a budget line for AI, sponsors vendor relationships, and directs functional teams to cooperate with deployment efforts. That structure works while the champion holds authority. It does not survive a reorg, a budget cycle, or a succession because the program's credibility in each function was borrowed from the champion's organizational authority rather than earned from functional value delivered to planning, operations, and compliance.

The durable alternative is a champion coalition built from the working level up. A planning manager who has used AI-assisted forecasting to defend a load growth assumption in an IRP filing owns the AI program's credibility in planning. A senior dispatcher who has used an AI topology optimization advisory tool and can describe both its utility and its limitations to peers owns the AI program's credibility in operations. A compliance lead who has worked through the NERC accountability documentation for AI-assisted reliability decisions owns the AI program's credibility in compliance. When these three professionals are in the coalition and each owns credibility in their own function, the program survives a reorg because its foundation is functional value, not hierarchical authority.

Identifying and Cultivating Champions by Function

The champion identification process begins with a diagnostic question for each function: who in this function has the operational credibility and technical curiosity to become the person their peers ask about AI advisory tools? The answer is almost never the person who is most enthusiastic about AI technology. It is the person who is most respected for their operational judgment, who is professionally curious about new methods, and who is not so committed to existing approaches that they cannot evaluate a new tool honestly.

Enthusiasm about AI is a disqualifier for the champion role because it produces exactly the wrong kind of advocacy. An enthusiastic champion says the tool always works and peers stop believing them after the first visible failure. A champion who says the tool works well in these specific scenarios and requires additional scrutiny in these others, and who has the operational credibility to back that assessment, maintains peer trust indefinitely. The champion's role is not to sell the tool; it is to translate AI capability into operational value in terms their function understands and cares about.

In the planning function, the champion role is best suited to a load forecasting lead or transmission planner who works directly with the planning data that AI tools ingest. This professional understands the quality and limitations of the data, can evaluate the AI tool's outputs against the underlying assumptions, and can speak to planning peers and regulatory audiences about the AI-assisted analysis in terms that are defensible in a rate case. The planning champion's specific value is regulatory defensibility: making AI-assisted planning credible to commissions and intervenors who will challenge the methodology.

In the operations function, the champion role is best suited to a senior dispatcher or substation operator with substantial territory experience and strong peer relationships. This professional has direct operational experience with the advisory tool in actual operational conditions, can describe specific scenarios where the tool's recommendations were valuable and specific scenarios where they exercised an override, and is trusted by colleagues to give an honest assessment of both. The operations champion's specific value is trust transfer: peers who would not trust the tool's output based on documentation will trust the assessment of a colleague they respect who has used it in actual operations.

In the compliance function, the champion role is best suited to a NERC compliance professional who works on reliability standards and is familiar with the accountability documentation requirements for reliability-relevant decisions. This professional understands the regulatory framework within which AI advisory tools must operate, can identify the documentation requirements that ensure AI-assisted decisions are defensible in a NERC audit, and can work with the program team to ensure that the governance structure maintains NERC-required human accountability. The compliance champion's specific value is audit durability: ensuring that AI-assisted decisions are documented in a way that survives NERC scrutiny regardless of who is in the room when the auditor asks questions.

The Cross-Functional Coalition Structure

A champion coalition is not a committee. Committees diffuse accountability; coalitions concentrate it. Each champion in the coalition owns specific deliverables for their function and specific communication responsibilities to their function's leadership and peers. The coalition's governance structure should be minimal enough to convene quickly and substantial enough to make decisions that carry organizational weight.

The coalition structure that works in a utility environment has four elements. The first is a defined scope: each champion's role is scoped to specific AI tools and specific use cases within their function. A planning champion who is responsible for AI-assisted forecasting tools has a defined scope; if the coalition's mandate expands to topology optimization tools, a new operations champion is added rather than overloading the existing champion with an out-of-function scope. Scope creep in the coalition is the primary mechanism by which champions become overwhelmed and their advocacy becomes superficial.

The second element is a shared knowledge base. The coalition maintains a shared repository of operational evidence: shadow-mode assessment results, override analysis, post-event reviews, and regulatory interactions where AI-assisted analysis was used. This repository is the factual foundation for each champion's advocacy in their own function. When the planning champion defends AI-assisted forecasting methodology in a rate case, they draw on documented evidence from the repository rather than vendor claims. When the compliance champion prepares NERC audit documentation, they draw on the same repository to demonstrate that accountability structures are in place and functioning.

The third element is a clear communication structure. Each champion has defined communication responsibilities: a regular briefing to their function's leadership on AI tool performance and governance; a peer communication channel through which operational staff can report concerns, ask questions, and share observations; and a coalition-level communication in which champions share what they are hearing from their functions. The coalition's ability to surface operational concerns before they become reliability problems depends on these communication channels being active and trusted by operational staff.

The fourth element is a succession plan. A coalition that has not identified successors for each champion position is one departure away from losing the functional credibility that took months to build. Champion succession should be part of the applied AI project work from the upskilling program: a champion identifies and mentors a potential successor from the training program's advanced cohort, ensuring that the champion's operational knowledge of the tool and its limitations is transferred rather than lost when the champion transitions to a new role.

The Reorg Survival Test

A useful diagnostic for a champion coalition is the reorg survival test: if the executive sponsor of the AI program left tomorrow, would each champion be able to independently defend the program's value to the new organizational structure? The answer requires that each champion has three things: documented evidence of functional value delivered by AI tools in their specific function; a peer network within their function that trusts their assessment of the tool; and a clear understanding of the program's governance structure, specifically the accountability boundaries and documentation requirements that make the program defensible to NERC and to leadership. A coalition in which every champion can pass the reorg survival test is a coalition that does not depend on any single individual for its organizational legitimacy.

Building the Compliance Champion's Accountability Documentation

The compliance champion's most operationally important deliverable is a documentation framework that makes AI-assisted reliability decisions defensible in a NERC audit without requiring the executive sponsor to be in the room. This framework has three components.

The first component is the decision accountability record. Every reliability-relevant decision made with AI advisory tool support generates a record that includes: the specific NERC standard or requirement governing the decision; the operator's or engineer's name and qualification status; the AI advisory available at the time of the decision; the operator's decision and its stated basis; and the timestamp sequence showing that the advisory was reviewed before the decision was made. This record is the documentary foundation for demonstrating that the AI tool was an information source in a human decision-making process, not the decision-maker.

The second component is the governance documentation. The advisory-never-autonomous policy is published and signed. The scope of each AI advisory tool is defined in writing. The formal process required to change that scope is documented. The competency requirements for operators and engineers who use each AI tool are documented. The training completion and competency assessment records for each operator and engineer are maintained. This governance documentation demonstrates to a NERC auditor that the organizational structure explicitly preserves human accountability rather than relying on informal verbal commitments.

The third component is the performance monitoring record. Shadow-mode assessment results, override log summaries (categorized by scenario type, not by individual operator), and post-event review findings for AI-assisted decisions are maintained as a continuous operational record. This record demonstrates that the utility is monitoring AI tool performance, surfacing failures, and improving the knowledge grounding and configuration of AI tools based on operational evidence. A NERC auditor who sees a performance monitoring record is seeing evidence of a reliability management process, not an unmonitored technology deployment.

CIP-003-9, enforceable from April 1, 2026, specifically addresses vendor electronic remote access and supply-chain security for low-impact BES Cyber Systems. CIP-012-2 (effective July 1, 2026, adding availability to the confidentiality and integrity protections of the prior CIP-012-1) governs the real-time data communications that AI advisory tools use as their primary input layer between control centers. Both have direct implications for AI advisory tool deployments. The compliance champion who understands both standards can ensure that the AI tool deployment's data architecture, access controls, and vendor connectivity satisfy these requirements before deployment rather than discovering compliance gaps after deployment when remediation is expensive.

Worked Example: The Coalition That Survived Two Reorgs

A large vertically integrated IOU deployed AI advisory tools across three functions over a three-year period: an AI load forecasting tool in the planning department, a topology optimization advisory tool in the control room, and an AI relay setting drafting tool in the protection engineering department. The initial deployment was sponsored by a VP of Reliability who was deeply committed to the program and drove it through two budget cycles and one vendor transition.

Eighteen months into the deployment, the VP of Reliability retired. The incoming executive did not have the same background in AI tool deployment and was not a natural advocate for the program. Six months after the retirement, a reorganization merged the planning and operations departments under a new EVP who had no history with the program. Three months after the merger, the program's budget was cut by 40 percent in the first review cycle.

The program survived both transitions and the budget cut because of a coalition that had been quietly built during the VP of Reliability's tenure, not by the VP directly but by a program manager who had seen single-champion programs fail at a previous employer and was determined not to repeat the pattern. The coalition had three members: a load forecasting manager who had successfully used AI-assisted forecasting to defend a 15 percent upward revision to the utility's ten-year load growth forecast in a contested IRP proceeding, with documentation that convinced the commission's technical staff despite challenges from two intervenors; a senior dispatcher who had been involved in the shadow-mode assessment, had advocated for the topology optimization tool among control room peers, and had documented 14 cases in the past 12 months where the advisory had flagged a potential constraint violation that the dispatcher had verified and confirmed; and a NERC reliability compliance engineer who had developed the decision accountability documentation framework for AI-assisted reliability decisions and had used it in a practice NERC audit preparation exercise six months before the VP of Reliability retired.

When the incoming executive asked the new EVP to explain why the AI program should be funded in the reorganized department, the EVP had three documents ready: the IRP proceeding outcome with the load forecasting manager's documented methodology; the dispatcher's 12-month operational log; and the compliance engineer's audit-ready documentation package. The budget cut was reversed. The program continued. Neither the VP of Reliability's departure nor the reorganization ended the program because the program's credibility was distributed across three working-level professionals who each owned it in their own function.

The program manager who built the coalition identified two factors as most important in retrospect. First, the champions were selected for their operational credibility rather than their enthusiasm for the technology. None of the three was a technology advocate; all three were operations and regulatory professionals who had found specific, documented value in the tools they advocated for. Second, the shared knowledge base made each champion's advocacy factually grounded. The load forecasting manager did not advocate for AI forecasting in the abstract; she advocated for this forecast, in this IRP proceeding, with this methodology documentation. The dispatcher did not advocate for topology optimization in the abstract; he advocated for this tool, on this territory, with this 12-month operational record.

Key Takeaways

• Single-champion AI programs do not survive leadership changes or reorganizations; the most common failure mode for utility AI programs is organizational, not technical.
• A champion coalition built from the working level up is durable because its credibility is earned through functional value delivered to planning, operations, and compliance, not borrowed from hierarchical authority.
• Champions are selected for operational credibility and peer trust, not for AI enthusiasm; an enthusiastic champion who claims the tool always works loses credibility the first time it does not.
• The planning champion's specific value is regulatory defensibility; the operations champion's specific value is trust transfer among peers; the compliance champion's specific value is audit durability under NERC scrutiny.
• CIP-003-9 (vendor electronic remote access and supply-chain security for low-impact BES Cyber Systems, enforceable April 1, 2026) and CIP-012-2 (confidentiality, integrity, and availability of inter-control-center real-time data, effective July 1, 2026) have direct implications for AI advisory tool deployments; the compliance champion who understands both can ensure data architecture and vendor access are compliant before deployment.
• A coalition that has not identified successors for each champion position is one departure away from losing the functional credibility that took months to build; champion succession is an operational necessity, not an administrative formality.
• The reorg survival test is a useful diagnostic: if the executive sponsor left tomorrow, could each champion independently defend the program's value to the new organizational structure based on documented functional evidence?