From Pilot to Profit: The COO's Case for AI Ownership
I’ve looked across industries and the evidence is clear: AI’s ROI stalls when it’s not in ops’ hands.
The enterprise landscape is littered with "AI-aware" companies. They have pilots, posters, and impressive press releases. What they don't have is P&L impact. Around 70-90% of AI projects never scale beyond the lab, and 95% fail to produce measurable financial returns. Research from RAND shows AI projects fail at twice the rate of traditional IT initiatives.
This isn’t a technology problem. The real breakdown sits in governance, ownership, and adoption. AI keeps stalling because the wrong teams own it. (Sorry, guys)
Who Owns AI Matters
The data on this disconnect is unavoidable. When ownership aligns with outcome accountability, the numbers move. The further ownership moves from pure IT and closer to the business, the higher the rate of success.
Why Ownership Determines Success
Assigning ownership isn’t just a political choice. It directly determines whether AI projects deliver measurable business outcomes or remain stuck in pilot purgatory. The following explains why the right ownership structure drives success and why IT-led models consistently fall short.
An IT-led approach focuses on stability, security, and cost control. Its success is measured in system uptime, availability, and IT cost reduction. That's the right metric for IT, not for enterprise transformation.
Operations-led ownership focuses on performance, margin, and speed. Its success is measured in business outcomes such as cycle time reduction, margin improvement, and customer satisfaction.
This isn't a new story. ERP only delivered results when it moved from an IT project to a business-driven process re-engineering initiative. DevOps only worked when responsibility was shared between development and operations. AI is following the same path.
IT-Led AI Failures: The Data Speaks
The following cases highlight the structural and organizational barriers that prevent these projects from achieving real business impact.
Pilot Purgatory: Many companies run hundreds of experimental AI use cases that never reach production. One study found 88% of all AI pilots fail to scale. Experimentation alone isn't enough without authority to integrate into operations.
Investing Before Understanding: An IBM survey found 64% of CEOs admitted to investing in AI before fully understanding its business value. This reflects a technology-first approach that doesn't align with enterprise outcomes.
Operational Mismatch: In industrial AI, over 60% of IT-led projects stall because the architecture, while technically sound, fails under operational conditions. Systems built without operational context cannot deliver results.
Positive Proof: Operations-Led Wins
When AI ownership moves to operations, outcomes improve dramatically. The focus shifts from deploying technology to improving processes, which drives measurable business impact.
Manufacturing (EBITDA Impact): A European equipment manufacturer's COO led a generative AI review of core operations across manufacturing, procurement, and supply chain. The initiative produced a prioritized roadmap for €300 million in EBITDA improvement. Only an ops-led mandate could enforce this cross-functional change.
Insurance (Process Speed & CX): An insurer implemented AI agents for end-to-end claims processing. The result was a 40% reduction in claim handling time and a 15-point increase in Net Promoter Score. The ROI came from linking operational processes directly to executive KPIs.
Supply Chain (Efficiency): Moglix deployed AI for vendor discovery within the sourcing team. This operations-led effort achieved a 4x improvement in team efficiency. Ownership of both the problem and the solution enabled real workflow transformation.
Winning the AI Turf War
Undoubtedly, shifting AI accountability to operations will create friction. IT’s concerns about security, compliance, and data governance are valid. At the same time, letting each business unit run AI independently leads to a fragmented, high-risk environment.
The solution is a hybrid Hub-and-Spoke model. In this setup, the COO leads the central hub, which sets governance, standards, and enterprise strategy, while business units operate as spokes within those guardrails.
Key steps for COOs to secure operational authority:
Establish Authority: Create a central AI Center of Excellence (CoE) that reports directly to the COO. This hub centralizes strategy, oversight, and technical resources.
Trade Metrics for Mandate: Implement a new scorecard. Operations takes responsibility for business outcomes, such as margin and cycle time, while IT retains responsibility for platform metrics like uptime and security. This ensures accountability aligns with financial impact.
Codify Authority: Formalize a charter that defines decision rights and funding for both hub and spokes. Clear roles prevent overlap and disputes.
Align Incentives: Use a co-investment approach. When business units co-fund the hub, they become stakeholders in success rather than passive participants.
The New Scorecard: Measuring What Matters
A new operating model requires a new scorecard. Tracking the wrong metrics keeps AI stuck in IT mode. Operations-led AI must be measured like a business driver, not a technical utility.
These KPIs must be reported in C-suite reviews. Research from McKinsey shows that CEO-level oversight of AI correlates most strongly with real financial impact.
Risk Management and Guardrails
Centralizing AI governance under the COO doesn't compromise security or compliance. The Hub-and-Spoke model strengthens oversight while enabling operational impact. Key guardrails include:
Centralized Model Governance: The hub adopts a formal framework such as the National Institute of Standards and Technology (NIST) AI Risk Management Framework to govern the AI lifecycle.
Operational Controls: Mandatory Human-in-the-Loop escalation paths for high-stakes decisions prevent unchecked automation.
Mandatory Audit Trails: Immutable logging of all agent and model decisions ensures accountability and traceability.
Roadmap: Taking Operational Ownership
Phase 1: First 90 Days – Establish the Foundation
Action: Launch a holistic AI Readiness Assessment, establish the COO-led CoE, and formalize the AI Charter with C-suite ratification.
KPI: Charter ratified and readiness baseline established.
Rationale: Foundation matters. Without a baseline and formal governance, pilots remain experimental.
Phase 2: First 9 Months – Build Credibility with Wins
Action: Shift the CoE from doing to enabling. Embed spoke teams in 2–3 priority business units and scale a high-ROI operational project from pilot to production.
KPI: 2–3 pilots fully scaled, performance tracked on the operational scorecard.
Rationale: Quick wins demonstrate value, fund future projects, and build momentum for broader adoption.
Phase 3: First 18 Months – Scale and Institutionalize
Action: Mature into a federated model where business units operate autonomously within hub guardrails. Use early wins to fund a self-sustaining AI program.
KPI: Program is P&L positive with measurable enterprise-wide impact on margin and operating expense.
Rationale: Scaling reinforces authority, embeds accountability, and creates a continuous cycle of improvement.
Rebuttals and Anticipated Objections
Even with a clear operational mandate, you'll face skepticism. IT leaders, business executives, and even board members may question whether this model can handle risk, data ownership, or technical complexity. The following addresses the most common objections and explains why the Hub-and-Spoke approach is designed to succeed where IT-led initiatives often fail.
Objection 1: "Operations cannot manage model risk." Response: The Hub-and-Spoke model centralizes risk governance in the Ops-led CoE. By enforcing the NIST AI Risk Management Framework for all projects, risk is controlled and transparent. This is safer than leaving IT pilots siloed and unmanaged.
Objection 2: "IT owns the data." Response: The hybrid model creates a partnership. The Hub (Ops/IT) owns the platform and infrastructure. The Spokes (business units) own domain-specific data stewardship and are accountable for data quality. This ensures both control and operational relevance.
Objection 3: "Operations lacks AI talent." Response: The structure addresses this directly. The Hub centralizes scarce technical and data science talent. The Spokes provide critical domain knowledge and operational expertise. This combination delivers results far more reliably than an IT-led team guessing operational needs.
Mandate for Change
AI won't generate predictable enterprise value unless accountability aligns with outcomes. Technology alone doesn't create profit—ownership does. The data is clear: operations-led AI consistently delivers measurable business impact, while IT-led initiatives often stall in pilot purgatory.
The question every executive must answer is this: does your AI team drive enterprise profit or just platform uptime? Only one of those truly matters to the Board and only one will scale.