The Governance Outsourcing Problem: What OpenAI's Consulting Alliances Actually Mean

On February 23, 2026, OpenAI announced “Frontier Alliances” --- multiyear partnerships with Accenture, Boston Consulting Group, Capgemini, and McKinsey to deploy its enterprise AI platform. The partnerships divide neatly: McKinsey and BCG handle strategy, operating models, and change management. Accenture and Capgemini handle technical implementation and data integration.

This is a significant structural move. But the conversation about what it means has focused on the wrong question. Most coverage asks whether this helps OpenAI catch Anthropic in enterprise revenue. The more consequential question is simpler: when consulting firms become the deployment channel for AI agents, who owns the governance framework?

The ERP Parallel Nobody Wants to Discuss

We have seen this architecture before. In the early 2000s, SAP and Oracle deployed enterprise resource planning systems through the same consulting firms. McKinsey defined the strategy. Accenture built the implementation. The enterprise wrote the check.

The result was a generation of ERP deployments where the governance layer --- who could access what data, which business rules were encoded into the system, how exceptions were handled --- was embedded in the consulting firm’s methodology, not the enterprise’s operational knowledge. When the consultants left, the enterprise owned the system but not the logic that governed it.

Project failure rates during that era ran between 50% and 75%, according to Standish Group research. The failures were rarely technical. The technology worked. What failed was the alignment between the governance assumptions baked into the implementation and the actual operational reality of the enterprise.

AI agent deployment through consulting firms creates the same structural risk, amplified by a critical difference: ERP systems followed deterministic rules. AI agents follow probabilistic patterns. When an SAP workflow went wrong, you could trace the rule that caused it. When an AI agent makes a poor decision, you need governance infrastructure that can attribute, audit, and reverse that decision. The governance requirements are strictly harder.

What the Partnerships Actually Do

The CNBC reporting and OpenAI’s announcement describe a two-tier model. McKinsey and BCG will help enterprises create “AI co-worker strategies” and “operating models.” Accenture and Capgemini will handle technical deployment --- connecting agents to enterprise data, scaling deployments across organizations, embedding Frontier into existing technology systems.

This mirrors the consulting industry’s standard delivery model. The strategy firms define what gets built. The implementation firms build it. The client operates it.

But Frontier is not an ERP system. It is what OpenAI calls an “intelligence layer” --- middleware that stitches together an enterprise’s CRM, ERP, ticketing systems, and data warehouses to give AI agents unified context. The agents then act autonomously within that context, making decisions, completing tasks, and operating across functional boundaries.

The governance question is who defines the boundaries. When McKinsey creates your “AI co-worker operating model,” they are defining which decisions agents can make, what data they can access, and what oversight applies. When Capgemini connects Frontier to your enterprise systems, they are implementing the permissions, identity management, and audit trails that constitute your governance layer.

This is not a criticism of consulting firms. It is a structural observation. The entities defining and implementing your AI governance are organizations whose primary expertise is project delivery, not sustained operational governance. They will leave. The governance framework they built stays.

The Non-Exclusivity Signal

None of these consulting partnerships are exclusive. McKinsey has been deploying Google’s Gemini since 2024. Accenture partnered with Anthropic in December 2025. BCG and Capgemini work across multiple AI platforms.

This is important for two reasons.

First, it means the consulting firms will develop platform-agnostic governance methodologies. They will not build governance frameworks tailored to Frontier’s specific capabilities and constraints. They will build frameworks that can be reused across OpenAI, Anthropic, and Google deployments, because that is how consulting firms scale.

Second, it means the governance your enterprise gets is not OpenAI’s governance vision or your engineering team’s governance requirements. It is the consulting firm’s methodology. The methodology that worked for a financial services client in Q1 will be adapted for a healthcare client in Q2. Cross-industry governance patterns are useful, but they are not the same as governance designed for your specific risk profile, data architecture, and operational context.

The Futurum Group’s Warning

Five analysts at Futurum Group published a detailed assessment of Frontier on February 9, 2026. Their core observation is worth quoting directly: governance is “the primary determinant of which platforms enterprises trust first. Model capability may attract attention, but governance determines deployment.”

They identified a specific gap: “The main OpenAI security and compliance landing page does not yet even list OpenAI Frontier as a product.”

This is not a minor oversight. It means that as consulting firms prepare to deploy Frontier across enterprise environments, the platform’s own governance and security posture is undocumented. The consulting firms are building governance practices around a platform whose governance surface is still being defined.

Futurum also flags a race dynamic: “The first platforms to convince enterprises they can safely delegate real authority to agents will gain disproportionate traction, even if their agent capabilities are not the most advanced.” Speed to governance credibility matters more than model performance.

The Real Question for Enterprises

The question is not whether OpenAI’s Frontier Alliances are good for OpenAI. They obviously are. The enterprise revenue gap with Anthropic --- roughly 40% vs. 80-85% from business customers --- creates existential pressure to close the distance, and consulting partnerships are the fastest channel.

The question is whether this is good for the enterprises that engage these partnerships.

If your AI governance framework is designed by a strategy consulting firm and implemented by a systems integrator, you have outsourced the most consequential layer of your AI infrastructure. You own the agents. You own the data. But you may not own the rules that govern how the agents use the data.

This is not hypothetical. It is the default outcome of the model OpenAI just announced. Every governance decision embedded in the implementation --- which agents can access which systems, what permissions apply, how decisions are audited, what escalation paths exist --- will be shaped by the consulting firm’s methodology, not your organization’s operational expertise.

The alternative is not to avoid consulting partners. Most enterprises need the scaling capacity, cross-industry experience, and implementation expertise that these firms provide. The alternative is to maintain governance sovereignty. Treat governance as an organizational competency, not a project deliverable. Ensure that the framework the consulting firm implements is one your organization designed, understands, and can operate independently.

What This Signals

The Frontier Alliances signal three things:

Enterprise AI is moving from DIY to managed deployment. The era of internal AI teams building custom agent frameworks is ending for most large enterprises. The consulting channel will become the primary deployment path, just as it did for ERP, CRM, and cloud infrastructure.

Governance will be standardized by consultants, not engineers. As McKinsey, BCG, Accenture, and Capgemini develop AI governance methodologies, those methodologies will become de facto industry standards. Enterprises that do not have their own governance frameworks will inherit the consulting firm’s.

The governance ownership question is now urgent. Once a consulting firm has implemented your AI governance layer, changing it requires re-engaging the consulting firm or rebuilding the framework internally. The window for establishing governance sovereignty is before the implementation begins, not after.

Fernando Alvarez, Capgemini’s chief strategy officer, described the partnerships with disarming clarity: “If it was a walk in the park, OpenAI would have done it by themselves.” He is right. Enterprise AI deployment is hard. But difficulty does not eliminate the need for governance ownership. It makes it more important.

The question every enterprise should ask before signing a Frontier Alliance engagement is straightforward: after the consulting firm leaves, who in your organization understands and can modify the governance framework that controls your AI agents?

If the answer is nobody, the partnership has not solved your governance problem. It has outsourced it.