Why enterprise architects must move AI from experimentation to governed, measurable execution
By Paul Gadbois
Most organizations don’t have an AI strategy problem. They have an execution problem.
The ambition is already there. Boards have approved it. CIOs have funded platforms. Business units are experimenting. Employees are using generative AI in pockets across the enterprise. Vendors are promising everything from productivity gains to autonomous workflows and intelligent agents.
And yet, leadership teams keep coming back to the same unresolved question: How does any of this turn into repeatable business value?
The answer isn’t another roadmap. It’s an operating model.
An AI operating model defines how strategy, capabilities, processes, data, platforms, governance, people, and measurement come together to actually deliver outcomes. It’s the difference between experimentation and execution.
This is where the modern enterprise architect becomes essential—not as a framework custodian or a technology reviewer, but as the connective tissue between ambition and results.
AI Is Forcing Operational Change
The market signal is clear: AI is no longer just a tool—it’s reshaping how organizations operate.
Gartner reported in April 2026 that 80% of CEOs expect AI to significantly change their operational capabilities, accelerating a shift toward autonomous business models. [1] At the same time, McKinsey’s 2025 State of AI research shows that while adoption is widespread, meaningful impact is uneven.
The organizations seeing real value aren’t just deploying tools. They’re redesigning workflows, redefining accountability, and embedding governance into how work gets done. [2]
That distinction matters. A tool improves a task. A platform improves a function. An operating model transforms how the enterprise runs.
AI doesn’t stay neatly within IT. It cuts across business capabilities, data domains, value streams, risk frameworks, and customer experiences. Scaling it safely and effectively requires architectural thinking.
The Real Problem: Under-Architected AI
Many AI initiatives stall because they’re framed too narrowly—as use cases instead of capability shifts.
A use case asks: “What can we automate?” A capability lens asks: “What must we do better as an organization?”
For example, deploying a chatbot is a use case. Improving customer issue resolution is a capability tied to a broader value stream.
That shift in perspective changes the questions leaders must answer: Which value stream is impacted? What outcome actually improves—for customers, employees, or partners? What data is required, and who owns it? Which decisions can be automated versus augmented? What controls are needed for accuracy, privacy, and auditability? What process changes are required first? How will success be measured?
This is where many AI efforts fall short—they’re simply under-architected.
From Strategy to Operating Model
An AI strategy defines intent. An AI operating model defines execution.
A strong operating model answers six critical questions: Value—what business outcomes matter most? Capabilities—what must the organization do differently? Workflows—which processes need redesign, not just automation? Data and Technology—what data, platforms, and integrations are required? Governance—who is accountable for decisions, risk, and performance? Measurement—how will success be tracked and improved?
These elements must work together as a single system.
Strategy without capabilities becomes aspiration. Delivery without governance creates risk. Governance without delivery becomes bureaucracy.
Keeping AI Grounded with Goal GPS™
One practical way to prevent fragmented AI efforts is the Goal GPS™ framework.
It starts with aspirations, and ushers appropriate solutions through design and execution phases to ensure that the desired value is delivered.
It starts with the vision and works downward to ensure alignment across the enterprise: Goal—what are we trying to improve? Critical Success Factors—what must go right? Capabilities—what must we do better? Value Streams—where does the work actually flow? Technology—what enables the change? Governance—how is it controlled and managed? Execution—how is it delivered and sustained?
This approach avoids a common failure: choosing AI tools before understanding the outcomes they’re meant to deliver.
Governance Must Be Built In
AI governance can’t be an afterthought or a final compliance checkpoint.
Frameworks like NIST’s AI Risk Management Framework and regulations such as the EU AI Act make one thing clear: governance must be embedded from the start. [3] [4]
That means defining approval processes for AI-enabled capabilities, clear accountability for outcomes, data, and model behavior, human oversight within workflows, monitoring for performance, bias, and drift, incident escalation paths, vendor evaluation standards, and retirement criteria for outdated AI capabilities.
The goal isn’t to slow innovation—it’s to make it scalable, explainable, and safe.
The Evolving Role of the Enterprise Architect
In the AI era, the enterprise architect becomes a translator between strategy and execution.
Their role is to help leaders understand the enterprise as an interconnected system of capabilities, value streams, data flows, decision points, technology platforms, and governance structures.
AI impacts every layer of that system. It can accelerate outcomes—but it can also amplify flaws.
The architect’s job is to make those tradeoffs visible before they scale. That requires moving beyond static diagrams to more dynamic, decision-focused artifacts such as capability heatmaps, value stream impact maps, AI decision inventories, data readiness assessments, governance models, adoption plans, and value realization metrics.
These tools help leaders decide where to invest, what to prioritize, and how to manage risk.
A Practical Starting Point: The AI Capability Portfolio
One of the most effective ways to move forward is to build an AI capability portfolio.
This isn’t a list of tools—it’s a structured view of where AI can improve the business and how those opportunities should be prioritized.
Each capability should include business outcome, capability impact, value stream location, AI pattern, data readiness, risk profile, human involvement, success metrics, and governance ownership.
This creates a bridge between ambition and execution, giving leadership a clear way to evaluate tradeoffs across value, risk, and readiness.
Practical Takeaways
- Treat AI as an operating model transformation, not a software rollout.
- Anchor initiatives in business outcomes, not experimentation.
- Use capabilities as the foundation for alignment.
- Redesign workflows before applying AI.
- Embed governance from the beginning.
- Build a portfolio view of AI investments.
- Position enterprise architecture as the integration point.
Closing Thought
AI won’t reward organizations that simply deploy more tools. It will reward those that rethink how value is created, governed, measured, and improved. That’s not a technology challenge—it’s an architecture challenge. The organizations that succeed won’t start by asking, “Where can we use AI?” They’ll ask a more important question: What must we become capable of doing?
That’s when and where the real work begins.
About the Author
Paul Gadbois is a technology executive, enterprise architect, author, and advisor focused on helping organizations align strategy, business capabilities, and enabling technology. He is the author of The Modern Enterprise Architect and creator of The Goal GPS™ Framework. Learn more at paulgadbois.com.
References
[1] Gartner. “Gartner Survey Reveals 80% of CEOs Say AI Will Force Operational Capability Overhauls.” April 23, 2026. https://www.gartner.com/en/newsroom/press-releases/2026-04-23-gartner-survey-reveals-80-percent-of-ceos-say-artificial-intelligence-will-force-operational-capability-overhauls
[2] McKinsey & Company. “The State of AI: Global Survey 2025.” November 5, 2025. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
[3] National Institute of Standards and Technology. “AI Risk Management Framework.” https://www.nist.gov/itl/ai-risk-management-framework[4] European Commission. “AI Act.” https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
[4] European Commission. “AI Act.” https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai