Operationalizing GenAI at Enterprise Scale: What Our AI Experts Have To Say

The generative AI enterprise transformation is in full swing, and every company is racing to figure out what it means, how to use it, and how to move quickly enough to keep up with the market. Yet beneath the urgency, one truth has become clear: experimentation is easy. Operationalization is not.

Most companies aren't struggling with ideas. They're piloting GenAI across teams, launching copilots, automating support functions, and building internal tools. But behind the flurry of activity, very few are seeing measurable impact.

Industry data reveals the scale of these challenges:

• 80% of AI projects fail due to unclear problem statements
• 1/3 of companies haven’t trained a quarter of their employees on AI usage
• 48% of executives lack quality data for enterprise-wide adoption

These are not technical limitations, but signs of low organizational readiness.

At Propeller, we've seen this dynamic play out across the tech industry organizations and clients we work with. In a recent roundtable discussion with consulting leaders Mark Fitzgerald, María Sara Roberts, and Addison Iwanaga, we explored what it really takes to close the gap between GenAI ambition and enterprise-scale value.

We’ll cover the five critical areas where organizations must focus to move from pilot to production:

1. Understanding where tech companies currently stand
2. Building operational foundations
3. Creating governance structures that scale
4. Transforming workforce dynamics
5. Measuring Real ROI

# 1. Where Tech Companies Stand With AI: Leading in Experiments, Lagging in Scale

Tech companies have been among the fastest to explore GenAI because they have the culture, infrastructure, and appetite to pilot early. That experimentation muscle has been a strength, but it has also brought about challenges.

# The Pilot Purgatory Problem

Many early pilots were launched in isolation, focusing on individual teams without planning for broader organizational integration. This dynamic creates "pilot purgatory." Organizations are flooded with GenAI experimentation but lack the structure, alignment, or prioritization needed to operationalize their efforts. Teams get caught chasing the next new thing rather than scaling what works.

Poll data from the session reinforced this trend. Most respondents described their organization's GenAI posture as "promising" or "experimental." Very few selected "mature" or "scaled." This pattern signals that experimentation alone isn't translating to embedded value.

# Three Key Shifts for Better Pilots

To break this cycle, companies should make three key shifts when designing and vetting pilots:

• Broaden pilot scope vertically, not just laterally. Design pilots that cut across functions to expose cross-functional gaps and surface critical friction points early.
• Start impact conversations upfront. While exact ROI may be hard to predict, teams should align on a directional understanding of feasibility and business value before investing heavily.
• Incorporate business and people change from the beginning. Many pilots fail not because the technology didn't work, but because organizations neglected to invest in the process and behavior changes required to support adoption at scale.

Mark Fitzgerald describes why so many GenAI pilots get stuck in “pilot purgatory.”

# 2. Operationalizing AI: From Pilots to Everyday Workflow

Moving beyond successful pilots requires a fundamental shift in approach. If experimentation is the easy part, operationalization is where most organizations stumble. Companies get swept up in GenAI excitement, launching flashy demos or building standalone tools while overlooking the harder work of embedding AI into decision-making and daily workflows.

# The Three Foundations of Operationalization

Operationalizing GenAI rests on three critical foundations:

• Governance: Strong AI efforts begin with curiosity, but scaling requires structure. Governance defines what's allowed, prevents tool sprawl, and addresses safety risks like hallucinations. It also clarifies access: who can use GenAI internally and what systems can connect to it.
• Integration: AI adoption slows when it feels like another disconnected tool. GenAI should fit naturally into existing workflows without introducing friction and requiring employees to work around established systems.
• Architecture: Flexibility is essential given the rapid technology evolution. Establishing a clear baseline for "good enough" helps organizations move forward, while adaptable architecture enables evolution alongside the technology.

These challenges were evident in our webinar poll data. Respondents identified their main barriers as difficulty integrating AI into existing workflows, unclear ROI, and lack of clear ownership — highlighting that the most significant obstacles are organizational, not technical.

# People, Process, and Change: The Real Lift of Operationalizing AI

These technical enablers only work when paired with a people-first approach. Different stakeholders across the organization have varying priorities and comfort levels with change, and successful operationalization requires understanding and addressing these diverse needs rather than imposing a one-size-fits-all solution.

Related Content: Center People and Processes in Your Next AI Implementation

# Data Readiness: The Silent Blocker

One commonly overlooked aspect of operationalization is internal data readiness. Organizations often underestimate the effort required to extract information from various systems and make it usable across different business levels. Inconsistent architecture, governance gaps, and unclear access controls can delay implementation or necessitate costly rework.

Related Content: Transforming Organizational Data into AI-Ready Assets

Addison Iwanaga outlines the three foundations that turn pilots into a daily workflow.

# 3. AI Governance & Org Models That Scale: Turning Energy into Enterprise Value

Beyond operational foundations, successful GenAI scaling requires clear organizational ownership and governance structures. Without these elements, even the most promising pilots remain trapped in experimental limbo.

In our webinar, most leaders said IT or Engineering currently "owns" GenAI, with no clear owner coming a close second. Only a small slice pointed to Data/AI teams or Business Units. The data confirms that ownership is still fragmented—and this fragmentation is slowing momentum.

Without clear ownership, GenAI efforts stall in pilot mode. Tools may be developed and tested successfully, but remain disconnected from business goals, underfunded, or duplicated elsewhere. Organizations making real progress treat AI like a core business capability with budget, ownership, and a clear path to scale.

# Effective Organizational Models

Several structural models prove especially effective:

• AI task forces or centers of excellence act as internal service layers, providing standards, templates, and support for teams building and deploying GenAI solutions.
• Federated innovation models balance consistency and speed. A central group sets policies and ensures responsible use, while individual teams adapt tools and approaches based on unique business needs.
• Cross-functional AI squads embed GenAI into delivery teams that move fast without cutting corners, bringing together product, data, legal, and UX expertise.

These structures accelerate adoption by promoting reuse. Instead of starting from scratch, teams build on vetted tools and infrastructure that already meet internal standards — a mindset shift toward "why not use what we already have?" before building something new.

Executive alignment also plays a critical role. While CEOs often see AI as strategic, there's a significant drop-off in GenAI readiness across the rest of the C-suite. Without broad leadership buy-in, sustaining investment and prioritizing cross-functional coordination become difficult.

Related Content: AI Governance: How an AI Council Could Make or Break Your Strategy

María Sara explains why clear ownership and Centers of Excellence are prerequisites for scale.

# 4. Workforce Transformation: Designing Human & AI Teams

Beyond organizational structure, GenAI's most profound impact lies in how it's fundamentally reshaping work itself. AI will eliminate some roles, create new ones, and force companies to rethink how teams are organized and what skills they prioritize.

Rethinking Jobs as Collections of Tasks

Rather than viewing jobs as fixed roles, leaders should break them down into tasks. Some tasks are repetitive or low-value — ideal for automation. Others are strategic, creative, or relational — areas where human strengths continue to shine. AI serves two key functions:

1. Taking over routine tasks entirely
2. Acting as an assistant to support people in doing more of what they do best

Related Content: How to Prioritize Agentic AI Use Cases That Scale

# Managing the Human Side of Change

But the shift isn't only structural—it's also deeply personal. People often equate their value with the tasks they do. When those tasks are automated, the change can feel like a threat to identity and worth. Leaders need to reframe the conversation around the team's contribution to the business mission, not whether a person owns every step of a process.

Moreover, resistance often fades when AI is positioned as a way to remove the most mundane parts of a job. When people see that AI frees them up to focus on higher-value, more enjoyable work, they're more likely to engage.

Success, though, requires clear roles, expectations, and accountability frameworks as AI becomes part of team workflows. Leaders should define when AI can be used, who's responsible for reviewing output, and how decisions should be made in a shared environment. Thoughtful change management is essential — people need the tools, support, and psychological safety to learn and experiment.

Related Content: Why Treating AI as a Teammate Unlocks Its Full Potential

Mark and Addison discuss redesigning roles into tasks and managing a blended human-plus-AI workforce.

# 5. Measuring AI ROI: Moving Beyond Productivity Metrics

Even with strong governance and workforce planning in place, organizations still face the persistent challenge of proving GenAI's financial impact. Adoption may be high, but many organizations struggle to translate experimentation into bottom-line results.

# A Multi-Dimensional Approach to Value Measurement

Organizations leading on AI maturity approach value measurement systematically, looking across three key dimensions:

• Operational Health - System stability, controlled latency, hallucinations, and error rates. Metrics like uptime, automated evaluation scores, and adoption rates track whether AI runs well enough to support scale.
• Business and Financial Impact — Real value comes from translating productivity gains into measurable business outcomes. Are you improving forecasting? Increasing revenue? Reducing effort in ways that impact the P&L?
• Strategic and Risk Reduction - As GenAI scales, it should support the organization's strategic goals through enabling growth, reducing security exposure, or scaling safely through stronger data governance. This category is harder to measure, but critical for long-term adoption.

Despite excitement around GenAI, many companies aren't seeing the expected financial impact. Teams default to measuring what's easy — latency, token costs, output volume — instead of what matters, creating a disconnect between usage and business value.

# A Practical ROI Example

To address this gap, some organizations are shifting how they define success. A practical example: A team using a GenAI co-pilot tool measured a 56% productivity uplift across 2,000 developers, translating to 30 minutes saved per person per day. Multiplied by workdays and loaded labor rates, this amounts to approximately $200,000 in time saved. When compared to the $45,000 annual license cost, the investment returned more than 4x in value — a clear case for scale.

This kind of ROI story resonates with modern CFOs, who want to understand whether the tools are producing outcomes commensurate with their cost.

# Best Practices for ROI Measurement

To improve ROI measurement:

• Define hypotheses early. Even a rough guess at expected value gives teams a target to compare against.
• Build multi-layered scorecards. Track technical performance, usage, and business outcomes side by side.
• Tie governance to ROI. Consider safety, compliance, and operational risk in your ROI calculation — not just cost.
• Right-size the tech stack. Use smaller models where possible. Split complex tasks into components and ask which parts require advanced models and which don't.
• Enable feedback loops. GenAI performance improves over time. Make iteration part of your measurement process.

Related Content: Measuring AI ROI: How to Build an AI Strategy That Captures Business Value

María Sara explains the 4x ROI rule and how to convert productivity gains into real dollars.

# Conclusion: Scale Comes From What Surrounds the Model

GenAI scales when the organization around it is ready.

That readiness comes from investing in the fundamentals that help organizations absorb and apply GenAI effectively:

• Strong governance that defines ownership, risk tolerance, and decision rights
• A culture that supports experimentation, iteration, and learning
• Workforce planning that addresses how jobs, teams, and responsibilities will evolve

Organizations that scale GenAI successfully stay grounded. They keep the business problem front and center, define desired outcomes, and recognize that scaling is not about adding more tools. It's about reshaping how the organization works.

When GenAI becomes part of how people think, collaborate, and operate — not just a tool they use — it becomes infrastructure. That's when the real value starts to show up.

# Let's Talk About Your GenAI Roadmap

Curious how these principles apply to your organization? Book a conversation with Propeller's AI experts to accelerate your journey from pilot to enterprise-scale impact.