AI Is Not the Strategy: Why Process-First Thinking Still Wins

It’s a story that has become all too familiar: A shiny new AI capability grabs attention within an organization. A promising use case is identified, and a team is assigned to develop and deploy. The pilot is launched quickly and often successfully, at least on the surface. The output looks promising—the tool works.

And then, very little changes. Adoption stalls. Teams revert to old ways of working. The tool becomes another isolated solution in an already crowded landscape. What went wrong? In most cases, the issue isn’t the technology itself, it’s the sequence of the design and deployment process.

In February 2026, APQC members convened for a roundtable discussion on how digital tools and emerging technologies are reshaping the process management landscape. APQC’s Madison Lundquist (Research Lead, Process and Performance Management) led the conversation that explored key themes including AI governance, integrating AI across processes, and identifying where digital capabilities can drive greater efficiency and performance.

Panelists included Jeff Varney (Director of Advisory Services, APQC), Jim Zurn (Sr. Manager, Process Engineering, CVS), and Dave Jackett (Digital & Transformation Supply Chain Leader, General Mills). Each provided valuable perspective from the field on how organizations can best leverage technology to improve business outcomes. Across industries and roles, one message came through consistently: Organizations are willing to embrace AI, but are struggling to apply it effectively within the context of process. This article is the first of two in a series designed to help organizations move from experimentation to sustainable impact.

The Most Common Mistake: Putting Technology First

The most common mistake in AI-enabled transformation is leading with technology before process. Instead of clearly designing how a process should operate, teams attempt to design solutions around vaguely defined problems. This “technology-first” approach often results in fragmented implementations, unclear ownership, and low adoption.

AI hype exacerbates the issue. When the tools themselves are positioned as inherently transformative, organizations skip foundational design work, assuming the technology will compensate for process gaps. In practice, the opposite occurs. AI amplifies the underlying process conditions. Poorly defined processes produce inconsistent outputs at greater speed. Misaligned ownership leads to localized optimization rather than end-to-end value. And without communication on desired outcomes and a focus on the people who will use the technology, even technically successful deployments fail to change how work gets done.

This dynamic shows up clearly in how organizations evaluate new tools. Rather than focusing on features or novelty, leading teams assess whether a solution materially improves how the process performs. As Zurn explained, “Ultimately, we look at whether the tool improves critical-to-quality elements and makes the process more reliable and consistent.” This discipline helps avoid deploying technology that operates adjacent to the process instead of improving it.

From an advisory perspective, Varney sees a similar pattern across organizations. Many expect technology to compensate for unclear process design, which creates more complexity over time. As he noted, “Technology is not a silver bullet. It requires a lot of work and diligence.”

A related challenge emerges when early adopters move faster than governance structures. While experimentation is necessary, it can create divergence—multiple tools, inconsistent approaches, and limited scalability—if not anchored to a shared process framework. 

The Solution: Lead with Process

AI accelerates and exposes gaps in whatever process you apply it to, for better or worse. This means that in order to design effective solutions, you must first analyze your processes and identify exactly what the tool will solve for.

Organizations that realize value from AI treat process design as the primary system and technology as an enabler within it. This includes embedding process owners and change agents directly into technology development efforts. This ensures that digital solutions are built around defined process outcomes rather than abstract capabilities.

In practice, this often requires a more disciplined approach to how tools are introduced. Zurn described a structured lifecycle in which solutions are piloted, evaluated for risk and fit within the IT environment, and then reassessed after implementation to confirm they are delivering value. “We separate hype from reality by piloting tools and then reevaluating within six to nine months,” he explained. This ensures that tools are not just deployed but integrated into how the process operates over time.

Alignment between process architecture and digital roadmaps is also critical. Without it, organizations optimize individual components rather than the end-to-end system. Equally important is defining adoption upfront. Leading organizations establish clear expectations for how both the process and the technology will be used and track that usage visibly, often through leadership dashboards.

Jackett emphasized that this process-first mindset also shapes how AI capabilities are developed. Rather than expecting immediate performance, his teams iteratively refine outputs based on how decisions are actually made within the process. “Think of agents like interns. When they first come in, they're going to need a lot of coaching,” he noted. This approach anchors technology in real operational context rather than abstract capability.

"Measuring what matters" becomes increasingly important in this environment. Instead of focusing solely on tool performance, organizations track value at the decision level:

• Are decisions faster?

• Are outcomes more consistent?

• Is risk reduced?

With the right metrics in place, a tool will be deemed successful not merely because it was deployed, but because of sustained, standardized use within a well-defined process.

The Hidden Metric: Adoption

Failures in digital transformation are often misdiagnosed as technology issues. In reality, they stem from unmanaged adoption—tools that exist but are not meaningfully integrated into how work is performed.

Adoption, not capability, is the most reliable indicator of whether AI will deliver value. But many organizations underestimate how uneven the adoption curve can be. Early adopters often embrace new tools quickly, while broader use across the enterprise lags due to unclear expectations, lack of trust, or misalignment with existing workflows. Experimentation outpaces governance, and individuals or teams deploy tools independently, creating fragmentation that ultimately slows enterprise-wide adoption.

Zurn emphasized that even technically strong solutions fail if they cannot scale across teams. “Adoptability is critical. If people can’t realistically use it, scale it, or share outputs across groups, it won’t stick.” His experience highlights that adoption is not just about willingness—it is about whether the tool fits into the realities of enterprise work.

Jackett described how adoption evolves over time, particularly when AI is introduced into decision-making processes. Early adoption may be limited while users build trust in the outputs. As he noted, agents require “a lot of coaching” before they reliably reflect how experienced professionals make decisions. Only after that iterative refinement does adoption increase and automation expand.

High-performing organizations address this by making adoption visible and accountable. A tool used by a small group may demonstrate potential, but real value emerges only when it is embedded into shared processes and used consistently across teams.

AI as an Intern: Rethinking the Role of Agents in Process Work

To move beyond hype, organizations need a practical mental model for AI agents. A useful framing has emerged where users can think of AI agents as interns who can:

• Handle structured, repeatable tasks

• Require guidance and tuning

• Need guardrails and supervision

• Improve with feedback and iteration

• Should not be handed mission-critical authority on day one
   

The intern analogy reflects real implementation experience, where AI agents are introduced with clear expectations for oversight and gradual responsibility. “There’s always going to be some form of human-agent teaming,” Jackett points out, and it's important to define roles on both sides.

In early deployment, agents often struggle because organizations expect them to operate autonomously right away. Instead, it's critical to treat AI as part of a human-agent team:

• Humans define the process.

• AI supports decision-making.

• Adoption is earned through trust.

• Automation increases only after consistent performance.

Varney reinforced the importance of maintaining realistic expectations for AI autonomy. Organizations that succeed treat AI as part of a broader system that requires governance, training, and iteration. As he noted, “Technology is not a silver bullet,” but over time, agents may take on more responsibility, allowing human roles to shift from execution to oversight and governance. That transition is deliberate and requires upfront training and calibration to deliver long-term efficiency gains.

AI Accelerates Good Process—It Doesn’t Fix Broken Ones

AI delivers the greatest value in environments where foundational elements are already in place:

• Clearly defined and standardized processes

• Trusted, high-quality data

• Aligned governance structures

• End-to-end process ownership

Zurn’s evaluation criteria reinforce this principle. By focusing on whether tools improve reliability and consistency, his team ensures that AI enhances a stable foundation rather than masking underlying issues. “Ultimately, we look at whether the tool improves and makes the process more reliable and consistent.”

And Jackett’s experience with AI-driven recommendations further illustrates the point. Early outputs required refinement before they aligned with real-world decision-making, reinforcing that value emerges only after the process and technology are calibrated together. As he noted, AI agents need coaching before they can be trusted at scale.

In these contexts, AI can reduce cycle time, automate decisions, and improve consistency at scale. Organizations have reported significant efficiency gains, including large reductions in manual effort and substantial financial impact tied to improved decision-making. By contrast, in highly adaptive or craft-based processes—where tacit knowledge and contextual judgment dominate—AI plays a more limited role. It can assist but not fully replace human expertise.

The common denominator: Ensuring process excellence before technology adoption lays the foundation for sustainable transformation.

Solving the Process + AI Puzzle

AI is an accelerator, not a substitute for process design. Looking ahead, the capabilities that differentiate organizations will not be tied to specific tools, but to how effectively they integrate each piece of the process puzzle:

• Process design

• Technology enablement

• Data architecture

• Governance models

• Human expertise

To ensure your organization is leveraging AI in a way that reflects process-first thinking, consider the following before undertaking your next technology deployment:

• Identify processes from an end-to-end perspective and define what needs to improve before designing the tool.

• Engage process owners early in the development process. This not only ensures a more relevant, useful tool, it aids in securing buy-in and adoption later.

• Make adoption expected, not optional. This requires leadership support and visible usage metrics and dashboards.

• Measure the success of the deployment with key metrics that focus on business value, expected outcomes, and better decision-making.

By putting process before technology, you can prevent new tools from languishing unused and instead be highly adopted and useful solutions in your organization's AI toolbox.