“Much of the effort and attention around AI for the last several years has been around technical developments,” wrote Babson College professor Tom Davenport in a recent Substack post, “The AI World Moves Toward Organizational Deployment.” “New model announced! New benchmark surpassed! New contract for massive data centers! New world-class technologists hired! You know the drill.”

“I am happy to say, however, that things are beginning to change,” he added. “AI companies are beginning to realize something that many corporate executives knew intuitively. What matters isn’t the technology — OK, that’s important too — but the ability of organizations to deploy it effectively and get value from it.”

That observation captures what I believe is one of the most significant developments in AI today. For the past several years, the conversation has focused overwhelmingly on increasingly powerful models and ever larger computing infrastructures. Those advances remain essential, but they are no longer sufficient. As AI matures, the central challenge is becoming organizational rather than technological: How do companies redesign work, management, and decision making to capture AI’s potential?

Several recent initiatives illustrate this shift.

Last May, Stanford University launched the AI and Organization Lab, “a new research center that will establish an empirical science of how AI transforms workplace coordination and organizational performance.” Led by Melissa Valentine, Professor of Management Science & Engineering, the lab is one of the first major AI research centers focused primarily on organizational rather than technical questions.

“We’re at a critical juncture where AI is being deployed across organizations at unprecedented speed, yet we have limited empirical understanding of its actual effects on how people work together,” wrote Professor Valentine. “This lab will generate the rigorous, evidence-based research needed to guide organizations toward AI implementations that genuinely augment human potential.”

To explore these questions, the lab, together with Google DeepMind, launched the AI for Organizations Grand Challenge, inviting computer scientists and management researchers worldwide to submit proposals on the future of organizational collaboration. More than 200 teams from 156 universities responded.

The winning proposal, — from Stanford Graduate School of Business researchers Yankai Wang and Amir Goldberg, — seeks to build an AI-based model that learns how successful teams coordinate their work and predicts which sequences of actions are most likely to succeed in different situations.

“Coordination unfolds through a series of human interactions,” they wrote. “People send emails, hold meetings, edit documents — but we don’t know what makes one sequence of actions effective in one situation versus another. We want to develop a framework to help leaders understand the dynamics of coordination and make decisions grounded in organizational science instead of having to trust someone’s instinct.”

The emergence of such research reflects a growing recognition that successful AI adoption depends as much on understanding organizations as on improving algorithms.

Davenport also points to initiatives recently announced by leading AI providers that recognize this same reality.

OpenAI, for example, announced Frontier Alliance partnerships with BCG, McKinsey, Accenture, and Capgemini to help customers implement AI at scale. It had previously begun hiring “forward deployed engineers” — software engineers who work directly with customers to customize and deploy AI systems in production rather than simply demonstrating prototypes. Anthropic has similarly created an Applied AI capability in partnership with several private equity firms to establish a new AI services company.

These initiatives acknowledge that customers need much more than access to powerful AI models. They need help integrating those models into existing workflows, business processes, and operational environments. But, as Davenport notes,  these deployment efforts remain heavily focused on technology.

“Granted, neither of these organizations are particularly focused on the human issues involved in putting AI systems into production,” he wrote. “The deployment-oriented roles they have created are primarily focused on technology architecture, optimizing performance, and other technical aspects of AI implementation. The umbrella and somewhat generic term ‘change management’ doesn’t appear anywhere in job descriptions or marketing copy.”

That omission may become one of the biggest obstacles to successful AI deployment.

A recently published Harvard Business Review (HBR) article highlights one particularly important approach for dealing with human issues..

“Most organizations treat AI adoption as a technology challenge — software rollout to be managed by IT and celebrated by the C-suite,” wrote Julia Shin and Sandra Sucher in “AI Adoption Is Overloading Your Middle Managers.” To understand how people throughout organizations were actually using AI, they interviewed partners, managers, and junior consultants at two major consulting firms.

“What emerged was not a technology story but an organizational one,” they wrote. “The pressure point was consistent across both firms. Our research suggests where AI adoption actually succeeds or fails: the middle layer of management.”

Their findings reveal what might be called a capability-reality gap.

Senior partners are focused on AI’s strategic potential — expanding services, accelerating delivery, and reimagining business models. Junior consultants report dramatic productivity gains, with tasks that once took days now taking hours or even minutes.

Middle managers, however, occupy a very different reality. They are expected to validate AI-generated work, identify errors, coach junior employees in AI techniques, establish quality standards, answer clients’ questions about AI, and continue delivering projects under the same, or even greater time pressure.

The HBR article illustrates what the daily routine of a typical middle manager might be like:

• Beginning the day learning new prompting techniques before the team arrives.
• Meeting with clients who want to understand how AI is being used.
• Reviewing AI-generated work for subtle errors and weak analysis.
• Coaching junior consultants who increasingly rely on AI rather than learning foundational consulting skills.
• Documenting successful prompts and workflows so others can reuse them.

As the authors note, “Across interviews, versions of this story came up again and again. … While our research focused on consulting, the patterns we found… are likely familiar to leaders across knowledge-intensive industries.”

The interviews identified three reasons why middle managers are becoming overloaded.

First, learning remains informal while delivery expectations remain unchanged. Managers are expected to experiment with AI, discover effective workflows, and teach others, but organizations rarely provide dedicated time to do so or formal support.

Second, incentives reward the wrong behaviors. Traditional performance metrics still emphasize billable hours and individual productivity rather than mentoring colleagues, sharing successful prompts, or developing organizational AI capabilities.

Third, senior leadership and middle management increasingly operate in different realities. Executives see AI as a strategic opportunity, while middle managers confront the day-to-day challenges of determining whether AI-generated work is actually accurate, useful, and ready for clients.

Perhaps the most important long-term concern is protecting the leadership pipeline.

Traditionally, junior consultants learned by closely observing experienced managers — how they structured complex projects, challenged questionable analyses, exercised judgment, and built trusted client relationships. AI can now generate polished presentations and reports remarkably quickly. But it cannot teach judgment, experience, or professional intuition.

As Shin and Sucher observe in their HBR article, “A junior can now produce a polished deliverable quickly. What still takes time to learn is how to tell when an analysis is plausible but weak, whether the recommendations make sense, or how to challenge a client without losing trust.”

Their conclusion extends well beyond consulting.

“The difference in AI adoption isn’t about the technology,” they write. “It’s whether leadership has built the support structure around the people who make AI work in practice.”

That may ultimately be the defining challenge of the next phase of AI.

The extraordinary progress of AI over the past several years has been driven largely by advances in models, algorithms, and computing infrastructure. Those technological advances will undoubtedly continue. But realizing their economic and organizational value will increasingly depend on a different set of capabilities: redesigning work, developing new management practices, investing in organizational learning, and helping people adapt to new ways of working.

In other words, the AI revolution is entering a new phase. The frontier is no longer just building better AI. It is learning how organizations can deploy it effectively.