Three AI adoption Trends in 2026

Our recent book, 10X ORG, by Alexey Krvitsky, Craig Larman, and me, is intended to be a wake-up call: Leaders should reconsider their org design to obtain the compounding performance impact of AI. 

While most companies are adopting AI in some form, it occurred to me that the projected mass layoffs are not happening (yet). This made me wonder: How are established organizations incorporating AI? A small research study confirmed my expectations: the trend is to add AI on top of the existing organizational design. The short-term AI challenges leaders face now are tool selection, staying in control of the growing number of emerging AI initiatives, and mushrooming shadow IT. Organizational design is not a relevant topic of discussion.

Looking at how larger organizations are actually adopting AI in 2026, three dominant approaches have emerged. They differ in sophistication and ambition. But they share something important: all three, in their current form, actively support keeping the existing organizational design unchanged. 

Use-Case and Centre of Excellence model

The most popular of adoption approach. A central team—typically sitting within an innovation or analytics function—is added to the org chart. Their task is to discover, prioritize, and govern AI initiatives (AI use cases) across the organization. Business domains execute them locally. A governance layer ensures alignment with strategic objectives. A change community acts as the connection between central AI ambition and day-to-day operational business reality. Their role is to signal opportunities for sharing learning and reduce duplication across teams and domains.

This model works. It has been proven at scale across major financial institutions, logistics companies, and infrastructure organizations. It creates organizational memory, builds AI literacy without demanding restructuring, and gives AI investment a legitimate place in the budget process. From an org design perspective, the AI-CoE is a staff function bolted onto the existing organization. It serves the org. It does not change it.

The structural risk lies precisely in its political elegance. By design, the CoE has no mandate to challenge existing structures—even when the use cases it discovers reveal that a handoff between two departments is unnecessary or that a coordination role exists solely to compensate for a process that AI could handle entirely. Over time, the governance board becomes a graveyard for initiatives that are technically sound but organizationally inconvenient, and the discovery pipeline fills with initiatives that will deliver performance gains limited to the scope the teams are working on.

The use-case model is very popular, as it builds capability and organizational confidence without deep organizational redesign. Companies are likely to see a local efficiency and productivity boost, but AI will not provide compounding gains across the whole org as it is contained by the existing organizational limitations.

Agentic AI and workflow redesign

Instead of automating isolated use cases, this adoption approach consists of deploying autonomous AI agents capable of planning and executing multi-step processes end-to-end—across systems, across departments, and without continuous human input. A recent enterprise survey found that while only 8.6% (1) of companies currently have AI agents deployed in production, 85% (2) expect to customize autonomous agents for their specific workflows within two years. The pace of movement here is significant.

What agentic AI actually does to an organization is more subtle than most adoption roadmaps acknowledge. It does not remove silos. It removes the friction between silos for routine work. The complexity is not resolved; it is redistributed and becomes invisible. Handoffs that previously required coordination between three departments are automated. The coordination work is handled by AI. But the three departments remain separate, with their own budgets, managers, and KPIs. The org chart remains unchanged, but the work it was designed to organize has fundamentally shifted underneath it.

This creates a specific kind of organizational tension. Job content changes as employees shift from execution and coordination to judgment, exception-handling, and oversight, but job titles, reporting lines, and performance frameworks are unlikely to keep pace. Middle management layers, which primarily move information and manage handoffs between departments, find themselves structurally preserved but functionally hollowed out. The technology has redesigned the workflow. The organization that the workflow runs through remains unchanged.

Over time, agentic AI makes the org design problem visible in a way that use-case approaches allow organizations to ignore. When a single AI agent does what three departments used to hand off between each other, the question is no longer abstract: why do those three departments still exist in their current form? This approach might lead to organizational redesign after all.

Operating-system level AI

A third approach is still rare but increasingly discussed among the most advanced adopters. Operating-system-level AI is the condition in which AI functions as the organizational infrastructure—continuously coordinating work, information, decisions, and resources across the entire enterprise. This is not a simple feature to add. It requires creating a technical environment in which the organization operates. 

The analogy: Just as a phone's operating system doesn't do one thing but manages everything simultaneously—memory, applications, connections, and power—invisibly and continuously in the background. OS-level AI becomes the layer that runs the organization itself. Routing work, surfacing risk, allocating resources, and connecting systems are all part of the infrastructure that simply operates.

Most people dismiss this approach as unrealistic. Consistent with findings from Deloitte's 2026 State of AI in the Enterprise survey (3), the prerequisites for this level of adoption point to three organizational conditions: clean, connected data across all systems, organizational trust in AI-driven decisions, and — critically — leadership genuinely willing to redesign how the company operates, not merely add AI to existing processes. 

Almost no organizations have all three today. The data dependency alone is disqualifying for most organizations: in most organizations, governance has not kept pace with AI adoption, and the underlying data infrastructure required by agentic and OS-level AI exposes decades of technical debt that use-case approaches could quietly sidestep.

Similar to Agentic AI, implementing OS-level AI without org redesign doesn't fail dramatically. The existing structure strangles the benefits of AI, inhibiting AI from becoming a competitive advantage. Some examples: A decision that should take seconds waits three days for a human approval chain that nobody redesigned. Data that the system needs sits behind a departmental wall that nobody dismantled. A manager overrides the AI recommendation because the AI's conclusion threatens their team's headcount.

Without redesign, this model becomes just another failed use case for which the technology will take the blame.

What This Means for Org Designers

Drawing a new org chart takes an afternoon. The pain is in the execution. Organizations are made of people collaborating and contributing in specific roles. Once a role exists and someone performs it, it develops its own legitimacy, its dependencies, its political weight. For example: removing a coordinator role doesn't just eliminate a box on a chart—it threatens someone's identity, their reason for existing, and their status and power. This is one of the reasons why organizational systems have a powerful gravitational pull toward keeping things as they are.

There is a false assumption running through AI adoption programs today: artificial intelligence can deliver high performance gains without deep organizational change. You can have the efficiency gains, the speed, and the competitive advantage of AI without touching the structure, the roles, the power dynamics, or the reporting lines that have defined how your organization has worked for decades. 

Senior leadership and shareholders have always wanted more than the organization could comfortably deliver. More speed, more efficiency, more results with fewer resources. The AI promise currently seems both seductive and dangerous because it appears to offer everything without any sacrifice. The gains without the pains: the performance without the restructuring.

If I were a business leader, I would love to believe the claim is true. 

Unfortunately, things don’t work this way.

Deloitte’s 2026 report (3) shows that agentic workflows are spreading faster than the governance models designed to manage them, and nearly half of executives report that turning responsible AI principles into operational processes remains an unresolved challenge. This is not a compliance problem. It is an org design problem wearing a technology hat.

The organizations that will realize the most value from AI in the next three years are not necessarily those with the most sophisticated models or the best CoE teams. They are the ones willing to continuously ask the harder questions that AI keeps surfacing but organizations keep deferring: Does our current structure still reflect how value is actually created—or does it reflect how value was created before the technology changed?

Org Topologies proposes to first consider the org design to make it fit for purpose and then adopt AI to get compounding performance gains at the whole company level. This approach requires first slowing down to accelerate later. Often, that’s not a senior manager’s favorite choice. 

AI adoptions that do not consider the org design will amplify the organizational dysfunctions over time. My fear is that AI will become so advanced that it can easily hide problems and operate smoothly even in a messy organizational design. But for how long? I predict this approach will give 10X problems instead of a 10X performance gain.

Roland Flemm, Org Topologies 2026

---

(1) a TechRepublic article on trends in enterprise AI adoption for 2026, which referenced a survey of 120,000+ enterprise respondents conducted between March 2025 and January 2026 by Recon Analytics. https://www.techrepublic.com/article/ai-adoption-trends-enterprise/

(2) enterprise AI adoption, referencing a CIO roadmap. https://aircall.io/en-gb/blog/ai-enterprise-adoption/

(3) The State of AI in the Enterprise Deloitte's 2026 AI report tracking adoption and impact: https://www.deloitte.com/ce/en/issues/generative-ai/state-of-ai-in-enterprise.html