The Counterintuitive Finding
According to McKinsey's 2026 Technology Trends Outlook, a significant portion of businesses are hitting a wall with one of the most advanced forms of AI. The report found that productivity actually declined in nearly 30% of companies after their teams
began using agentic AI tools. This runs directly counter to the narrative that new AI systems are a guaranteed ticket to efficiency. The finding brings a critical, and often ignored, detail to the forefront: technology alone is not a solution. The problem, as McKinsey points out, isn't the AI itself, but how it's being implemented—or rather, forced into existing structures.
What Is Agentic AI?
To understand the issue, it’s important to clarify what agentic AI is. Unlike earlier AI, such as chatbots or basic automation scripts that follow predefined rules, AI agents are systems designed to operate autonomously. They can understand a goal, create a multi-step plan, interact with different software systems, and execute complex tasks with minimal human intervention. For example, where a simple bot might log a customer complaint, an AI agent could take that complaint, schedule a follow-up with a support team, update the customer relationship management (CRM) software, and generate a draft response for human review. It’s this ability to plan and act that has made businesses invest billions.
More Activity, Less Value
The core of the productivity problem lies in a misunderstanding of what AI-driven work looks like. The McKinsey report highlights a phenomenon where using agentic AI leads to a surge in activity without a corresponding increase in valuable output. In one study cited, AI tools boosted coding activity by a massive 180%, but the number of actual product releases only rose by 30%. This demonstrates a critical disconnect. Teams are generating more code, more drafts, and more data, but this flurry of AI-assisted activity doesn't automatically translate into more finished products or better business outcomes. It simply creates more work to manage, review, and integrate, overwhelming the very workflows it was meant to improve.
The Missing Piece: Workflow Redesign
The 30% of companies that saw productivity fall made a common mistake: they layered powerful new technology onto outdated processes. Simply plugging an AI agent into an existing, human-centric workflow often creates friction. These legacy workflows have built-in handoffs, review cycles, and approval chains that were designed for the pace and limitations of human teams. When an AI agent executes its part of the task at machine speed, it can create bottlenecks as the work piles up, waiting for a human to complete the next manual step. High-performing companies—those seeing significant productivity gains—do things differently. McKinsey found that top accelerators are more than twice as likely to redesign their processes before incorporating AI. They don't just automate tasks; they fundamentally rethink how work gets done from start to finish.
Building for Success
Effective workflow redesign involves more than just buying software; it requires a strategic overhaul of roles and responsibilities. This means breaking processes down into individual tasks and deciding which are best suited for AI (speed, scale, data analysis) and which require human judgment (ambiguity, strategy, high-consequence decisions). According to the research, companies that successfully integrate AI are 3.3 times more likely to use it to transform their business rather than just chase efficiency gains. They also address the human element directly. The McKinsey report noted a significant trust gap, with 46% of software developers actively distrusting the accuracy of AI tools. Successful organizations close this gap through training, clear governance, and by building verification systems that allow teams to trust, but also verify, AI-generated work.
