AI Has Gone Corporate, but It's Still Not Pulling Its Weight

Three years after generative AI crashed into the enterprise world like a caffeinated intern with a PowerPoint deck, most companies are still fumbling through the same awkward pilot phases they were in last year. That is the topline from McKinsey’s latest global AI survey, an annual pulse check on how close we are to the machine overlords, or at least to AI doing something genuinely useful at work.

Spoiler: We are still waiting.

McKinsey’s November 2025 report, “The State of AI: Agents, Innovation, and Transformation” (read it here), reads like the corporate version of a teenager who just discovered ChatGPT and believes it can write their term paper. Ambition is sky-high. Execution is not.

Let us break it down.

Everyone’s Using AI. Nobody’s Scaling It.

According to the report, 88 per cent of surveyed organisations are using artificial intelligence in at least one business function. That is a tidy leap from 78 per cent last year. However, a clear pattern emerges on closer inspection. Most implementations are still stuck in the shallow end of the pool. Only one-third of companies say they have begun to scale AI across the enterprise.

In other words, there are plenty of proof-of-concepts, but not much in the way of production-level impact. Companies are treating AI like a beta app. It is fun to demo, but too unstable to trust with actual business processes.

And when McKinsey refers to “scaling,” it does not mean widespread adoption across departments. In many cases, it means someone in IT automated a service desk ticketing system and declared it “transformation.”

Meet the Agentic AI Hype Machine

One of the major themes in the report is the emergence of “agentic AI.” This refers to autonomous systems capable of planning, deciding, and executing multiple steps without human involvement. Think of automated processes like rescheduling meetings, reordering inventory, or preparing data for financial forecasts.

Sixty-two per cent of respondents say their organisations are at least experimenting with AI agents. That might sound impressive, but only 23 per cent are attempting to scale them. Among those, most are doing so in just one or two business functions.

These systems are not broken. The challenge lies in plugging them into tangled legacy infrastructure, dealing with conflicting priorities, and overcoming human resistance to change. Even McKinsey concedes that progress is inconsistent. IT and knowledge management are leading the way, mostly because they already have structured workflows. Introducing agents into supply chain operations or marketing departments is a far messier proposition.

The Corporate AI Stack Is Mostly Built on Hope and Slideware

So what is holding enterprise AI back? The short answer is everything.

McKinsey’s so-called “high performers,” a mere 6 per cent of companies that report more than 5 per cent EBIT (earnings before interest and tax) from AI, tend to have several features in common. They invest in redesigning workflows, incorporate human oversight, commit substantial budgets, and build robust platforms with governance structures to support AI use. These elements are unglamorous and expensive, but essential.

Everyone else is dabbling.

Despite the widespread use of AI, only 39 per cent of companies report any measurable EBIT impact. Most of these gains are modest. The rest are still chasing cost reductions rather than pursuing innovation. According to the report, organisations that see the greatest return are the ones using AI to drive growth, improve customer satisfaction, and differentiate themselves from competitors.

As McKinsey’s Tara Balakrishnan puts it, “What stands out most about the high performers is their level of ambition.”

If You Want AI to Work, You Have to Rewire Everything

This is not hyperbole. High performers are nearly three times more likely to have fundamentally redesigned their workflows to support AI. Not just tweaked or patched, but completely rebuilt.

The report references McKinsey’s Rewired framework, which outlines six pillars required for real digital transformation. These include strategy, talent, operating model, technology, data, and scaling. Without aligning all six, AI efforts are unlikely to succeed.

One crucial practice is keeping humans in the loop. Organisations that establish clear rules for when human oversight is needed are far more likely to extract value. AI does not eliminate people; it enhances them. This is not HAL 9000. It is more like Clippy after a doctorate.

Concerns about job losses are not unfounded, but they may be exaggerated. Only 32 per cent of respondents expect workforce reductions of more than 3 per cent next year. In fact, large companies are actively hiring for AI-related roles, especially in software engineering, data engineering, and machine learning.

Risks? What Risks?

This is where the disconnect becomes especially obvious. More than half of respondents report experiencing at least one negative consequence from AI. Inaccuracy is the most common issue. Despite this, explainability, or the ability to understand how an AI model makes decisions, remains one of the least-addressed risks.

That is astonishing. Imagine allowing an opaque algorithm to decide who gets a loan or which customers are worth pursuing, and then failing to question its logic.

To their credit, high-performing companies take risk management more seriously. This is partly because they are more likely to use AI in high-stakes scenarios, such as legal, financial, and intellectual property domains. They have more to lose, and they know it.

The Money Question: Is AI Worth It?

Among high performers, one-third spend more than 20 per cent of their digital budgets on AI. For comparison, only 10 per cent of the rest do the same. This is not casual experimentation. It is a deliberate bet, and it appears to be paying off.

These companies are not just cutting costs. They are generating revenue, building better customer experiences, gaining market share, and improving employee productivity.

However, the majority of organisations are not prepared to make such deep investments. They want AI to magically transform their businesses without the heavy lifting. That is why 70 per cent report using AI across multiple business functions, but fewer than half have actually redesigned those functions to benefit from the technology.

AI’s Corporate Era Is Real. The Results Are Still Loading.

McKinsey’s 2025 AI report captures a corporate culture caught in transition. The tools exist. The intent is forming. But the outcomes are lagging.

There is an increasingly visible gap between organisations that talk about AI and those that use it to change how they operate. That divide is only going to grow.

If you are serious about making AI work, then it is time to stop tinkering with prototypes and start doing the hard work. This includes redesigning workflows, building infrastructure, establishing governance, and aligning leadership. Without these, your AI projects are likely to remain just that—projects.

And if your only goal is to save money, be warned. The organisations seeing the greatest returns are the ones using AI to create, not just cut.

Further Reading:

• One Year of Agentic AI: Lessons from the Trenches
• Rewired: Digital and AI Leaders Are Leaving the Rest Behind
• McKinsey’s Full 2025 State of AI Report

Let us hope the 2026 report features more building and less buzz.