According to the statistics, artificial intelligence has already won. McKinsey, Stanford, and OpenAI say that more companies are using AI, and workers praise its usefulness. From an external perspective, it appears that the task is complete. According to the most recent McKinsey State of AI survey, 88% of businesses now use AI for at least one task, up from 78% a year ago. Stanford’s 2025 AI Index shows that the number of businesses using AI went from 55% to 78% in just one year. OpenAI’s State of Enterprise AI report tells us that three out of four workers using AI say it improves the speed or quality of their work
As the Head of IT Engineering & Operations at a manufacturing company, I see a very different picture. There is a big difference between trying something out and really changing. After crossing this gap, here are my confessions: the things I wish every CEO, CFO, and business leader knew before giving money to the next AI project.
Confession #1: Your Biggest AI Problem Isn’t Your Budget
Most plans focus on getting funding: licenses, GPU clusters, and vendor contracts. But the model is almost never the real problem. Despite the widespread use of AI, McKinsey’s own 2025 data indicates that its financial impact is not significant. The real problems are basic issues like disorganized data, unclear responsibility for results, and old processes that haven’t been updated to take advantage of AI’s abilities.
You can spend twice as much on AI and still fail. The real investment is in how well your company can handle change. The technology is ready, but our business models aren’t.
Confession #2: We are Building a Zoo of Tools, Not an AI Backbone
Most companies have made a bunch of separate solutions or use cases in their rush to come up with new ideas. Sales has its CRM assistant, HR has its policy chatbot, and marketing has its three copywriting tools. These are pilot projects that are charming but fragile.
Data shows that the difference between leaders and laggards isn’t the tools but how well they work together. We need an AI backbone, which includes a shared data layer, a platform for deployment that is governed, and a catalog of AI capabilities that can be used again and again. By connecting to this core, every new use case should become cheaper and faster. We are still a zoo of strange experiments without it, not an intelligent ecosystem.
Confession #3: Regulation Is Here, But Our Governance Is Not
Many boardrooms are wrong to think that AI regulation is years away. The EU AI Act is already in effect. It bans some things and requires powerful models to be open about how they work. Simultaneously, public authorities are formulating basic rules for agentic AI systems. The reality is that the first bans on “unacceptable” AI practices took effect in early 2025, including social scoring and certain types of emotion recognition, according to Le Monde. From August 2025 on, obligations for providers of general-purpose AI models (GPAI) started to apply, with transparency and risk-management requirements for the most capable models, as published by the European Commission.
However, many companies tend to overlook governance. We have AI built into ERPs and SaaS tools, and early agentic workflows are up and running, but we don’t have clear rules for logging, oversight, and human-in-the-loop controls. This is a very important risk for the Head of IT. Regulatory scrutiny won’t just stop at our vendors; it will also include us as the users and controllers.
Confession #4: We still run IT like a service desk in a world that needs a nervous system
Many people still think of IT as a team that resolves problems and manages software. In that model, AI is just another request from a user. But when AI has an effect on credit decisions, pricing, quality control, or predictive maintenance, the stakes are very high.
IT now has to help build the company’s digital nervous system. This means being able to answer tough questions: What real-time data signals do we need? Which decisions do we completely automate, and which do we only help with? How do we explain, check, and fix things when the system makes a mistake? In manufacturing, an AI mistake isn’t just a bad email; it can also mean scrap, downtime, or a safety problem. IT leaders need to be there with Operations, Quality, and Finance from the beginning.
Confession #5: Your AI strategy will fail without a human operating system
The hardest thing to accept is that the technology is the easy part. The long-term success of every AI project depends on “soft” factors that are very difficult to deal with.
Do managers think that AI will take away their power? Do teams receive bonuses for working together or find themselves in trouble for problems that come up? Is there psychological safety to say, “The model is wrong here”
Using AI requires a shift in decision-making processes, authority structures, and job definitions. If you don’t acquire this new human operating system, you’ll have to confront expensive tools and habits that don’t change.
If I could request three strategic shifts from leadership, they would be:
1. Measure impact rather than activity. Don’t ask, “How many AI projects do we have?” “Where in our P&L is AI moving the needle? ” is a good place to start. Every initiative must specify its desired cost, risk, or revenue driver along with a precise measurement strategy.
2. Fund the Backbone, Not Just the Zoo: Give your tech executives the tools they need to create a single, controlled platform for data and AI. Although it seems slower at first, this foundational work is the only way to eventually achieve scalable, secure, and effective transformation.
3. Treat AI as Organizational Change First: Align roles, rewards, and culture with new ways of doing things. Make it safe to try new things and important to question the system.
If these confessions resonate with what you see inside your organization, you’re not alone. In my book Life in the Digital Bubble, I explore how AI and digital systems are reshaping not just IT, but also work, families, and society over the next three decades. And if you’re ready to turn AI from a noisy collection of projects into a clear operating model, my digital transformation and AI consulting services are focused on helping leaders design that next phase with structure, realism, and confidence.