We bought AI for everyone but taught no one how to use it

Written by Angelina Darrisaw I founded C-Suite Coach, a leadership development consultancy, in 2015, in part because I saw a disconnect between people and their leaders. Over time, that work has grown to include AI readiness and helping people use artificial intelligence efficiently and ethically. So this year, I paid close attention as companies shifted from rewarding employees for how much AI they used to restricting how much they could use based on spending.

As leaders, we were told adoption was the goal. By March, The New York Times reported that employees at Meta and OpenAI were competing on internal leaderboards that tracked how much AI each person used. By May, The Wall Street Journal reported that some companies had burned through a full year’s AI budget in just three months, and limits soon followed.

Any finance leader seeing that would have set guardrails, and I would have done the same if spending was racing ahead of expectations. Still, while a cap may control how much people use AI, it does not answer the bigger question of how well they are using it. The way we use AI also has costs beyond the office.

In April, the NAACP sued xAI, alleging that the company was operating 27 gas turbines without an air permit to power its data center in predominantly Black communities in the Memphis area. xAI argued that it relied on Mississippi’s determination that no permit was needed. The case is still being decided.

According to a United Nations University report, 80% to 90% of AI’s energy goes toward everyday requests. Each time we ask ChatGPT a question or create a prompt, we add to the demand power plants must meet. In a 2025 national survey by the nonprofit Jobs for the Future, 53% of Black respondents said they used AI tools daily or weekly, most often for their own learning, while 71% said AI made them feel the need to develop new skills.

Corporate training, however, has not kept pace. In a November 2025 survey analyzed by economists at the Federal Reserve Bank of New York, 39% of employed respondents said they had used AI at work in the past year, but only about 16% said their employer offered any AI training. Most people using AI at work were never trained, yet some companies still judged them by how much they used it.

Many companies bought these tools the way they buy other software licenses. They purchased seats, announced a transformation, and trained no one, so adoption became expected and enforced while workers’ skills remained underdeveloped. The gap shows up in McKinsey’s 2026 global survey, where eight in 10 respondents said AI improved their own productivity, but only 37% said it contributed to their organization’s earnings, roughly the same share as the year before.

Using AI well is a leadership skill, and it looks a lot like delegating well. Picture someone typing “write a marketing plan” into a prompt and expecting a masterpiece. A manager who expects people to read minds gets weak work back from both people and AI.

Good managers give clear instructions, provide context, show what success looks like, and review the work before it goes out. Delegating well also means giving feedback when the first draft comes back. When AI gets something wrong, many people abandon the chat and start over, just as a manager might take back an assignment instead of explaining how it could improve.

Strong leaders coach others by sharing what is working and what is not. The same habit leads to better results from AI. How often have we assumed an employee lacked skill when we had never clearly delegated the task in the first place?

Many of us are setting next year’s budgets this quarter, whether we lead a division or own a business with 10 employees. That makes this a good time to connect every AI expense with training requirements. In a small business, wasted AI use shows up as lost time, rework, and mistakes that reach clients, so teaching people to use these tools well directly protects the margin.

I would begin by measuring how employees are using AI today. We already rate buildings and appliances by how efficiently they use energy, and AI use at work deserves the same kind of measurement, which is why my firm built one. One number I would track is output adoption: the share of AI-assisted work that people actually go on to use.

I would measure it again in 90 days. In a 10-person business, that can be as simple as asking each week which AI drafts were sent, which were revised, and which were discarded. I would also ask the most skilled users on each team to teach the people around them.

More than anything, I would ask owners and senior leaders to let their teams watch them review their own AI-assisted work, because teams learn to use tools the way they see their leaders use them. Imagine how different next year’s AI budget conversation would be if it started with what our people know how to do. What if we asked how much of the work AI produced last quarter was actually used?

What if every spending limit came with an hour of coaching? Paying for a tool and leaving people without the skills to use it, while the power it requires is generated in someone else’s neighborhood, is negligence. The people on our teams are already using AI every day, and the least we owe them is to show them how to use it well.

Angelina Darrisaw is the founder and CEO of C-Suite Coach, a leadership development and executive coaching consultancy, and the creator of the AI Efficiency Score, which measures how effectively organizations and their employees use AI.