AI can make one employee faster and your company slower
The hidden costs leaders may be missing
I learned long before ChatGPT that the promise of technology is usually more complicated than it first appears. For six years, I sold educational technology to colleges and other institutions across the Northeast. A computer based program could assess a student's learning gaps, create a customized learning path, extend learning beyond the classroom and/or augment a lesson plan to include visuals, audio and awesome graphics.
It felt like a game changer. It extended the reach of teachers and helped students who did not learn well through traditional instruction. Once we moved beyond the fear that technology would replace the teacher, educators embraced it.
But administrators and teachers often underestimated what implementation required. Teachers needed preparation. Institutions needed infrastructure. Workflows had to change. Someone had to manage the system, review the results, and integrate it into instruction. The software made one part of learning more efficient while creating financial and human costs elsewhere.
At the same time, I heard students describe something the spreadsheets could not fully capture.
The technology gave them another way to learn. Instead of simply being told what they had done wrong, they could see how to do it correctly, practice, receive feedback, and try again. If it took six attempts, the software did not become impatient or change its tone. The technology was supplementing good teaching without pretending to replace the teacher. Those experiences taught me to hold two truths at once.
Technology can expand human capability. It can also create costs and demands that leaders fail to see when they measure only the most obvious benefit.
We are making the same mistake with AI.
Faster output is not the same as better performance
If AI helps someone produce an email, proposal, analysis, or presentation in thirty minutes instead of two hours, we call the difference a productivity gain. But what happens next? How much time does someone else spend checking facts, clarifying the thinking, correcting the tone, or repairing mistakes?
If one employee saves ninety minutes and three coworkers each lose thirty minutes dealing with the result, the company has not gained anything. It has simply moved the work.
Boston Consulting Group surveyed nearly 12,000 employees and found that 47 percent now spend more time managing and directing AI than doing the work itself. Seventy two percent said AI had changed the skills expected of them, yet only 36 percent believed they had received adequate preparation. That is a revealing gap.
Companies are giving people responsibility for supervising AI without necessarily teaching them how to judge its work, improve it, or recognize when it should not be used. The real cost appears in the manager rewriting the recommendation, the salesperson unable to explain it, or the customer receiving something technically correct but emotionally tone deaf. This doesn't mean AI is failing. It means our definition of productivity is too narrow.
The costs that do not appear on the AI invoice
The subscription price may be the easiest cost to measure. It may also be the least important.
The first hidden cost is rework. Poor AI output often arrives looking finished. Someone still has to determine what is accurate, what is missing, and whether the conclusions make sense.
The second cost is oversight. AI does not eliminate management. People must provide context, review the output, manage exceptions, and remain accountable for the result. We all know how bad it can hallucinate.
The third cost is lost judgment. We build judgment by doing difficult things repeatedly. If AI removes too much of that productive struggle, people may create acceptable answers without developing the understanding needed to recognize a bad one.
The fourth cost is confusion about ownership. When AI assisted work is wrong, responsibility can become murky. When no one knows who owns the result, accountability weakens.
The fifth cost is misused capacity. Even when AI saves time, many companies have not decided what should happen to it. Does it create room for better service, thinking, training, or relationships? Or merely an expectation for more output? Saving time has little strategic value if no one decides how to reinvest it.
AI can help people practice, not just produce
AI can become a great training partner. When I was coaching at IMPACT, we used AI to help company sales leaders practice getting an appointment using ChatGPT. We used the audio to rehearse a difficult conversation, respond to an objection, or practice delivering feedback. We recorded the attempts, had them listen to themselves, identify the behavioral cue they missed, and try again in real time. With a clear goal, useful feedback, and repeated correction, that became deliberate practice that helped them build the behavior rather than simply understand the concept.
And, that distinction matters. AI can supply patience, repetition, and a safe place to make mistakes. It cannot supply lived experience, understand every emotional undercurrent, or fully appreciate the history and power dynamics inside a company. It can generate an example. It cannot have lived it. That is why I do not believe the strongest model is AI instead of a coach. It is human coaching supported by AI practice between conversations.
AI supplies unlimited repetition. The human coach supplies judgment, context, experience, and accountability.
Why this is a leadership issue
Another AI policy or training session may help. Neither is enough. This is an operating question. Leadership teams must define good work, where human judgment matters, and how value will be measured. Before calling an AI initiative successful, leaders should be able to answer six questions:
What business outcome improved?
Whose time was saved?
Whose workload increased?
How are we measuring accuracy, usefulness, and quality?
Can the employee explain and defend the work without the tool?
How are we reinvesting the time AI saves?
These questions change the conversation from adoption to performance.
AI enters the culture, management practices, expectations, and accountability structures that already exist. If roles are unclear, AI can make them less clear. If quality standards are inconsistent, AI can multiply that inconsistency. If people already pass unfinished thinking to one another, AI can help them do it faster.
The reverse is also true. In a healthy organization, AI can remove low value work, create room for better decisions, and help people practice the skills their performance depends on. The difference is not simply the tool. It is the clarity and leadership surrounding its use.
Measure the whole system
Leaders do not need to become AI experts. They do need to stop confusing speed with value. The right question is not, “How much faster did we produce this?” It is, “Did this improve the performance of the whole system?” That includes the person using the tool, the colleaguesreceiving the work, the customer affected by it, and the business outcome it was supposed to improve.
Before investing in the next tool, take a harder look at the ones already being used. Find out where they are improving judgment and where they are replacing it. Ask where they are removing work and where they are merely transferring it. Because efficiency without accountability is not productivity. It is work passed to someone else.
Comments