Journal

The Dashboard Is Green

It is Monday morning at a mid-sized distributor, and the operations review is going well. The screen at the front of the room is green from corner to corner. A year ago, leadership put an AI system in place to draft replies to customer emails. Before that, a customer who wrote in with a problem waited about nine hours for an answer. Now the replies arrive already drafted, waiting for someone to approve them, and the wait is down to forty minutes. Response time is green. Volume per person is green. Customer satisfaction held steady, so that tile is green too. The operations manager walks through each one, the room is pleased, and they have every reason to be. Somebody worked hard to move those numbers.

At the end of the table sits the senior service rep. She has been with the company fourteen years, and for most of those years, when a customer called angry, she was the one who called them back. She did it herself because she was good at it, and because the angry calls were where she learned things. One week last year, customers who had never spoken to each other kept telling her the same thing about one of the rechargeable units the company sells: it got hot to the touch while it charged. She asked each of them which unit they had, how long it had been plugged in, and where it was sitting when it happened. The answers pointed at one production lot. She walked the list to the quality manager, and the lot was pulled before anyone was burned. Nobody asked her to do that. It was what happened when a person with fourteen years of listening behind her spent her afternoons on the phone.

Her job now is approving drafted replies, about 140 a day. The customer's message sits above each draft, but the reply is already written, so she gives that message less attention than she once gave a caller. She fixes a word here and there and sends it on. She is faster than she has ever been. What stopped was the call: the voice, the follow-up questions, the reason to stop and ask which unit and where it was sitting. This year she has flagged nothing, and there was plenty to flag.

She is not one person. She is a composite of something playing out in companies of every size, and there is a fair chance you have her on your team right now. Nobody in that room made a bad decision. They made a lot of good small ones, and every one of them showed up on the screen as a win.

This is the third entry in a series on leading with AI. The first entry, The Decision You Cannot Hand Off, was about the leader alone with a finished answer before it goes out the door. The second, Who Signs for It?, was about the person that answer landed on, who asked why and found no owner. This one is about the system around both of them: what happens to a whole team when it starts running on speed.

The phone call was inefficient, and everyone at that table knew it. It took twenty minutes when an email took two, it could not be scheduled around, and half the time it opened with someone yelling. It was also the only place in the entire process where a person with judgment heard the customer in their own voice. She heard the tone. She heard which complaints came with a tired sigh and which came with real anger. She heard the thing a customer mentioned at the very end, almost in passing, that would never have made it into an email. When the company took the call out, it took out the delay, which was the whole point. The judgment that lived inside the delay went out with it, and the loss never got written down, because the judgment had never been written down in the first place. It was simply part of how she did her job.

The research is starting to show what she lives every day. In an eight-month study published in Harvard Business Review in February 2026, researchers followed how generative AI changed work habits at a U.S.-based technology company of about 200 employees. People worked at a faster pace, took on a broader scope of tasks, and stretched work into more hours of the day, often without anyone asking them to. The researchers warn that leaders may not see what the gains are costing them until it is too late.

Read that next to the service rep. Her day is fuller than it has ever been. She touches more customers before lunch than she used to reach in a week, and by every measure on the screen she is doing more. The part only fourteen years could give, the listening, has gotten smaller. The work got faster while her part in it got thinner, and those are two different measurements. The dashboard only takes one of them.

You cannot point to the decision that did this. Somebody approved the drafting tool, and that was reasonable. Somebody set a response-time target, and that was reasonable too. Somebody moved the callbacks to email so the team could hit the target, and somebody else noticed the senior rep was the fastest approver in the department and gave her the biggest queue. Each choice made sense on the day it was made, and each one arrived with a number to back it up.

I wrote about this in RemAIning Human, in a chapter I called "The Efficiency Trap." Here is how I put it in the book: "Each individual change appears reasonable. Each one can be justified on its own." The damage comes from how the changes stack, and once they do, "People optimize for the metrics. The deeper purpose of the work becomes harder to see." The service rep did exactly what the system asked of her. She optimized for the metric she was given and got very good at it, and the reason the company hired her slipped out of view without anyone ever deciding it should.

Leaders have kept score on the wrong thing before. In Scoreboard Leadership I wrote about Wells Fargo's "Eight is Great" sales target, where the number became the mission and employees who raised a warning through the company's own ethics line were fired. With AI, the scoreboard can shift underneath everyone at once, and it never feels like anyone chose it.

I like efficiency. I have spent more than thirty years building companies, and I have chased it the entire time. It gets dangerous when it becomes the only thing a leader can see. I learned that keeping a fast pace safe for people took work that never showed up on any screen.

At MWI we operated at a very high pace, all the time. The work was always in front of us, and there was always more of it coming. Speed was never the question for us. How our people were doing under it was.

What made that pace possible was the part that stayed out of sight. We had to be very aware of how our people were doing, and I mean how they were doing as humans, beyond where their projects stood. Were they under too much pressure? Were they able to perform in a way where they felt like they were having success? Were they coming in every day rested and ready to go to work, or were they becoming overwhelmed by the pace and the volume of work that was always in front of all of us? None of those questions had a tile on a dashboard. Somebody had to keep asking them, and pay attention to what came back.

We also worked hard on team. One phrase we used constantly in those years was "nobody is on an island." There were always multiple voices connected to a process or a project, and we made sure our people stayed open to working together on it.

That was the real work behind the speed: making sure the systems, the environment and the culture all lined up, so we could move fast and not beat people up. The speed was safe because leaders did that work on purpose.

AI takes part of that work out without anyone deciding to. In a company like that distributor, a lot of the human work used to happen whether anyone planned for it or not, because the slow steps forced it. The phone call made someone listen. The handoff between two people put a second set of eyes on the problem. AI removes that friction, which is exactly why companies buy it, and when the friction goes, the human work it used to force goes right along with it. The only way that work survives is a leader putting it back deliberately.

The national numbers point the same direction. Gallup reported in July that, as AI use climbed during the first half of 2026, slightly fewer U.S. employees felt their opinions count, and slightly fewer felt connected to their organization's purpose. Hold those two findings up next to the service rep. Her opinion used to count for something specific: last year, it kept a defect from becoming a recall. She knew exactly why she was there. Now she approves drafts, and both of those things have quietly gotten harder for her to feel.

The more useful finding sits inside the same study. Workers who have a clear AI plan, use AI frequently, get active support from their manager and are themselves engaged are nearly three times as likely to give AI the highest productivity rating as frequent users overall, 50 percent against 17. The tool is the same across all of those workers, and the highest ratings come far more often where a manager is actively involved with the person using it. The best results belong to the places where a leader stayed in the room.

Most companies will never read a Gallup report. Few companies of forty or eighty people run engagement surveys, and the numbers would be too small to tell them much if they did. At that size the cost arrives a different way, as a problem that went uncaught.

Now give the distributor six months. Somewhere in that stretch, the lid on one of its pressure cookers starts coming loose while the pot is still under pressure. She would have heard it on the phone. The defect keeps shipping that whole time. Customers still write in about it, and each one gets a polite, well-written reply with an apology and a replacement, so the case closes and the tile stays green. Every individual email is handled well. The pattern running across them is never acted on, because the one person closest to it was busy reading drafts. Months later a customer is burned and a claim letter arrives. Someone pulls the reply history and finds the same complaint, answered politely dozens of times, every case closed inside the hour, and none of it ever gets tied back to the day the phone calls stopped. A green number with nobody behind it is a guess dressed up as a fact.

So what does a leader do on Monday? The same kind of work we did at MWI, now on purpose, because the machine no longer forces it.

First, every time a step comes out of a process, the person who owned that step answers one question for the leader before anyone reports the gain: "What did that step let you notice?" Sometimes the honest answer is nothing, and the step can go with a clear conscience. Sometimes the answer is the most important thing that person did all week, and you only learn that by asking before the step is gone.

Second, the leader keeps the human part that carried judgment, even when it costs speed, and says out loud why. At the distributor, that means the senior rep calls back the five angriest customers every week. Response time on those five goes up, and the leader says so in the Monday meeting and explains the reason. When a leader names a trade like that in front of the team, people learn what the company actually values, and they learn it faster than any memo could teach them.

Third, in one-on-ones, the leader asks, "What part of your work don't you recognize anymore?" It sits alongside the questions we asked at MWI about pressure, success and rest, pointed at a new problem. The answer is information the dashboard cannot produce, and a leader should treat it as seriously as anything on the screen.

All of this is slower. It hands back some of the gain, and some leaders will hear that and decide it is not worth the time. I would ask them to price those six months first. Speed that nobody is watching tends to send the bill later, and it usually arrives bigger than the time you saved.

The first entry in this series asked what the output does not know about this person. The second asked whose name is on it. This one asks something just as plain: what did we stop hearing?

Go back to that Monday. Same room, same green screen. The operations manager finishes the last tile, and before he moves to the next agenda item, he turns to the end of the table and asks the senior rep what she is not hearing anymore. She hesitates, because it has been a year since anyone asked her anything like that. Then she tells him. Three customers in the last month mentioned the lid on one of the pressure cookers coming loose while they cooked, each one in passing, each one in an email she approved and sent. She had noticed it the way you notice something at the edge of your vision, and she had nowhere to put it. The meeting runs ten minutes long. Nobody minds.

Nothing on the screen changed that morning. The tiles were just as green when the meeting ended as when it started. What changed was that the leader treated the person at the end of the table as a source of information again, instead of as a line on a throughput report, and she walked out knowing her ears still mattered to somebody.

Speed is easy to measure, and what a person notices is the reason you hired her. Those two things can live in the same company, and the best companies I have seen keep both. A machine can make the work faster. Only a leader can decide what the work is for.

When did your team last get faster at something, and what did you stop hearing because of it? If you are working through that in your own company, I would like to hear about it. Let's have a conversation. It's free.