Jeff built 15 AI bots, then asked what they had changed about him

The story

Jeff built 15 AI bots, then asked what they had changed about him

I have more than fifteen AI bots and agents working for me right now. Research. Sales. My inbox. My finances. Coaching prep. I even gave them names, like Annie Inbox and Fred Finance. Building them keeps getting easier: xAI's Grok Bot lets you name a bot, give it a job and start working, with no code anywhere in the steps, and OpenAI's Dots and Meta's Muse put a standing agent inside tools people already pay for.

Not long ago I asked myself a question I did not enjoy answering. What have any of them actually changed about me?

3.6 billion people opened a Meta app on an average day in June, the company reported in July. Getting software into human hands has never been easier. And yet Gallup's newest study of 15,482 American employees, published in September, found that 54 percent have not used AI in their job at all.

Look at the last two numbers together. Among the people who do use these tools, speed is the benefit they report most and better work is reported far less often. So the tools are reaching people, and mostly they make the same work happen faster.

My sons have a word for someone who is strong. Swoll. They will also tell you that owning barbells does not get you there, and plenty of garages hold a squat rack draped in laundry.

Back to my question. Honestly, not much had changed. I had built the gym and skipped the workouts.

My wife Mindy, who teaches leaders on stage, once said something I have never forgotten: "There is a difference between the transfer of information and transformation." That same gap sits between owning these tools, even building your own, and being changed by using them.

As Jason Malec and I work with leaders, we keep seeing the same pattern. The tools start paying off once a practice forms around them. That means a new habit about what you do, and, harder, deciding what you no longer need to do yourself so the tool can do it. Most of us keep doing work out of habit long after something else could handle it.

More this week

Five leaders putting AI to good use

Each of these five people made one specific decision about how AI should be used.

Faith communities had little say in how social media was designed. These five are getting involved while AI is still being designed.

KPMG asked 48,000 people whether they check AI's work

KPMG and the University of Melbourne surveyed more than 48,000 people across 47 countries. On AI at work, they report that two thirds rely on what AI produces without evaluating whether it is accurate, and that 56 percent are making mistakes in their work because of it.

The missing step is at the end, after the draft comes back. A pilot flies most of a trip on autopilot and still takes off and lands the plane. We call this the 10/80/10 rule.

The first ten percent is yours. The middle eighty you can give away. The last ten comes back to you: read it, check it, correct it, or start over. A pilot cannot blame the autopilot for a bad landing, and you cannot blame AI for a mistake in work you send. Before you open the tool, write down the goal, who it is for, and what a good result looks like. When the draft comes back, check it yourself before it goes anywhere.

The practice

Run one 45-minute session with your team

Think about how much of your week, and your team's week, goes to coordination: email, scheduling, follow-ups, and meetings to plan other meetings.

A chief of staff is the person who handles that work for a leader. They keep track of everything so you can spend your attention on what matters most. They brief you before you walk in, track what you promised, draft the hard message, and ask the question that tests your assumption before you decide. The median US chief of staff is paid about $210,000 a year. Most leaders will never have one.

An AI assistant set up to do that job can take on much of the coordination work. It prepares decisions and never makes them. Any decision that is legal, financial, or affects someone's job belongs to a named person who answers for it.

Run this as a team session rather than sending a memo. Microsoft surveyed 20,000 knowledge workers who use AI and found that organizational factors — culture, manager support, talent practices — accounted for 67 percent of the AI impact people reported, against 32 percent for individual mindset and behavior. Microsoft says plainly that this is an association rather than a proven cause. Even with that caveat, the finding suggests that what you set up as a leader matters more than how motivated your team feels.

If you have not used an AI assistant this way yourself, use one for a week before you run the session.

After the session, expect three different responses. The person who resists usually has a real concern, such as their job or the quality of their work, so ask which single task they would hand off if they trusted the tool, and do that one task with them. The person who is quietly using AI already may be using it badly, so hold a 30-minute session each month where people show how they use it. The person who is redesigning whole workflows is doing the work your plans for AI depend on, so find them, give them a budget, and back them publicly when an experiment fails.

Add one more item to each person's context page: the two or three things they most want their work to accomplish this year. Then, every Friday, ask the assistant to compare that week's calendar with that list and answer two questions. How much of my week went to the things I said matter most? What did I agree to that I should have declined?

Higher ground

James is writing about scripture, not software, but he describes the problem Jeff found with his fifteen bots. When we hear good advice and agree with it, it is easy to feel that we have already acted on it. James calls that feeling self-deception.

An AI assistant can show you where your week went. It cannot tell you whether that was the right place for it to go. That takes prayer, and honest conversation with people who know you well. Most of us drift from what matters slowly, by agreeing to one reasonable request after another, which is why the Friday review is worth ten minutes every week.

The next step

Base Camp is where we build these practices alongside other leaders until the habits hold: one ninety-minute session a month with founders and executives, plus a written briefing every week.

See what Base Camp is →

Then hit reply and tell Jeff what you handed off this week, and what you did with the time. We read every one.

Climbing with you,
Jeff and Jason