Table of content
Intro
At Marakana, we sold three to five day classes to some of the most advanced engineering teams on the planet. Android, Java, Hadoop, Scala, whatever the technology of the day was. And we had a running joke about it. Half or more of the value we provided was not our content. It was not our instructors. It was simply booking that time and place on their calendar.
Those engineers could have learned any of it on their own. The material was out there. They just didn't, because they were busy, and things happen.
That joke turns out to be the most important idea in corporate learning right now. AI training for employees fails when it is built as single-player content, a library people are expected to work through alone. It works when it is built as a multiplayer container: a set time, a set place, and people who expect you to show up. The container is the curriculum.
What is a learning container?
A learning container is a recurring, shared commitment to learn: a fixed time, a defined place, a real goal, and people who expect you to be there. A content library is the opposite. It waits for people to find the time on their own, and most never do.
You already know the forms. A standing tech talk. A weekly demo. A hackathon. A cohort-based program. A community of practice that meets every other Thursday. What they share is not the format. It is that someone notices when you don't show up.
Why don't employees finish on-demand AI training?
Not because the content is bad. Because nobody is waiting for them.
Everyone knows how to work out. Most of us still join a gym in January and are gone by February. The people who keep going are in a running group, a CrossFit class, a masters swim program. For a decade, I ran at 6AM every Tuesday and Thursday with my neighbors. Plenty of those mornings were dark, cold and raining. I went because I had promised them I would show up.
Learning works the same way, and the people running AI enablement are watching it happen. In recent conversations with engineering enablement and L&D leaders, at companies from a few hundred people to tens of thousands, the same pattern kept coming up.
One learning lead described building a comprehensive on-demand AI library, then looking at the numbers: "Realized nobody was doing that at all." Live sessions took off instead. At a company with tens of thousands of employees, an AI infrastructure lead told us optional lunch-and-learns get around a hundred views. The only time he had seen their LMS be really effective was for mandatory training.
Neither of those is a content problem. The material was there. What they didn't have was a reason for anyone to be in the room.
What happens when you give people time but no container?
The obvious fix is to protect time. It helps. It isn't enough.
One company gives employees a protected afternoon every week to experiment with AI tools. Good instinct. But some people came back asking: "Is there a course? Should I be doing something?" Time without structure is a gym membership. It is still single-player.
Mandates fail from the other side. One enablement leader described his CEO telling the entire company it had to use AI. It caught on like wildfire, and a lot of what people built in the free-for-all that followed was, in his words, "garbage." Looking back, he would have paired the mandate with education on how and when to use it.
A container is not a calendar block, and it is not an order. It is a time, a place, a goal and other people.
What does AI training that works look like?
The programs these leaders said actually worked had three things in common. People learned together. They learned on real work. And success was measured by what they built, not what they watched.
An engineering education leader at a large software company ran early AI training where everyone who took the course then had to write real code and open a pull request, with the results reported to their directors. "We didn't really care if they'd use the material or not," he said. "We just wanted them to be able to do the thing."
Another company took one product team, gave it the tools and a short trial to rebuild how it worked, then turned that team into coaches for the teams that came next. An engineering enablement lead elsewhere said hackathons, and pairing people with teams further along in AI adoption, worked better than anything self-paced.
None of these are content strategies. They are container strategies.
How do you build AI training for employees that people actually finish?
Across the programs these leaders described, and the ones we have run for engineering teams over the years, the same moves keep showing up.
Put it on the calendar, on a rhythm. A recurring slot beats a big launch. The value is in the habit, not the kickoff.
Make real work the curriculum. People learn faster on their own code, tickets and projects than on sample exercises. One engineering manager told us he onboards new developers by having them debug real support tickets for a month instead of reading the codebase.
Measure what people build, not what they complete. Completion rates tell you who clicked through. A pull request, a shipped workflow or a solved ticket tells you who learned.
Let your experts teach. The engineer who figured out a tool last month is a better teacher for it than any course. Give them a podium: a tech talk, a demo, a pairing session.
Start with one team, then let it coach the next. A team that has already changed how it works is the most credible curriculum you have.
Use mandates for the floor, containers for the ceiling. Mandatory training can get everyone to a baseline. Only containers get people past it.
What changes when content is free?
This matters more now than it did when we were booking three-day classes. AI has made content creation close to free. Any team can generate a polished AI course in an afternoon. If your enablement strategy is still a library, you are competing on the one thing that just lost its value.
What AI cannot generate is the room. The standing tech talk where an engineer shows what broke last week. The hackathon where someone learns a tool because the team needs it by Friday. The pairing session where the person who figured it out sits next to the person who hasn't.
That room is also where an organization's most valuable knowledge finally surfaces: the undocumented, hard-won know-how people only share when someone is there to hear it. A course library never captures that. A container does, every week, for free.
The companies getting AI adoption right are not buying more content. They are building more containers, and letting their own experts fill them.
That is what a learning culture actually is. Not a library people are allowed to use. A set of containers people are expected to show up to.
Nobody on those 6AM runs ever taught me anything about running. My neighbors just expected me to be there. That was the whole program.
