You mostly do not have to. The team is already adopting it on their own, ahead of any program you would run. In one survey, 87 percent of employees said they use AI voluntarily even though only 28 percent of companies require it, a pull running roughly three to one ahead of the mandate.

That changes what the founder is actually solving for. The adoption playbook most businesses inherit assumes resistance: a team that will not move unless pushed, so the work is the mandate, the training week, the checkbox in the review. But the data describes the opposite situation. The people are out in front, experimenting on their own initiative, and the systems around them are what lag. The bottleneck is not willingness. It is that all that voluntary use happens on personal logins, in private chat histories, in ways nobody coordinated and nobody else can see or reuse.

About half of employed US adults now use AI at least a few times a year, yet only around one in ten strongly agree it has transformed how work gets done across the organization.
Gallup, 2026

So the real question is not how to create willingness that already exists. It is how to keep the willingness from evaporating into scattered personal habits. Gallup found that while roughly half of employed US adults now use AI at least a few times a year, only about one in ten strongly agree it has transformed how work gets done org-wide. Individuals pull. The shared operation lags. That gap, not a motivation problem, is what the founder is there to close.

You make it something they reach for by removing a friction they already hate, not by selling them on a tool. Pull not push. Find the task that is high effort and low satisfaction, the one everyone redoes and nobody enjoys, and fix that one first. A tool that takes away a real, felt friction gets passed from one person to the next without a campaign.

The pull is not slowing down, either, which means there is plenty of it to point somewhere useful. Pew found about 21 percent of US workers now do at least some of their work with AI, up from 16 percent in a single year, with the growth driven by individual uptake rather than company rollouts. People keep finding their own reasons to use it. The founder who tries to be the reason is competing with a force that is already stronger than any internal memo.

The move that works is to give the pull a destination. When one person finds an AI setup that genuinely saves them an afternoon, the difference between that staying a private trick and becoming how the team works is whether there is a shared place for it to live. Point the pull at the operation, and adoption spreads sideways on its own.

Pull beats push because pushing is both unnecessary and counterproductive, and the evidence for the second half is getting loud. The mandate produces compliance, not adoption, and it has a habit of being publicly reversed. A language-learning company that announced it would evaluate employees on their AI usage walked it back within the year, with the CEO saying plainly “I’m not going to force you”.

A year after announcing AI use would be assessed in performance reviews, the CEO backtracked: “I’m not going to force you.”
Fortune, 2026

The reason the walk-back keeps happening is that the mandate is solving a problem the business does not have. You do not need to force a behavior the team is already volunteering. What forcing actually does is replace a genuine, useful pull with a defensive one, where people use AI to be seen using it rather than because it removed something they hated. That is motion without adoption, and it leaves the founder owning the whole effort.

The founder-led version of this is sharper than the enterprise version, because in a smaller business the founder is right there, close enough to see exactly which friction is worst and to fix it fast. The job is not to be the chief evangelist. It is to be the one who notices the high-effort, low-satisfaction task, removes it for one person, and makes sure the fix is shareable. The pull does the spreading.

This is the difference between scattered AI and a shared operation. Everyone running their own version of a chat tool is sprawl that looks like adoption and leaves the founder exactly as central as before. What turns the existing pull into something the business keeps is a place where one person’s working setup becomes the team’s and gets stronger the more it is used. That shared rail is what a tool that fits keeps getting reached for, and it is the reason adoption compounds instead of resetting. JynAI built Works, an AI Business OS, to be that rail: not another tool to push onto the team, but the place the pull they already have finally has somewhere to go.

Here is where the show beats the tell. The pain is that one person’s good AI setup dies in a private chat history and never becomes the team’s. The Works mechanism that relieves it is Learns Your Business, where context, voice, customers, and the sequences that worked accumulate in a shared workspace, so a six-month-old setup runs on today’s Works without a rebuild. The gain is adoption that spreads sideways without you driving it. The affordability that makes this real for a founder-led business is the $49 Pro tier, capability at a price that does not require a budget meeting. And it is not theory for us: deployed across six teams, with revenue per employee landing two to three times higher, the fuller account of how that played out is its own story. The pull existed there too. It only mattered once it had a shared place to land.

Get team-led AI adoption. Get early access. Or grab the Bottleneck teardown first to find the one friction worth fixing on your team.

Which high-effort, low-satisfaction tasks should I fix first to make adoption spread on its own?

Start with the task everyone redoes and nobody enjoys: the weekly report nobody trusts, the follow-up sequence that always slips, the research that eats an afternoon before a call. Pick the one with the worst effort-to-satisfaction ratio, because relieving it produces the most obvious before-and-after. When the person who used it tells the next person without being asked, you have found the right one. See the failure mode when no pull exists for the inverse case.

Why does a tool that solves a real friction spread on its own?

Peer-to-peer word-of-mouth is the mechanism, and it outperforms any mandate because the messenger has personal evidence. When one person’s afternoon got saved, they have a concrete before-and-after to show a colleague, and that carries far more trust than a rollout email. The Pew data shows the same pattern at scale: individual uptake grew from 16 to 21 percent of US workers in a single year, driven by people finding their own reasons, not by company programs.

Should I require my team to use AI?

No. Requiring it solves a problem you do not have, since voluntary use already outruns mandated use by roughly three to one, and the mandate tends to get publicly reversed. Requiring it also swaps a genuine pull for a defensive one, where people use AI to be seen complying. Remove friction instead, and let the pull that already exists do the work.

How do I get bottom-up AI adoption instead of top-down adoption?

Bottom-up adoption starts with the first person whose friction you remove, not with a program. Fix the worst task for that person, make sure the fix lives in a shared system rather than their private chat history, and then watch whether they pass it on without being asked. If they do, it is a signal the fix is real. The Gallup data adds a precise lever: employees who strongly agree their manager actively supports their AI use are far more likely to use it weekly, yet only 28 percent report having that support. Being visibly behind the fix is the support the team needs.

What is the failure mode when there is no pull or the pull has no home?

The polite silence: the team nods at the rollout and then quietly reverts to the old way, because the tool was never built for how they work together. Each person is privately deciding what is safe, the know-how stays siloed, and the founder ends up exactly as central as before. The full reading of that failure mode is in the polite silence.

1. Entrepreneur, Your Employees Are Using AI Whether You Know It or Not, 2026
2. Gallup, Rising AI Adoption Spurs Workforce Changes, 2026
3. Pew Research Center, About 1 in 5 U.S. workers now use AI in their job, up since last year, 2025
4. Fortune, Duolingo CEO backs off from evaluating employees on their AI usage, 2026
   
   

Keep reading:

• The polite silence, the failure mode when there is no pull
• Why a tool that fits keeps getting reached for
• Team Runs With Me