OneDigital tells a story in Workforce Intelligence that I now repeat to almost every owner I work with. The firm launched one of its AI coworkers to a group of employees. The knowledge base was solid and the outputs were accurate, yet adoption stalled at 35% of the people who could use it.
The team’s first instinct was to extend the pilot and rework the tool. The fix turned out to be something else. They replaced the single launch event with a facilitated session every two weeks, run by the person who owned the coworker. Within six weeks, adoption reached 72%, and nobody had touched the software.
“The Coworker had been ready, while the adoption infrastructure had not.”
Workforce Intelligence, Chapter 8
For a small business, the adoption infrastructure is mostly one person: whoever runs the team. This post explains why, and gives you a 90-minute session you can run this month instead of a corporate training program.
What the Research Says About Training
Robin Singhvi, who runs OneDigital’s AI Coworker program, sums up the book’s approach to training in one line:
“AI adoption happens when you help people do today’s work better and with more confidence, rather than when you ask them to believe in the future.”
Robin Singhvi, quoted in Workforce Intelligence, Chapter 8
The data backs him up. SHRM’s 2025 research, which surveyed 1,812 US workers, found that 97% of employees who rated their organization’s integration of AI and people as excellent were satisfied with their AI training. Where integration was rated fair, satisfaction fell to 21%. Where it was poor, 18%.
In most of those organizations the training content was similar. What differed was everything around it: whether managers used the tools, whether the work itself changed, and whether anyone followed up after the session ended.
Small businesses face a second problem. The US Chamber of Commerce Foundation’s Main Street AI Monitor found that 19% of small business workers say AI adoption at their organization was driven mostly by employees exploring tools on their own. Only 11% said it came from organizational guidance. About one in ten had been offered formal AI training at all.
Put those findings together and you get a familiar picture. Some of your team are already using AI, and they taught themselves. Others are avoiding it and hoping nobody asks. What turns either group into a capability, rather than a risk, is what their manager does next.
Why the Manager Matters Most
Gallup has found that 70% of the variance in team engagement comes down to the manager. AI doesn’t change that. It makes it easier to see, because using AI is a visible behavior.
When a team watches their manager work through a real problem with AI, including the prompts that didn’t work, they learn that this is now part of the job. When a manager talks about AI in meetings but visibly keeps working the old way, the team learns that too, and no memo will override it.
In a small business there is usually no layer between you and your team, so the manager in question is often you.
The Six Manager Behaviors
The book’s half-day workshop for people managers covers six behaviors. All of them are free and all of them can be seen by the team.
| Behavior | What it looks like in a small business |
|---|---|
| 1. Use AI daily, where people can see | Work with it in normal meetings and working sessions, not only on your own |
| 2. Share what you are learning | Show the prompt that failed and what you changed, as well as the wins |
| 3. Make AI fluency a personal goal | Name it as something you are working on, and say how it is going |
| 4. Run the demo yourself | Don’t delegate the demonstration. The team reads priorities from who presents |
| 5. Protect learning time | Put it in the calendar, and react calmly when an experiment goes wrong |
| 6. Bring AI into real client work | Use it on live work, where the results actually matter |
The second behavior is the one most managers skip. Sharing a failed prompt teaches the actual skill, and it tells everyone else that being imperfect at something new is acceptable here. The sixth matters most, according to the book, because using AI on real client work is what convinces people it is worth the effort.
The Barrier Is Belief, Not Skill
The book’s authors noticed something odd inside their own firm. Many of the same people using AI for quick, one-off tasks at work were using it far more deeply at home, as a fitness coach, a parenting sounding board or a planning partner, with long conversations that built on earlier ones. They already had the skill. At work, they asked it to summarize a document.
The book calls the reason “authenticity concern”: the worry that if AI helped with the thinking, the work is no longer fully yours. That worry is reasonable for anyone who has spent decades building their expertise, and telling people to get over it won’t work. The book’s answer is that the person directing the work still owns the thinking:
“The AI can push and pull on ideas with you tirelessly until you have a refined conclusion or deliverable. And by that logic, the thinking is still yours.”
Workforce Intelligence, Chapter 4
Say this out loud to your team. Left unspoken, authenticity concern limits adoption to the handful of people who were going to experiment anyway.
What to Keep From the Corporate Model
OneDigital uses a five-tier fluency model: the AI user, the manager, the coworker supervisor, the AI fellow and the builder. Every employee completes a 90-minute hands-on workshop before getting access, and every people manager completes a half-day session.
A small business needs the first three tiers and can ignore the last two. Your enthusiastic early adopter is already doing the fellow’s job without a title, and your software vendors are your builders.
In an owner-led business, the supervisor tier is usually you, and the manager tier may be one other person or nobody. A training cascade doesn’t work with fifteen people. What works is a short, regular session run by the owner on real work.
The 90-Minute Session
This combines the book’s workshop and manager session into one sitting.
Before you start (15 minutes). Pick a hard, unresolved problem from this month’s work. Choose something real that is still bothering you, not a demonstration example.
Opening (5 minutes). Explain what you are doing in plain terms. For example: “I’m going to work on this problem in front of you, because I’m not good at this yet and I want us to get good at it together.”
The work (50 minutes). Work the problem with AI on a shared screen, using real prompts. When a prompt doesn’t work, say so and explain what you are changing. The iteration is the skill you are teaching.
Handover (20 minutes). Ask each person to bring one of their own real problems to the next session. Book the next session before anyone leaves the room.
Repeat it monthly. As the 35% to 72% story shows, a regular rhythm matters more than the content of any single session.
Your Next Steps
This week, send your team a short, honest message. Something like: “Some of you are already using AI and some of you aren’t sure where to start. We’re going to work on this together, and I’d rather we learn it in the open than quietly on our own.”
Then book your first 90-minute session for this month, with your own hardest problem as the agenda.
If you are still deciding where AI should go first, the AI Opportunity Scorecard takes two minutes and tells you whether a piece of work is the right place to start. If you need ideas for your hardest problem, start with the Redesign column of the 2027 AI Transformation Worksheet.
If you’d like help planning the session or coaching your managers through it, the Focus HR consulting team works with Arizona small businesses on exactly this.
Talk to the Focus HR teamSee all 8 posts in The 2027 AI Action Plan for Small Business
About the author: Clint Parry, MBA, SHRM-SCP is a Senior Business Consultant at Focus HR, where he helps Arizona small businesses with HR strategy, compliance and workforce planning. Focus HR is powered by OneDigital. This post is part of The 2027 AI Action Plan for Small Business series.


