Last month, J said something that caught me off guard.

“You know what percentage of the team actually uses those AI tools we bought?”

I guessed 50%. Then he pulled up the backend data and told me: 15%.

Fifteen percent. I spent time researching, testing, writing documentation, even recorded demo videos. But most people watched and forgot, went back to manually taking notes in meetings, then spent two hours after the meeting organizing everything.

My first reaction was—maybe the tool isn’t good enough? Should I switch to a different one?

But later I realized, the problem was never the tool.

Your AI Coach Lives in “Another World”

This is the most common scenario I observed: companies buy AI meeting assistants, hand out accounts, run training sessions, and then… nothing.

Because there’s a wall between AI tools and the daily workflow. Nobody thinks to open it during the meeting. After the meeting, everyone rushes to handle the next task—who remembers to ask AI “help me organize the meeting highlights”?

I was the same way. I used to treat AI as an “extra step”—finish the meeting, then go find AI, paste the transcript, give it commands. Just thinking about that process felt exhausting, let alone expecting the team to do it every time.

The turning point was when I reversed this order.

Not after the meeting—bring AI in before the meeting starts.

Stuff AI Into the Meeting, Not After

I made a simple adjustment: before each meeting, I’d drop a message to AI—the meeting topic, goals, decisions needed. Then let AI prepare a framework before the meeting starts: what questions to ask, what data to confirm, what conclusions from previous related meetings were.

During the meeting, AI runs simultaneously. Not that quiet recording in the background—there’s an actual structure guiding the discussion direction.

When the meeting ends, I don’t ask AI to “summarize the key points”—that’s just the most basic usage. I use the three-layer questioning method:

Layer 1: Basic Organization. What conclusions were reached, what action items, who’s responsible for what. Most people do this, but many stop here.

Layer 2: Context Mapping. How did the discussion unfold? Who raised what points? Where were the disagreements? How were those disagreements resolved? This layer helps you see the meeting’s “process,” not just the “results.”

Layer 3: Strategic Analysis. What assumptions underlie this meeting’s decisions? What risks do those assumptions carry? If market conditions change, which decision is most likely to have problems?

When I dropped the Layer 3 analysis results to the team on Slack, J’s reaction was: “Wait, did AI analyze this itself?”

Yep, the same tool you think can only organize meeting notes.

Teams Don’t Use AI Because They’re Lazy—They’re Afraid

This took me a while to figure out.

At first, I thought teams didn’t use AI because it was too much trouble, or didn’t see the value. But actually talking to them revealed a deeper worry—“if I use AI to do this, does that mean I’m not capable enough?”

This psychological barrier is harder to solve than any technical issue.

My approach: don’t position AI as a “tool that does the work for you,” but a “coach that helps you see blind spots.” During meetings, I’d directly say: “Let me ask AI what it thinks about this,” as naturally as you’d say “let me look something up.”

When the boss uses it themselves, and uses it in a casual, unpretentious way, the team’s psychological threshold drops.

I also did one thing: once a week, I’d share an AI “unexpected insight” from a meeting. Not the “wow AI is amazing” kind of showing off, but something that actually helped us avoid a bad decision, or pointed out a contradiction none of us noticed.

Three months later, the team’s AI tool adoption went from 15% to 82%.

Not because I found a better tool, or ran more training sessions. Because AI is no longer an “extra thing”—it’s in the workflow, as natural as holding a meeting.

Data Speaks, But You Have to Give It a Chance

I later looked back at the before-and-after numbers:

MetricBefore IntegrationAfter Integration
AI Tool Adoption15%82%
Meeting Notes Organization TimeAverage 45 minutesAverage 8 minutes
Post-Meeting Action Item Miss Rate~30%Under 5%
Team Proactively Asks AI QuestionsAlmost noneOver 20 times per week

What surprised me wasn’t the efficiency improvement—that was expected. What surprised me was the change in “post-meeting action item miss rate.” Before, after every meeting, a few things would slip through the cracks, and two weeks later someone would remember. Now AI automatically tracks after the meeting, flagging action items with no progress after three days.

It’s not AI managing people—AI is helping you remember what you promised to do.

Not Every Meeting Needs AI, But the Ones That Do Can’t Go Without

My approach now is tiered.

Daily standups, quick 15-minute syncs—don’t need AI, that just adds friction. But strategy meetings, cross-department coordination, any meeting requiring decisions—AI coach must be present.

The reason is simple: human attention in meetings is limited. When you’re thinking about how to respond to someone’s point, you can’t simultaneously think about “what’s the second-order impact of this decision.” AI can.

It won’t get tired, won’t get distracted, won’t hold back反对意见 just because the boss is in the room.

Of course, AI coach has its limits. It doesn’t understand office politics, doesn’t know why that project actually failed last time (what’s on the report versus what really happened are often different), and can’t read the subtle atmosphere in the meeting room.

So it’s a coach, not a boss. It shows you more, but the final call is yours.

The other day, while organizing these numbers, I thought about something—how much time do we spend on “post-meeting organization”? If we added up the time the whole company spends on meeting notes every week, we could probably ship another product.

But most people still think meeting notes are just meeting notes, nothing to change.

Huh, I used to think the same way.

Further Reading