The short version: AI meeting note tools like Otter, Fireflies and Zoom's built-in AI companion are brilliant at capturing what was said, and they are quietly making small teams worse at deciding what happens next. When everyone assumes "the AI has it", nobody owns the action, and things stop getting done. The fix isn't ditching the tool, it's adding one human step back in that AI can't do for you.
What's happening in small business meetings right now
I've watched this shift happen over about eighteen months. Two years ago, small teams took their own notes, badly, and half the actions still got missed. Now most of the small businesses I work with have some flavour of AI note-taker running in every call. Zoom's AI Companion, Otter.ai, Fireflies, Fathom, Microsoft Copilot in Teams. It joins silently, transcribes everything, spits out a tidy summary with bullet points and "action items" thirty seconds after you hang up.
On paper this is a gift. Nobody has to type while they're supposed to be listening. You get a searchable record. You can send the summary to someone who missed the call. All true.
What nobody tells you is what it does to how people behave during the meeting itself.
The meeting where nobody wrote anything down, including the AI's owner
Here's a real one. I was on a strategy call last year with a client, a seven-person marketing agency in the Midlands, going through their quarterly plan. Fireflies was running in the background as it always did for their calls. We agreed, clearly, out loud, that their ops manager would rewrite three client proposals using a new pricing structure by the following Friday.
Three weeks later, in the next monthly call, we're going through the same client proposals and they haven't moved. Nobody had touched them. When I asked what happened, the answer was almost funny in how honest it was: "I thought it was in the notes somewhere, I assumed someone would follow up." The founder thought the ops manager had it. The ops manager thought the founder would chase it once the summary landed in her inbox. The AI summary did, in fact, list the action item correctly, word for word, with a name attached. It sat in an email nobody reopened.
That's the bit the tool can't fix. It captured the decision perfectly. It did nothing to make anyone feel responsible for it, because a document doesn't feel like a person asking you a question.
Why this keeps happening (and it's not a tech problem)
Psychologists call this diffusion of responsibility, the same reason a crowd is slower to help someone in trouble than one person alone is. Add a "smart" record-keeper to a meeting and you get a version of it: when a group believes a task has been captured somewhere, individual members feel less personally on the hook for making sure it happens. The more official and polished the record looks, the stronger the effect, because a beautifully formatted AI summary looks like accountability even when it's created none.
Small teams get hit harder by this than big companies do, and that's the uncomfortable part most posts about AI note-taking skip over. A 200-person company has project managers, ticketing systems, someone whose entire job is chasing status updates. A five-person business has the founder, a part-time assistant, and now an AI summary standing in for all of that infrastructure. The tool is doing the easy 20% of the job, transcribing and summarising, and everyone is behaving as if it's doing the hard 80%, which is making sure things get done.
I'd put money on this: if you pulled the AI-generated action items from the last ten meetings across a typical small business and checked in three weeks later, at least a third would be untouched, sitting exactly where the AI left them, in a summary nobody reopened.
The fix: one human step the software will never do for you
This isn't an argument for dropping AI note-takers. They're faster and more accurate than a tired human trying to type and think at the same time. The fix is a five-minute ritual bolted onto the end of every meeting, before anyone closes the laptop.
- Step 1: End every meeting with sixty seconds of out-loud recap. Not "the AI will send it round", an actual person says, "So to confirm, three actions: Sarah does the pricing rewrite by Friday, I chase the supplier, Tom books the photographer." Saying it out loud, to a named person, creates a psychological commitment that a written summary can't.
- Step 2: One person owns the follow-up, and it's never the most senior person by default. In small teams the founder often assumes this job by habit and then gets too busy to do it. Assign it explicitly, ideally to someone whose role includes chasing, not deciding.
- Step 3: Actions move out of the AI summary within 24 hours. Copy them into whatever your team uses daily, a shared task board, a WhatsApp pinned message, a simple spreadsheet. If the action item's only home is the AI-generated document, it's already at risk of being forgotten.
- Step 4: Put a date on it, not "soon". AI summaries are decent at capturing that a task exists and weak at capturing genuine urgency, because the tone in a calm meeting rarely signals it. A human has to add the deadline back in.
- Step 5: Open the next meeting by checking the last one's actions, not the agenda. This is the single biggest change I made in my own weekly calls. First two minutes, every time: "Last time we said X, Y, Z. Where are we?" It takes ninety seconds and it's the thing that builds the habit of following through, far more than any note-taking software.
None of that is complicated. It's also the part every AI-meeting-tools listicle leaves out, because it's not a feature you can screenshot.
Which tools are worth using for this
For what it's worth, on accuracy and price, Fireflies and Otter.ai both do a solid job for teams under about 20 people, with free tiers that cover occasional use and paid plans running roughly $10 to $20 a month per user once you need unlimited transcription and integrations. Fathom has become popular specifically because it's free for individual use and syncs cleanly with Zoom. Zoom's own AI Companion and Microsoft's Copilot in Teams are worth it mainly if you're already paying for the wider Zoom or Microsoft 365 subscription, since the meeting notes come bundled rather than as an extra cost.
What none of them do well yet is flagging which action items are at risk of being dropped. They'll tell you what was said. None of them will nudge you three days later and say "this hasn't moved, and you both seemed to think the other person had it." That gap is exactly where step 5 above earns its keep.
The bigger pattern: AI is exposing weak execution, not causing it
This is worth sitting with, because it applies well beyond meetings. AI tools are very good at removing friction from the easy parts of running a business and completely indifferent to the hard parts. Content generation makes writing a blog post faster, it doesn't make your content strategy any sharper. Automated invoicing makes billing faster, it doesn't make you better at chasing late payers. Meeting notes make capturing decisions faster, they don't make anyone more likely to act on them.
Alex Hormozi talks about this constantly, in a different context, in his writing about why most businesses fail to execute even when they know exactly what to do; there's a good breakdown of his thinking in this piece on business lessons from Alex Hormozi, and the through-line is the same one I keep seeing with AI tools: knowing the right action and having it recorded somewhere is not remotely the same as doing it. Napoleon Hill wrote about this a century earlier under a different name, calling out indecision and the habit of letting good intentions sit unactioned as one of the main things separating people who build something from people who don't; it's covered well in this piece on business lessons from Napoleon Hill. AI didn't invent the gap between deciding and doing. It just made the deciding part so effortless that the doing part now stands out by comparison.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
I see the same pattern in content marketing teams. Ahrefs built a huge audience without ever chasing hype cycles, largely because their content team treats "we said we'd publish this" as a commitment with a name and date attached, not a note in a doc. Their approach is worth studying, and I've broken down how they did it in this piece on the Ahrefs marketing strategy. Small teams can copy that discipline without copying their budget.
Where this connects to the rest of your marketing, not just your meetings
The same "AI captured it so it must be handled" trap shows up outside meetings too. I see small business owners generate a batch of social captions with AI and assume the posting is basically done, when scheduling, engagement and replying to comments is the actual work. I see people build an AI chatbot and assume customer service is now "sorted" when nobody's checking what it's telling people. It's the same gap, just wearing a different outfit.
If you want AI to do more than record what happened and pull people into your business, the more useful move is pointing it at something interactive that does work while you sleep, like a calculator or quiz that qualifies leads on your site; I wrote about how brands use these in this piece on interactive calculators for engagement and conversion, and it's a good example of AI-adjacent tooling that does something a human wasn't already doing, rather than automating a step a human was doing anyway and quietly removing the accountability that came with it.
Brands that get this right tend to build entire systems around ownership, not just tools. Airbnb's early growth wasn't down to a clever piece of software, it was down to founders personally showing up to photograph listings and chase hosts, a story I go through in this piece on the Airbnb marketing strategy. Away built its brand the same deliberate way, covered in this piece on the Away marketing strategy. None of it was automated away. The tools changed, the need for a human to own the follow-through never did.
When AI meeting notes are worth it, and when they're not
They're worth it when the meeting is informational, a briefing, a training session, an update where the main value is the record itself. Nobody needs a human ritual to remember that a webinar happened; the transcript is the whole point.
They're worth watching closely when the meeting produces decisions and actions, which is most small business meetings that matter. That's exactly where the human recap, the named owner, and the "where are we" opener at the start of the next call earn their place. If your business is at the size where getting proper help setting this up would save more time than it costs, that's the kind of practical, unglamorous fix an AI implementation coach should be helping you put in place, not just recommending another tool.
The tools aren't the problem. The assumption that a good summary equals a finished job is the problem, and it's an easy one to fix once you notice it.
Free resource: The At-Risk Account Playbook Mini-Guide.
Frequently asked questions
Do AI meeting note tools reduce accountability, or is that just anecdotal?
It's a documented psychological pattern called diffusion of responsibility, well established in group behaviour research going back decades. AI note-takers amplify it because a polished, official-looking summary makes a team feel a task is handled even when nobody has claimed ownership of it. The tool captures the decision; it does nothing to create the feeling of personal responsibility that gets it done.
Which AI meeting note tool is best for a small business?
For teams under 20 people, Fireflies and Otter.ai both offer solid accuracy for around $10 to $20 a month per user on paid tiers, with usable free plans for light use. Fathom is a strong free option if you already run meetings through Zoom. If you're already paying for Zoom or Microsoft 365, their built-in AI companions are worth using since they're bundled rather than an extra cost.
How do I stop action items from AI meeting summaries getting forgotten?
Add one human step immediately after the AI summary lands: recap actions out loud with names and dates before the call ends, move them into your actual task system within 24 hours, and open the next meeting by checking on the last one's actions before anything else. The recap and the follow-up check are the two steps that change behaviour.
Is it worth using AI note-takers for every meeting?
They're most useful for informational meetings, briefings, training, updates, where the record itself is the value. For decision-making meetings, where actions and owners come out of the discussion, keep the AI summary as a reference but don't let it replace a spoken recap and a named owner before people leave the call.
Related reading: AI Meeting Note-Takers: What Nobody Tells You Before You Turn One On and The AI Notetaker Problem Nobody's Warning Small Business Owners About.
Related: user intent vs keywords what matters in modern digital marketing.
