Why Your Meeting Notes Aren't Building Institutional Knowledge

Thirty to forty percent of a consultant's billable capacity goes to meeting-related work – pre-meeting prep, live capture, and post-meeting documentation. That's the finding from floral.so's 2026 analysis of AI meeting intelligence for consulting practices, drawing on McKinsey's baseline research on how knowledge workers spend their time.

Most practitioners who see that number reach for the same answer: a transcription tool. Granola, Otter.ai, something that records and summarizes. Good tools. They solve real problems.

The transcript isn't the flywheel, though.

There's a distinction worth understanding here – between capture and institutional knowledge – because they require fundamentally different things. Meeting intelligence tools solve the first problem well. The second requires an architecture those tools don't provide.

What meeting tools actually solve

Granola and Otter.ai are capture tools. They record, transcribe, and summarize. The best ones generate action items and organize by meeting type. They eliminate the burden of real-time note-taking and give you a searchable record of what was said.

That's genuinely useful. The 30–40% capacity burden is partly a capture problem: consultants spending time writing up meetings they just attended, re-reading notes before the next one, reconstructing context from memory because nothing was recorded properly. Capture tools address all of this.

But once you've solved capture, a different problem becomes visible. You now have hundreds of transcripts that aren't talking to each other.

What institutional knowledge actually requires

Elephas.app's 2026 guide on knowledge bases for consultants quantifies what happens without a synthesis layer: independent research by knowledge management consultants puts the average knowledge worker at four to six hours per week re-researching things they've already found. For a solo consultant billing at $150–$300 per hour, that's $31,000 to $93,000 in lost productivity annually — before accounting for proposals that don't reference your best past work.

They call it the knowledge recreation tax.

Capture tools don't eliminate that tax. A transcript is a record of what happened. Institutional knowledge is a resource for what's next. Those are different things, and they require different infrastructure.

A transcript from Tuesday's meeting with Client A sits in your Granola folder alongside sixty others. When a similar situation arises with Client B three months later, you won't remember that the Client A conversation has anything useful. You might not even think to search for it. And if you do, "AI strategy roadblock" returns fourteen results and you're doing the work the tool was supposed to prevent.

The flywheel requires curation – something that reads the transcript, identifies what's reusable, tags it with the concepts that connect it to other engagements, and makes it retrievable when a relevant situation recurs. That's not transcription. That's synthesis.

The synthesis layer: where the flywheel lives

Firmem's 2026 benchmark data on boutique consulting firm capacity identifies the friction point precisely: at five or more concurrent engagements, context reconstruction at the start of each client interaction consumes 30 to 45 minutes. A searchable institutional knowledge layer removes three to four hours per week of that friction. Firms with strong post-engagement documentation and searchable past-work libraries scale 30 to 50 percent further than those without one.

That infrastructure – the thing that makes captured meetings useful across time and across engagements – is what the knowledge flywheel actually describes.

Floral.so puts it directly: "The firms that will lead in the next decade are not the ones with the most consultants. They're the ones whose knowledge base grows with every conversation." (AI Meeting Intelligence for Consultants, floral.so, 2026-01-27)

A knowledge base grows when someone curates it. Transcripts accumulate.

What the synthesis layer actually requires

The synthesis layer has three components that transcription tools don't provide. Pattern extraction – not just what was said, but what the meeting reveals about how this type of engagement tends to unfold, and what's reusable the next time a similar situation arises. Cross-engagement retrieval – connecting Client A's January conversation to Client B's current situation through something more precise than keyword search. And context persistence – arriving at a meeting already oriented, with relevant threads pulled from across the engagement history rather than three transcripts to re-read from scratch.

Most practitioners attempt some version of this themselves. They take extra notes. They maintain a folder structure. They review transcripts before important meetings. The problem isn't intent – it's that curation is slow, inconsistent across engagements, and compounds badly at scale. The consultant who reviews transcripts carefully at two concurrent clients is pulling highlights at five.

The architecture question

The useful question for most practices in 2026 isn't "are you using meeting intelligence tools?" The answer is probably yes, or you're thinking about it.

The useful question is what happens after the meeting ends and the transcript files.

Capture is largely a solved problem. The tools are good, inexpensive, and not meaningfully differentiated from each other for most practitioners. Granola and Otter.ai both do what they say they'll do.

The synthesis layer – curation, pattern extraction, cross-engagement retrieval – isn't a tool category yet. It's either something you've built for yourself, something you do inconsistently, or something that isn't happening at all.

The firms that compound knowledge across engagements aren't the ones with the best transcription tool. They're the ones that have resolved the architecture question: who holds the synthesis layer, and how does it feed into the work?

If you're running a practice where meetings are captured but knowledge isn't compounding – where each new engagement still starts largely from scratch – that's the gap worth closing.

It's not a note-taking problem. It's an intelligence architecture problem.

If you want to see what the synthesis layer looks like in a boutique advisory practice, email matthew@fieldway.org.

---

Related: Your Pipeline Is Healthy. Your Calendar Is Lying to You. | The 70% That Gets You to Your Best Work | Why Your CRM's AI Doesn't Know When a Client Is at Risk