The Myth of the Knowledge Management System

Everyone has a knowledge system. Nobody uses it.

The graveyard of KM initiatives

Every firm we see has SharePoint. Or Confluence. Or Box. Or some custom wiki someone built five years ago. The folder structure made sense to whoever created it. Now it's a maze. There's a "Client Deliverables" folder and a "Final Deliverables" folder and a "Deliverables (Old)" folder and nobody knows which one to use.

Partners tell us the same thing: "We tried to get everyone to document after engagements. It worked for a month. Then everyone got busy."

The people who actually know where things are? They're the senior folks who've been there 10+ years. One analyst told us he gave up on search entirely. He just calls people. "It's faster to text someone than to dig through SharePoint."

Now we're hearing a new one: "We're integrating LLM's with our SharePoint." We get the appeal. Natural language search. Ask questions, get answers. Sounds great, but here's what actually happens: The LLM doesn't know your firm. It doesn't know what "Phase 2" means at your shop. It doesn't know that "turnaround" and "restructuring" are the same thing internally. It doesn't know which docs are final vs. draft, or which proposals won vs. lost.

So you ask: "Find me a healthcare turnaround proposal." And it returns 50 results. Half are drafts. A third are from 2018. Some are mislabeled entirely. You spend 20 minutes sorting through garbage and give up. If your docs are mislabeled, buried, or duplicated, the LLM just hallucinates faster.

Why it keeps failing

The pattern is always the same:

KM systems fail because they require behavior change. And behavior change doesn't happen when people are billing 70+ hours a week.

You have to know exactly what you're looking for. "Q3 healthcare payer cost reduction" returns nothing. But the doc exists. It's just called Acme_Final_v3_JM_edits.pptx and it's in a folder called "2022 Projects."

The state of the data

Even if search worked perfectly, the underlying data is a mess.

There's Acme_Proposal_v1, Acme_Proposal_v2_final, Acme_Proposal_v2_final_JM, and Acme_Proposal_ACTUALLY_FINAL. Which one won?

Nobody knows. All four are in the system, none are labeled.

Then there's the stuff that never makes it in at all. The best work, the real insights, the client-specific analyses, the models that actually drove decisions, often lives on someone's laptop or in a client-specific Teams folder that gets archived after the engagement. It never gets sanitized, never gets uploaded, never gets shared. Too much effort, too little time. So your KM system ends up being a graveyard of half-finished drafts, duplicate versions, and sanitized templates stripped of everything useful, while the good stuff stays locked in people's heads or buried in folders nobody can access.

What actually has to change

What's required isn't better discipline, it's automation that replaces discipline:

The knowledge layer has to understand your firm's ontology. Not just generic text. It has to know the difference between a proposal and a deliverable, a draft and a final, a win and a loss.

But finding the right docs is only half the problem. The real unlock is what happens next. When a consultant needs to draft a proposal or build a deck, the system should do more than surface references, it should generate work that looks and feels like it came from your firm. Same structure, same tone, same formatting. Not some generic AI output that needs to be rewritten from scratch, but something that matches your templates and mirrors how your best people write.

And the consultant shouldn't have to spend 30 minutes hunting for the right examples first. The system should abstract that away entirely, automatically pulling in the most relevant past work based on context, so generation starts from the best available foundation without anyone having to find it manually. The goal isn't "here are 20 docs, good luck." It's "here's a first draft built on what's worked before."

Takeaway

Most firms will keep buying tools that require adoption. They'll keep mandating behavior change that doesn't stick. They'll keep wondering why the new system has the same problems as the old one. The winners will be the firms where the system does the work, not the consultants.