Where Your Company's Knowledge Actually Lives (and Why Nobody Can Find It)
Someone on your team spent twenty minutes last week looking for a document they knew existed. Not a document they hoped existed — one they had seen, had worked on, could describe. The final version of a proposal or perhaps a signed variation to a contract. The latest copy of a spreadsheet with the real numbers, not just a draft with estimates. They just could not find it, and after twenty minutes had to give up and rebuild it from memory, or interrupt the one colleague who would know.
We are so used to this kind of struggle that it barely registers as a problem, yet it should. Multiple delays and the use of incomplete or wrong data, across every area of the business, and you are looking at one of the largest un-costed line items most enterprises carry - from the smallest to the largest.
The underlying problem is almost always misdiagnosed as "we need better software" when it is actually something more structural.
The seven places your company's knowledge actually lives
Ask where your company's knowledge is kept and most people will name one or two systems — the shared drive, maybe the CRM. The honest answer is that it is spread across at least seven places, and no two of them talk to each other.
Email inboxes. The single largest unmanaged knowledge store in most companies is email. Decisions, approvals, the reasoning behind a price, the one message that explains why the contract says what it says — all of it sitting in individual mailboxes, along with attachments, orders, invoices, contracts, and more. Email is searchable but only by the person who owns the inbox, and only if they remember the right word.
Personal laptops and local drives. The desktop, the Downloads folder, the "working copy" that never made it back to the shared location. Knowledge here is invisible to everyone except its author and vanishes entirely when the laptop is wiped or the person leaves.
Shared network drives. Nominally the official store. In practice a decade of nested folders where Final, Final_v2, and Final_USE_THIS sit side by side and nobody is certain which one a client actually received.
SaaS application silos. Each tool the business adopted to solve one team's problem now holds a slice of the company's knowledge behind its own login. The mid-sized company today runs over a hundred separate SaaS applications,[^1] and a large share of them were brought in by a team without IT's involvement.[^2] Every one is another silo with its own search box and its own walls.
Chat. Teams and Slack have quietly become where decisions are actually made. The reasoning is in a thread. The thread scrolls away. Six weeks later nobody can reconstruct why the call was made, only that it was.
People's heads. The most valuable store and the only one with a leaving date. The colleague who "just knows" how the renewal process works, which client cannot be emailed on a Friday, why that integration is held together the way it is. Roughly four in ten of the things a role depends on exist only in the head of the person doing it.[^3]
The paper and PDF graveyard. Scanned contracts, signed forms, the supplier agreement that exists only as a photograph someone took on their phone. Technically retained. Practically unsearchable.
Nobody chose this
It is worth being clear about something, because the instinct when this is laid out is to assume someone has been negligent. They have not.
No one sat down and designed this, it just grew over time. Every system arrived as a reasonable answer to a real problem. Sales needed a pipeline, so a CRM appeared; finance needed approvals, so a tool was chosen. A team needed to move faster than the shared drive allowed, so they started keeping their working files somewhere they controlled. Each decision was locally sensible. Nobody was ever given the job of owning the whole — of asking where, across all of it, a given piece of knowledge is supposed to live and how anyone else is meant to find it.
This is entropy, not incompetence. Information systems, left alone, do not stay tidy. They fragment, because every individual and every team optimises for their own immediate friction, and the cost of that fragmentation lands on someone else, later, invisibly. Recognising that it was nobody's deliberate choice matters, because it changes the fix from "find who is at fault" to "decide who now owns the system as a system" — which is a very different and much more productive question.
The Real Cost is not in Systems or Storage.
Storage is cheap and getting cheaper and this has allowed fragmentation of information to continue — the visible cost looks trivial.
The real, uncounted cost is decision latency and decision accuracy. The meeting that cannot resolve because the number everyone needs is in a system only one absent person can open. A proposal is sent without context, as finding it would take longer than the deadline allowed. The same analysis is run twice in six months by two people who never knew the other had done it. Supplier orders remain in-force, a cost burden to the company, even when equivalent customer orders have ended.
Studies have tried to put a figure on this, but they are old and the headline numbers should be treated as directional rather than precise — even so they remain damning. McKinsey's work on information and collaboration found knowledge workers spend close to a fifth of the working week simply locating internal information or tracking down the colleague who has it.[^4] Earlier industry research priced the combined cost of failing to find information, and recreating what could not be found, in the millions per year for an enterprise of a thousand knowledge workers.[^5]
There is a sharper version of the cost than wasted minutes. It is the decision made without information, because finding it would take more time than the decision-maker had. Nobody logs that, but it surfaces later as a renegotiated contract, duplicated spend, a client who received the wrong version. It is the most expensive consequence of fragmentation and the hardest to attribute, because by the time it shows up the search that never happened is long forgotten.
The challenges underneath the obvious one
Even when something is found, three problems sit underneath, and they are why "just search harder" is not an answer.
Version ambiguity. Finding the document is only half the task. The other half is knowing whether it is the one that matters — the current one, the one the client signed, the one that is still true. When the same artefact exists in five places, "found it" and "found the right one" are different achievements, and the gap between them is where errors live.
Access asymmetry. Knowledge that exists is not knowledge that is reachable. It sits behind a login the person who needs it does not have, in an inbox they cannot see, on a drive scoped to a team they are not in. The information is technically retained and practically unavailable, which from the decision-maker's seat is indistinguishable from not existing.
The volume of data nobody is using. Most of what a company stores is dark — its content and value unknown to anyone — or redundant, obsolete, or trivial. One large multi-country survey of IT decision-makers put roughly half of stored data in the "dark" category and another third as redundant, obsolete or trivial, leaving only a sliver actively understood as business-critical.[^6] The practical effect is that the signal you need is buried in noise the organisation pays to keep and nobody curates, which makes every search harder than it should be.
And running under all of it: institutional memory loss. The person who knows leaves, and a measurable fraction of how the business actually works leaves with them — not the documented part, the part that was only ever in their head. It is the one store on the list that does not degrade gracefully. It goes all at once, on a notice period.
The Shape of the Problem
Put the pieces together and the shape becomes clear. This is not a software problem with a software answer. Buying a better search tool to point at seven disconnected, half-dark, version-ambiguous stores gives you faster access to the same confusion. It is a company-wide process and systems problem: the absence of anyone owning the question of where knowledge lives, how it is found, and who is allowed to reach it — across the organisation.
AI and Knowledge Management
AI relies on knowledge - either from its training, or from accessible memory systems - in order to make sensible recommendations or summaries. The cost of poorly managed information risks increasing significantly, especially as AI systems are known to hallucinate, or simply make things up.
Interestingly though, AI can become part of the solution to company-wide memory loss when used appropriately.
Want to see where your organisation stands? Take the AI Readiness Scorecard — a short, structured assessment with a tiered result and a sensible next step for where you actually are.
Ready to talk it through? Book a 30-minute discovery call — no sales script, just a conversation about your information landscape and what to do about it.
[^1]: SaaS application count per company — verify exact figure and source at cooldown. Working source: Productiv, The State of SaaS Sprawl / IT SaaS statistics (https://productiv.com/blog/it-saas-statistics/), reporting ~106–112 apps for mid-to-large companies (2024–25). Cross-checked against SellersCommerce SaaS statistics roundup. Pick one primary, cite it, drop the aggregator.
[^2]: Shadow-IT / shadow-SaaS share — verify at cooldown. Working figure: ~42–48% of SaaS apps in use are outside IT's control (2024–25, per Productiv and roundup sources above). Confirm wording before publish; this theme is developed fully in article 4 of the series.
[^3]: "Roughly four in ten" institutional knowledge unique to the individual — Panopto Workplace Knowledge and Productivity Report (cited via reworked.co and learntowin.com). Trace to the Panopto primary report and confirm the 42% figure and its definition before publish.
[^4]: McKinsey Global Institute, The Social Economy: Unlocking Value and Productivity Through Social Technologies (2012) — https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-social-economy. Finding: interaction workers spend ~28% of the week on email and ~19–20% searching for internal information or tracking down colleagues. Note in copy already flags this as dated/directional — keep that framing.
[^5]: IDC, The High Cost of Not Finding Information (Susan Feldman, 2001) — PDF mirror: https://computhink.com/wp-content/uploads/2015/10/IDC20on20The20High20Cost20Of20Not20Finding20Information.pdf; summary at KMWorld (https://www.kmworld.com/Articles/Editorial/Features/The-high-cost-of-not-finding-information-9534.aspx). $2.5–3.5M/yr per 1,000 knowledge workers. This is a 2001 figure — the article text explicitly dates and softens it; do NOT present as current. Founder: decide at cooldown whether to keep with the caveat or cut.
[^6]: Veritas Global Databerg Report (2016, Vanson Bourne, 2,550 senior IT decision-makers across 22 countries) — https://www.veritas.com/news-releases/2016-03-15-veritas-global-databerg-report-finds-85-percent-of-stored-data. 52% dark, 33% ROT, 15% business-critical. Dated but methodologically solid and still the most-cited source; the Splunk dark-data findings are a more recent cross-check if a fresher anchor is wanted.