How to stop losing information at work

The information exists. You just can't find it. Here's how to fix that without reorganising everything.

You're looking for the document from last month's client meeting. You know it exists. You wrote it, or someone on your team did. But you can't find it. You search Google Drive: nothing relevant in the first ten results. You search Slack: find a reference to it but not the document itself. You search your email: find the thread where someone shared it, but the link is to a Google Doc that's been moved. You ask on Slack: "Does anyone know where the Acme meeting notes are?" Someone responds twenty minutes later with a link. The information existed the whole time. Finding it took longer than the meeting that produced it.

This scenario happens multiple times per day in most organisations. Research shows knowledge workers spend 1.8 hours daily searching for information they already have. That's roughly 450 hours per year, per person, spent on finding rather than using. For a 20-person company, that's 9,000 hours annually, the equivalent of more than four full-time employees doing nothing but looking for things.

The information isn't lost in the sense of being deleted. It's lost in the sense of being unfindable, which has the same practical effect: you can't use what you can't find.

Why information gets lost

It's fragmented across tools. The average company uses 137 apps. Each one holds a piece of the organisation's knowledge. None of them share a search. Finding something requires knowing which tool it's in, and the more tools you use, the less likely you are to guess correctly on the first try.

Search is keyword-based. Most tool-native search matches exact keywords. If you search "meeting notes Acme" and the document is titled "Client Call Summary - Acme Corp Q3," the keyword search might miss it. You have to guess the exact words the author used, which in a multi-person organisation is an unreliable proposition.

Filing depends on humans. Manual filing (choosing folders, naming conventions, tagging) requires discipline from every person on the team. One person files meticulously. Another dumps everything in a root folder. A third uses a naming convention nobody else follows. The result is an information architecture that reflects the habits of a dozen different people rather than a coherent system.

Information decays. Documents go stale. Processes change but the documentation doesn't. Links break when files are moved. The information that was findable six months ago may no longer be accurate, which is worse than being unfindable because it misleads.

How to fix it (without reorganising everything)

The instinct when information feels lost is to reorganise: create a new folder structure, implement a naming convention, do a documentation sprint. These help temporarily and decay rapidly because they depend on ongoing human discipline that doesn't scale.

The structural fix is a layer that makes existing information findable without reorganising it.

Connect your tools to one search

Connect the tools where your information lives: Google Drive, Slack, email, Dropbox, Notion, and others. The information stays where it is. A single search spans all of them.

The "Acme meeting notes" are now findable whether they're in Google Drive, Slack, email, or Notion, from one search bar. The fragmentation problem doesn't go away (the tools are still separate) but it becomes invisible to the person searching.

Search by meaning rather than keyword

Semantic search finds information based on what it means rather than the exact words used. "The meeting notes from the Acme discussion last month" finds the document whether it's titled "Client Call Summary," "Acme Notes Q3," or "Meeting 2026-07-15." You describe what you're looking for and the system finds it.

This eliminates the naming convention problem. You don't need consistent naming across the organisation because the search doesn't depend on names. It depends on meaning.

Capture knowledge automatically

The information that's most frequently "lost" was never documented in a persistent form. It was shared verbally in a meeting, discussed in a Slack thread that scrolled away, or explained in an email that was read and archived.

Self-writing documentation captures this knowledge from the channels where it's naturally shared: meeting recordings are transcribed and summarised, Slack discussions are captured and structured, decisions are documented as they're made. The knowledge persists without anyone writing a separate document.

Let AI help find things

The AI assistant can answer questions about your organisation's accumulated information. "What did we decide about the pricing change?" isn't just a search query. It's a question that the AI can answer by synthesising across multiple sources: the meeting where it was discussed, the Slack thread where the team debated it, and the email where the final decision was communicated.

What changes

The shift is from "finding information requires knowing where it is" to "finding information requires knowing what it's about." The first depends on organisation. The second depends on search quality. Organisation is fragile (one person's mess breaks it for everyone). Search quality is robust (it works regardless of how anyone filed anything).

The 1.8 hours per day that your team currently spends searching doesn't drop to zero. Some searches are for information the organisation doesn't have, and some require human judgment about what's relevant. But the portion of search time spent looking for things that exist but can't be found, which is the majority, drops dramatically when every source is searchable from one place by meaning.

Frequently asked questions

Do we need to reorganise our existing files? No. The search layer works on top of your existing file structure, however messy it is. The semantic search finds information by content and meaning rather than by folder location or filename. Your messy Google Drive becomes searchable as-is.

What about information in people's heads? Self-writing documentation captures knowledge from the channels where people naturally share it (Slack, meetings, emails). The goal is to make the knowledge that's currently shared verbally persistent and searchable, which converts the most common type of "lost" information (things that were never documented) into findable knowledge.

How is this different from Google Workspace search? Google Workspace search only covers Google tools. It doesn't search Slack, email outside Gmail, CRM, project management tools, or meeting transcripts. A unified search spans all of your tools, which is where the cross-tool finding problem is solved.

Will the team actually use it? If searching is faster than asking a colleague (which it typically is from day one), the team uses it. The previous generation of knowledge tools failed because the search was poor. Semantic search that returns relevant results reliably earns adoption through usefulness.

How quickly does this make a difference? Connecting tools takes hours. The search begins returning results immediately. Most teams notice a reduction in "does anyone know where..." questions within the first few weeks.

What about information in tools we can't connect? For tools without a direct connection, email forwarding and web clipping provide manual capture paths. Forward important content to your library by email. Clip relevant pages from web-based tools. The information becomes searchable alongside everything from your connected tools.

How does this handle different file types? The search works across text documents, PDFs, presentations, spreadsheets, images, ebooks, audio and video transcripts, emails, Slack messages, and notes. The semantic search reads inside the content of each file type rather than only matching filenames.

What if different people have conflicting versions of the same information? The search surfaces all relevant results with their sources and timestamps. When two versions conflict, both appear with enough context for the searcher to determine which is current. Self-writing docs help resolve this by generating documentation from the most recent activity rather than maintaining outdated versions.

Can this help with compliance or audit requirements? Yes. A searchable record of decisions, communications, and processes provides an audit trail that's more complete than what most organisations can produce from manual documentation. Every claim in the self-writing documentation links back to its source, creating a verifiable chain of evidence.

Is there a limit to how much information the system can handle? The system is designed for large, growing libraries. Performance doesn't degrade with library size, which means the search at 50,000 items is as fast as the search at 500. This is a deliberate architectural choice to avoid the scaling problems that affect tools like Notion.

Related reading: The cost of scattered knowledge, The search tax, How to break down information silos, How to organise your digital life. Related pages: One search, Find anything, Search, Connections.