The 5 commandments of information discoverability
Even with supercomputers at our fingertips, knowledge is only as discoverable as we make it.
Sharing information freely makes it easier for everyone to do their jobs. For example, opening up project plans and retrospective notes can save teams from re-inventing the wheel, and makes aligning on goals much simpler. Documenting why a decision was made and what other options were explored helps teams understand the “why” behind where the company is headed and how their work fits in.
There’s the proactive kind of sharing, where you send out a link, document, or video to a select group of people because it’s relevant to what you’re working on together. There’s also the passive kind of sharing, where information lives in your company’s systems and can be accessed as needed by whoever needs it.
Collectively, this body of knowledge is more than just an archive. It’s fuel that can be used to empower teams and propel them forward. But knowledge is only as accessible as we make it. In order for passive sharing to work, information must be stored in a way that makes it easy to retrieve.
In the past, we accomplished this by giving documents descriptive titles so people could find what they need by searching a repository or browsing through network folders. Now, however, AI-powered search is increasingly popular, especially in larger companies with mountains of data stored in myriad systems.
Global search that reaches across systems and apps is a massive leap forward for discoverability. But beware: AI tools still struggle with unstructured data, and aren’t yet capable of extracting information from images accurately and consistently.
It’s time to adjust the way we present and organize information at work such that it’s easier for both humans and AI to access and understand. These are the five (new) commandments of knowledge discoverability.
Commandment zero: thou shalt not hoard information
Don’t keep knowledge tucked away where others (including AI) can’t find it, such as your email inbox, text messages, and documents saved only on your local hard drive. Information stored in private spaces essentially ceases to exist for anyone but its owner. If AI can’t “see” it, AI can’t surface it, leading to inaccurate or incomplete search results. Use tools like Confluence, SharePoint, shared network drives, or knowledge bases instead. And make sure to transfer useful information discussed in private channels to shared spaces so it’s visible to teammates who might need it.
1. Design for discoverability
Structure matters. Any formatting that makes it easier for humans to understand the information will also make it easier for AI.
Nobody likes “wall of text” documents, for example. In fact, AI technology tends to lose sight of the context when processing long blocks of text, which can cause it to return results that don’t match the searcher’s intent. Using section headlines goes a long way in helping AI (and humans!) understand the big picture, not just individual paragraphs.
Lists are also helpful. First, they provide a welcome visual break for human eyes. Second, when we create lists, we tend to be both concise and precise. As a result, lists become a rich source of information that AI can scan easily.
Last, use tags or labels to tie related documents together and show how they are related. E.g., “customer loyalty campaign” or “employee benefits.”
2. Prioritize clarity over volume
Information overload is real, so be thoughtful about exactly what (and how much) information you’re including in each document or record. Consider who the information is intended for and the context in which they’ll be accessing it. Tailored, nutrient-dense documentation is better than endless pages that go too far into the weeds or off on tangents. Be sure to make titles and section headlines clear and contextually rich, too – ”Q3 2025 Sales Performance Report” is more helpful to both humans and AI than “Q3 Report.”
Pro tip
Confluence users can take advantage of the Expand macro. Use it in cases where some people will want all the details they can get, while others just need the basics. This technique, known in the UX world as progressive disclosure, fights information overload for humans while still making all the details available to AI.
As you’re writing, address topics directly using clear language. This can be deceptively difficult for people who aren’t comfortable writing! We tend to ramble and use overly elevated language to cover up our insecurity. If this sounds like you, take heart: AI tools can make your drafts more concise, or even kick-start the writing process by generating a draft that you can refine.
This isn’t to say you should avoid the figures of speech, slang, cultural references, or humor that make knowledge easier and more enjoyable to take in. In fact, conversational language is preferable whenever it’s appropriate. Not only are our brains trained on casual conversation, but much of the data LLMs are trained on is also written in a casual tone, so AI will feel right at home. Plus, people don’t search for ways to “optimize our synergies across teams,” they simply ask whether other teams are working on a similar project. Yet another reason to go easy on the corporate-speak.
3. Strive for “open by default”
Knowledge must be available to those who need it, when they need it, in order for it to have any value. In companies with a culture of locking everything down, making knowledge discoverable may require a fundamental shift in how information is stored.
Instead of a “closed by default” approach where you have to make a case for opening documents up, make them open by default and justify your way to locking it. Obviously, some information needs to be kept private to most people, such as HR records, confidential financial performance data, or embargoed press releases. That’s fine. But your project plan, for example, should generally be unlocked (even before it’s finalized).
Pro tip
Everybody wants to make the best impression possible at work, but that doesn’t mean you have to hide in-progress work. Just include a “draft” or “WIP” disclaimer at the top and leave the document unlocked. This way, you can safeguard your professional reputation, while still making the information available to teammates and AI search.
Beyond individual documents, striving for open access means evaluating and optimizing your systems of record. As much as possible, make them open to AI tools and anyone who might benefit from the information they contain. If user seats are cost-prohibitive, ask about free or inexpensive read-only seats for infrequent users so access isn’t unnecessarily hindered by licensing models.
Also, the value of knowledge plummets when it can’t flow from one system to another. Are there five separate Jira Cloud subscriptions floating around your company? Consider consolidating them into a single instance. And be sure to integrate apps whenever possible.
4. Be multi-media savvy
Sharing information via video can be a life-saver for people who struggle with writing, as well as for people who absorb information better by listening. Video messages, like those you create with Loom, work great for department updates, tutorials, and other one-way communications. Humans receive all the same information as they would in an email, but with the added benefit of seeing the presenter’s body language and facial expressions, which carry rich contextual meaning. Plus, they can zip through the recording at double speed to save time. Platforms like Loom also generate a transcript that AI can search later.
Imagery, on the other hand, presents some challenges to be aware of. Diagrams, flowcharts, photos, and illustrations are fabulous ways of conveying conceptual and contextual information to humans. But even though AI technology is advancing all the time, it still struggles to understand flowcharts and diagrams accurately, even when optical character recognition technology is layered on top of an LLM. AI is also unreliable when it comes to extracting text that is embedded within an image (think meme captions, for example). And again: AI can’t retrieve what it can’t see.
Does this mean diagrams are dead? No. It just means diagrams and other images need to have clear, descriptive captions that AI can access.
If you can add alt-text, that’s even better. (Here’s how to do it in Confluence, PowerPoint, and Google Docs.) Alt-text is also what screen-reading software looks for, making your images accessible and useful to people with vision impairments. Depending on the image, your alt-text might include:
- The exact verbiage on a graphics-heavy slide
- A description of an illustration
- A summary of what a diagram explains
- The purpose and description of a flowchart
Just be sure to include the keywords people would be likely to use in search queries.
5. Keep knowledge fresh
Outdated content confuses people and AI alike, so regular refresh cycles are important. Ideally, teams will foster a culture of upkeep. That might mean carving time out each quarter to look through the archives for anything that is no longer current or relevant. Or it might be ad-hoc, where team members are encouraged to flag outdated content as they come across it.
It’s also worth refreshing your knowledge of what AI can and can’t do. Search technology is becoming more capable every day, so don’t delete those product updates from your AI vendor without reading them!
Open, free-flowing information fosters collaboration, reduces redundant inquiries, and ensures that knowledge is available to inform decisions. By embracing an “open by default” mindset and actively working to make both information and the systems that house it universally accessible (while respecting confidentiality needs), you can use your organization’s collective knowledge to empower teams and unleash more of their potential.
Special thanks to Sven Peters for his contributions to this article.
