Claude, Finding It

Enterprise search lets Claude read what's actually written inside your documents, across your drive, wiki, and chat, bounded strictly by your own access.

2:10 video3 min readWatch on YouTube

Someone assumes Claude would search everywhere across a company's systems once given the chance. It doesn't. It only reaches what you can already open yourself. The real question is narrower and more useful: can Claude find what your company already wrote down, using nothing more than your own existing access?

The buried answer problem

In any company that's been running for a while, the answer to a given question usually already exists somewhere, in a colleague's document, a slide deck, a chat thread. The problem isn't that the information is missing. It's that you can't reach it. And the natural next thought, "I already searched, it isn't there," is often wrong for a specific reason: normal search only matches a file's name, not what's actually written inside it. Call a document "Q3 notes" when it actually holds the reason your team switched payment processors, and searching for "payment processor" turns up nothing, because the filename and the content have no relationship to each other.

Reading inside, everywhere at once

Enterprise search fixes exactly that gap. It reads the words inside your documents rather than the label on the folder they're sitting in. It searches your drive, your wiki, your shared folders, and your chat, all at once, so you don't have to remember where you filed something, only roughly what it was about. And the search itself works in plain language, the way you'd ask a colleague a question, with no boolean operators or query syntax required.

Grounded, not generic

It goes a step further than just locating a file. It can pull the found content straight into the conversation. Ask "how did we decide on Stripe over PayPal, and why," and if anyone wrote that reasoning down anywhere, even casually in a meeting note, Claude retrieves the actual content and answers from it directly. That's the real difference this makes: without enterprise search, Claude reasons from general best practices. With it, Claude reasons from your team's own actual decisions.

Bounded by your own access

This is the part worth being precise about. Enterprise search only searches what you could already see using your own credentials and your own permissions. Nothing new gets unlocked. Finding that buried Q3 notes file doesn't mean Claude reached somewhere it otherwise couldn't; you could always have opened that file yourself, it just would have taken longer to find. The reverse is also true: not finding something doesn't mean it isn't written down somewhere. It might simply not be connected or indexed yet, which is why enterprise search does nothing until you actually point it at your sources. The first indexing pass takes a while to complete; searches after that are fast.

What this actually changes

Enterprise search doesn't give Claude new access to anything. It gives you faster reach over what you could already see, because it's reading what's actually written rather than only what a file happens to be named. That's a meaningful difference for anyone who has ever known an answer exists somewhere in the company and just couldn't find it.

Key takeaways

  • Normal file search matches filenames, not content, so a document with a vague or misleading name won't turn up even when it holds exactly the answer you need.
  • Enterprise search reads the actual content inside documents, across your drive, wiki, shared folders, and chat, in one plain-language query.
  • It can pull retrieved content directly into a conversation, letting Claude answer from your team's real decisions instead of generic advice.
  • Enterprise search is strictly bounded by your own existing access and permissions; it never reaches anything you couldn't already open yourself.
  • It requires an initial indexing pass before it works, and it stays silent about anything not yet connected or indexed.

Try it yourself

Try this prompt: "Help me set up enterprise search for my team. First, ask me which tools our documents actually live in, drive, wiki, shared folders, chat. Then tell me the three questions I should try first to prove it works, and one decision I should start documenting this week so it's searchable later." This walkthrough is part of Humanitarians AI's series on how Claude actually works, aimed at anyone meeting these features for the first time.

Chapters

  1. 0:00Cold open — can Claude search everywhere?
  2. 0:10The buried answer — filenames lie
  3. 0:35It reads inside, everywhere at once
  4. 0:54Grounded, not generic
  5. 1:15Bounded by your access — the anchor returns
  6. 1:39Carry-out
  7. 1:48Your turn
  8. 2:06Outro
Full transcript(auto-generated, with timestamps)

Cold open — can Claude search everywhere?

[0:00]Someone assumes Claude would search everywhere in the company. It only reaches what you can already open. So, can Claude find what your company already wrote down using just your own access?

The buried answer — filenames lie

[0:10]In any company that's been running a while, the answer already exists. It's in a colleague's doc, a slide deck, a chat thread. You just can't reach it. So, the natural next thought is, I already searched, it isn't there. But, normal search only matches the file's name, not what's inside it. Call a document Q3 notes when it actually holds the reason you switched payment processors, and searching payment processor finds nothing. The name and the content have nothing to do with each other. Enterprise search fixes exactly that. It

It reads inside, everywhere at once

[0:36]Reads the words inside your documents, not the label on the folder. It searches your drive, your wiki, your shared folders, and your chat all at once, so you don't have to remember where you put something, only roughly what it was about. And you just ask in plain language the way you'd ask a person, no boolean operators, no query syntax. It goes one step further than finding the

Grounded, not generic

[0:55]File. It can pull that content straight into the conversation. Ask how did we decide on Stripe over PayPal and why? And if anyone wrote that down, even casually in a meeting note, Claude retrieves the actual reasoning and answers from it. That's the difference between generic advice and grounded advice. Without it, Claude reasons from best practices. With it, Claude reasons from your own decisions.

Bounded by your access — the anchor returns

[1:15]Here's the part to be clear about. It searches only what you can already see, your own credentials, your own permissions, nothing new. So, finding that Q3 notes file doesn't mean Claude reached anywhere new. You could always have opened it yourself. And not finding something doesn't mean it isn't written down somewhere. It might just not be connected or indexed yet, which is also why it does nothing until you point it at your sources. The first index takes a while. And after that, searches are fast. Enterprise search doesn't give

Carry-out

[1:40]Claude new access. It gives you faster reach over what you can already see, because it reads what's actually written, not just what it's named. Your

Your turn

[1:48]Turn. Here's the prompt, read it with me. Help me set up enterprise search for my team. First, ask me which tools our documents actually live in, drive, wiki, shared folders, chat. Then, tell me the three questions I should try first to prove it works and one decision I should start documenting this week so it's searchable later. Liam In for Bear. Claude finding it. Liam In for Bear.

More from Claude

Humanitarians AI Lyrical Literacy Project