Your Company Already Has the Answers. Claude Just Made Them Findable
The small law firm that found their smoking gun on page 113 – and why every executive should care about RAG
I almost missed it.
Buried in Claude's latest update notes was a feature that really changes how companies can use AI. Not another marginal improvement or fancy new model. Something that solves a problem I've watched companies struggle with for years.
Claude Projects now lets you build knowledge bases up to 2 million tokens (about 5,000 pages of text). If that number means nothing to you, here's what matters: your teams can finally work with ALL their critical documents in one place. Not cherry-picked excerpts. Not the "most important" files. Everything.
The Problem Nobody Talks About
Last week, I sat in a boardroom watching a new CEO try to understand why his predecessor killed a product line that seemed profitable on paper. The answer was buried somewhere in five years of strategy documents, board presentations, and email threads. Three consultants spent two weeks finding it.
This is the dirty secret of modern business: we're drowning in our own knowledge. Every merger multiplies the problem. Every departing executive takes important context with them. Every strategic pivot orphans valuable insights.
Here's what Claude did about it. They kept their standard 200,000 token context window for the main conversation (think 500 pages you're actively working with), but added something new: Projects can now have a separate searchable library. You upload your documents – contracts, strategies, reports, whatever – and Claude uses RAG (Retrieval Augmented Generation) to search through them when you ask questions.
Why This Actually Matters
I'll be honest. When I first heard "RAG integration," my eyes glazed over. Another technical acronym. But then I watched it save a small law firm from drowning in discovery documents.
Picture this: boutique firm, product liability case, and opposing counsel just dumped 400 new documents on them. Classic intimidation tactic. The senior partner called me, practically laughing through his exhaustion: "They think we'll miss the smoking gun in all this noise."
We created a Project and his paralegals spent about an hour uploading everything. Depositions. Expert reports. Internal emails from the defendant. Manufacturing specs. Quality control reports. Years of correspondence between engineers who seemed increasingly worried about something.
Then I asked Claude: "Find me every instance where someone raised safety concerns about the product design."
In forty-five seconds, it found an engineer's email buried on page 113 of a document dump, connected it to a footnote in a quality report from two years earlier, and linked both to a deposition where the same engineer suddenly "couldn't recall" those exact concerns. The pattern was clear: they knew about the defect and buried it.
A team of associates would have needed three weeks and $50,000 in billable hours to maybe find those connections. Maybe.
The kicker? This wasn't some special enterprise feature that requires IT involvement. You literally just drag and drop files into your Project. Claude handles the rest.
What Smart Companies Are Doing Right Now
The companies getting ahead aren't waiting for perfect use cases. They're starting with their messiest, most document-heavy problems.
One small CPG company I work with built a regulatory compliance Project. Years of FDA communications, internal compliance documents, and legal opinions – all searchable in seconds. Their legal team went from spending days researching past cases to getting answers during the meeting where questions come up.
Another client, a tech company, created a "Customer Truth" Project. Every sales call transcript, support ticket, and feature request from their top 100 accounts. Product managers now make roadmap decisions based on actual customer language, not filtered summaries.
The pattern I'm seeing? Start with knowledge that's currently scattered across minds and hard drives. The stuff people email each other asking, "Do you remember that document about...?"
The Uncomfortable Truth
Your competitors are probably reading this too. The question isn't whether AI will change how companies handle information. It's whether you'll be the company that figures it out first or the one playing catch-up in eighteen months.
At $20/month for Claude Pro or $30/month per user for Claude Team (minimum 5 users, so $150/month total), this isn't a budget-breaking experiment. It's less than you spend on coffee for those endless document review meetings.
Here's my advice: pick your most frustrating information problem. The one where data exists but nobody can find it. Give a small team a few days to build a complete Project around it. Upload everything. Measure how long it takes to answer complex questions before and after.
I've been in this business long enough to recognize when something actually changes the game. This is one of those moments. Not because the technology is revolutionary, but because it solves a problem every company actually has.
The companies that get this right won't just make better decisions. They'll make them faster. And if the last few years taught us anything, speed might be the only real advantage left.
If you enjoyed this article, please subscribe to my newsletter and share it with others! Looking for help to use AI in your organization? Wondering where to start or what tools to use? Reply directly or reach out: steve@intelligencebyintent.com