I spent part of this week deep in the build process, working through the pipeline from raw business data to a working knowledge base. Cleaning data, chunking it, generating embeddings, loading it into a vector database.
It was not glamorous work. It was also the most important thing I have done this month.
Here is why.
The spreadsheet that changed how I see everything
I was on a client call this week with a company that has been in business for over 10 years. At some point they mentioned, almost in passing, that they have a spreadsheet. An automated Excel file. Built over a decade. It contains the accumulated logic of how their business actually thinks. Their pricing rules, their qualification criteria, the edge cases they have learned to handle, the patterns they have recognized over thousands of transactions.
It is, by any honest definition, the institutional brain of that business.
They had been trying to figure out how to migrate it into a CRM.
I had to stop them.
Migrating that spreadsheet into a CRM would not preserve its value. It would destroy it. A CRM is built around someone else's schema. It has fields and structures and workflows that were designed for a generalized version of a business like theirs. To get that data in, they would have had to flatten it, simplify it, cut the parts that did not fit.
Ten years of domain knowledge would have become a contact list with some notes attached.
What that spreadsheet actually is, once you understand what knowledge bases can do now, is a ready-made foundation for something fundamentally different. Not a generic assistant. Not a chatbot that knows their FAQs. A system that carries a decade of learned judgment and can apply it at scale.
I call this Encoded Expertise. Domain knowledge structured in a format AI can actually execute. It is the most valuable thing your business owns and it is probably stuck in a format nobody can use.
That is not a migration project. That is a build.
Why SaaS is the wrong frame
The assumption behind most business software is that you will adapt to fit the tool. You adjust your workflows, your naming conventions, your processes, your reporting structure to match what the software can do. You have been doing this for so long it probably does not feel like a compromise anymore. It just feels like how software works.
It does not have to work that way anymore.
A knowledge base built around your specific business is not a SaaS product. It does not ask you to bend. You feed it what you know. You structure that knowledge in the way your business actually thinks. And then you have a system that reflects ten, twenty, thirty years of accumulated expertise rather than a generic starting point that you have gradually tried to customize.
The gap between those two things is not a feature gap. It is a fundamental difference in what you are building.
Off-the-shelf software is someone else's answer to a generalized version of your problem. Encoded Expertise is your answer to your specific problem, built from what you have already learned.
The part nobody talks about
Here is the thing I keep running into, both in my builds this week and in every client situation I encounter.
The technology is not the hard part.
The hard part is the data. And almost nobody is talking about it honestly.
Most businesses have years of accumulated knowledge that is scattered across a dozen different places. Emails. Old documents. Spreadsheets that one person built and nobody else fully understands. SOPs that were written in 2019 and never updated. Meeting notes that live in someone's personal drive. Processes that have never been written down because the person who does them has always just been there to do them.
Before any of that can power a knowledge base, it has to be cleaned. Organized. Structured in a way that AI can actually retrieve and use. Duplicates removed. Outdated information flagged. The 20 percent of data that provides 80 percent of the value identified and prioritized.
This is not exciting work to pitch to a client. It does not make for a dramatic demo. But it is the difference between a knowledge base that actually works and one that hallucinates, gives inconsistent answers, and gets abandoned within three months.
Bad data going in means bad answers coming out. Every time. No exceptions.
The businesses that get this right, that do the unglamorous work of getting their knowledge into a clean, structured, AI-ready state, are not just building a better tool. They are building something their competitors cannot easily replicate, because the knowledge itself is proprietary. The tool is not the moat. The Encoded Expertise is.
What this means in practice
The ten-year Excel file is not unusual. Every business I talk to has a version of it. Knowledge that has been accumulating for years, in formats that were never designed to be shared with an AI system, that is sitting there waiting to be unlocked.
The question is not whether the tools exist to do something with it. They do. The question is whether the business is willing to do the work of surfacing what it actually knows before trying to build on top of it.
That work takes someone who understands both the business and the technology. And it is almost always the part of the project that gets underestimated or skipped entirely, which is exactly why most AI implementations underperform.
The businesses that invest in that foundation are not just going to have better AI tools. They are going to have a competitive asset that deepens over time as they continue to capture what they learn.
That is what built not bought actually means.
If your business has been accumulating knowledge for years and you want to know what it would take to turn it into something that actually works for you, book a 30-minute strategy call. We will identify the highest-leverage knowledge in your business and figure out where to start.
Not ready for a call? Just reply to this email and tell me what your version of the ten-year spreadsheet is. I read every reply.
See you next Tuesday.
Nick
PS - I have been building a platform that does this extraction work through voice-guided interviews. It tracks knowledge coverage by section and shows you exactly how much of your business is documented versus stuck in someone's head. Here is what it looks like for one business after 5 sessions:
33% coverage. Customers section at 80%. Operations at 50%. Vendors and Finances barely started. The red and green bars tell you exactly where the gaps are. If you want to see what this would look like for your business, reply and I will show you.
Built, Not Bought is a weekly newsletter about encoding expertise into systems that scale. Subscribe on LinkedIn or Beehiiv so you do not miss an issue.
The News Source 2.3 Million Americans Trust More Than CNN
Tired of spin? The Flyover delivers fast, fact-focused news across politics, business, sports, and more — free every morning. No agenda. No paywall. Join 2.3 million readers who trust us to start their day right.