Last fall, Leon Coe, AI strategist and the president of the Houston AI Club, joined an AI Power Hour with Build First and did something that blew me away: He built an entire CRM from scratch in an hour. (You can see the repo on Github here.)

Using just Claude Code and his own document archive of faux customer data and documentation about his business, he created not only the backend infrastructure that connected the dots, but a fully editable read-write frontend interface as well.

When I saw this, I thought, “Well, there goes B2B SaaS software…”

And then I spent the rest of the fall gearing up to build my own.

Yesterday’s launch of Claude Cowork confirmed a realization that many of us who have been building with Code Code have known for months.

AI’s real power isn’t just writing code, it’s acting like a second brain to help make sense of messy ideas, notes, and data over time.

Back when I started out in my career in tech, I used to work in sales and marketing, where my day was largely managed by the flow of inbound lead and pipeline management.

We used apps like Salesforce and Hubspot to help us make sense of our customer data. While this helped keep information centralized and organized, the only way the system worked was if every human on the team committed to updating the database the same way.

This meant our sales team spent an inordinate amount of time on manual tasks like data entry, keyword matching, and data cleanup. While we had the ability to create our own custom reports and sales dashboards, the amount of customization was limited based on the broad feature set of the entire tool.

In other words: We all became expert users of Salesforce, but we never learned how to build technology to invent and solve our own problems.

AI no-code tools like Claude Code (and now, Claude Cowork), reinforce this very clear paradigm shifting behavior.

Large language models read our sales data for semantic understanding, which means that the era of keyword-matching into tightly controlled database may soon be over. Not only that, but the prevalence of no-code building means that every sales person on the team might reasonably be able to build their own custom interfaces to help them do their work.

I tried this myself on Friday night and was pretty pleased with how far I got in just a three-hour build session. Here’s a view of my self-managed sales dashboard (with demo data).

The only reason I was able to build myself a personal sales dashboard in one night is not because I have access to more unique insights or tooling than anyone else.

It’s because I’ve spent the past year teaching myself how to think like an engineer.

I’ve played with dozens of AI tools and I’ve invested thousands of hours into tinkering with mini-builds and disposable apps, largely to practice what it feels like to put new digital artifacts out into the world.

This is the set of skills that I believe anyone who works with a computer will need to understand in order to thrive in the AI age.

While it’s certainly possible that the AI advances will continue at such a rapid rate that none of this will matter (because the AI will just abstract away any complexity and predict what we need with higher and higher accuracy), I still believe that learning how to command get technology to do what we want is an incredibly important life skill.

As I’ve learned first-hand, English is a perfectly good "programming language” to start with. One thing that has helped me immensely is by using an acronym called POP-IT to remind me to always start fresh builds by anchoring on the problem I am looking to solve. If this helps you too, I’d love to know what you build.

Last fall, Leon Coe, AI strategist and the president of the Houston AI Club, joined an AI Power Hour with Build First and did something that blew me away: He built an entire CRM from scratch in an hour. (You can see the repo on Github here.)

Using just Claude Code and his own document archive of faux customer data and documentation about his business, he created not only the backend infrastructure that connected the dots, but a fully editable read-write frontend interface as well.

When I saw this, I thought, “Well, there goes B2B SaaS software…”

And then I spent the rest of the fall gearing up to build my own.

Yesterday’s launch of Claude Cowork confirmed a realization that many of us who have been building with Code Code have known for months.

AI’s real power isn’t just writing code, it’s acting like a second brain to help make sense of messy ideas, notes, and data over time.

Back when I started out in my career in tech, I used to work in sales and marketing, where my day was largely managed by the flow of inbound lead and pipeline management.

We used apps like Salesforce and Hubspot to help us make sense of our customer data. While this helped keep information centralized and organized, the only way the system worked was if every human on the team committed to updating the database the same way.

This meant our sales team spent an inordinate amount of time on manual tasks like data entry, keyword matching, and data cleanup. While we had the ability to create our own custom reports and sales dashboards, the amount of customization was limited based on the broad feature set of the entire tool.

In other words: We all became expert users of Salesforce, but we never learned how to build technology to invent and solve our own problems.

AI no-code tools like Claude Code (and now, Claude Cowork), reinforce this very clear paradigm shifting behavior.

Large language models read our sales data for semantic understanding, which means that the era of keyword-matching into tightly controlled database may soon be over. Not only that, but the prevalence of no-code building means that every sales person on the team might reasonably be able to build their own custom interfaces to help them do their work.

I tried this myself on Friday night and was pretty pleased with how far I got in just a three-hour build session. Here’s a view of my self-managed sales dashboard (with demo data).

The only reason I was able to build myself a personal sales dashboard in one night is not because I have access to more unique insights or tooling than anyone else.

It’s because I’ve spent the past year teaching myself how to think like an engineer.

I’ve played with dozens of AI tools and I’ve invested thousands of hours into tinkering with mini-builds and disposable apps, largely to practice what it feels like to put new digital artifacts out into the world.

This is the set of skills that I believe anyone who works with a computer will need to understand in order to thrive in the AI age.

While it’s certainly possible that the AI advances will continue at such a rapid rate that none of this will matter (because the AI will just abstract away any complexity and predict what we need with higher and higher accuracy), I still believe that learning how to command get technology to do what we want is an incredibly important life skill.

As I’ve learned first-hand, English is a perfectly good "programming language” to start with. One thing that has helped me immensely is by using an acronym called POP-IT to remind me to always start fresh builds by anchoring on the problem I am looking to solve. If this helps you too, I’d love to know what you build.