Making AI Practical for Every Company
Personal newsflash, thoughts on how companies can use AI to increase productivity.
Exciting news: I'm starting as the Product Manager at maia, a company that's helping medium-sized businesses manage all their complex knowledge more effectively. You know, the kind of information buried in thousands of PDFs, spreadsheets, graphics, and more—crucial when you need to figure out how many of your products meet specific criteria or how many rounded door handles on the market are at least 2 cm thick. Real questions that aren't easily answered by tools like ChatGPT or Claude.
Note: I'm also excited to join a local company based in Germany with a hub in Leipzig. Developed in Germany, hosted in Europe, and fully GDPR compliant.
Maia's team is talented and passionate, and they've already developed an impressive product. As you'd expect, it relies heavily on generative AI!
Since this newsletter is all about "Making your business smarter with data & AI," I’ve got plenty of thoughts on this new challenge, which combines the tech I love with the problem-solving I've been doing throughout my career.
Let’s dive in—no particular order.
Today's newsletter is sponsored by my brand new free 3-Day AI-Assisted Decision-Making Challenge!
1. We, You, Everyone Needs to Build More Tools
Edison invented the light bulb in 1878—but it wasn't the first electric light. Electrical light existed for decades before Edison, and electricity was already being used in many ways. So why is Edison synonymous with invention? He created practical lights and combined them with a power system, making adoption easy. Electricity, after all, is a general-purpose technology, but general-purpose technologies carry what I call the Curse of Generality.
General-purpose tech can be used for almost anything, but that often means it's not effectively used for anything specific—unless we create two things: 1) Tools and 2) Practical systems that integrate those tools with the general technology in accessible ways.
Today, we have another general-purpose technology—AI. And like electricity, it’s impressive but doesn’t fit neatly into existing workflows.
And that’s okay! These amazing LLMs and generative AI tools have the Curse of Generality. What we need now are tools and systems to effectively harness them. I’m excited to be part of that effort.
2. A Corollary for Companies
Want to introduce ChatGPT in your company? Great! But it's not just about running a few training sessions and watching productivity skyrocket.
You need to invest heavily in systems and tools, which could take various forms:
Third-party tools, like Yoodli.ai for training staff, or Descript for fast video content creation—or, of course, maia.
Templates specific to your team’s needs—pre-written prompts for ChatGPT or Claude that streamline particular tasks.
Workflows that combine these templates with smaller tools.
Light software solutions that your IT department can build—whether it’s a few bash scripts, Chrome Extensions, or something more integrated.
The goal is to make sure your employees don't use a powerful tool without understanding it properly—i.e., use a powerful technology without a practical guide.
3. Data Is No Longer About Decisions—It's About Knowledge Work
For almost a decade, I’ve been saying that data empowers people to make better and faster decisions. But I've changed my mind.
It's not just generative AI and LLMs; it's been a gradual shift. Today, data and AI aren't just about making decisions—they’re about enabling knowledge work.
You could argue that knowledge work ultimately leads to decisions, but it's more than that. Knowledge work isn't called "decision work" for a reason. The core tasks are gathering, curating, analyzing knowledge, and then deciding.
Data and AI excel in supporting these aspects of knowledge work. Thinking of them merely as 'decision tools' underestimates their broader capabilities.
4. My Mental Model for Empowering Knowledge Work
Peter Drucker coined the term "knowledge workers." Today, many of us fit this description, with machines handling much of the physical work. In my view, knowledge work splits into three pillars:
Generating knowledge (researching, taking notes, thinking things through)
Analyzing the information and making decisions
Creating something new based on those decisions
For example, when I wrote my last blog post, I spent an hour researching on Twitter and reading various articles. That was knowledge generation. Then I sifted through my notes, filled in gaps, and discarded irrelevant points for about an hour—that was analyzing and deciding. Finally, I spent another hour writing it all out—creating something new.
Roughly, my time splits equally into these three activities.
5. The Challenge for Companies and Knowledge Workers
If most work can be divided into these three activity pillars—and AI excels in each—shouldn't companies be seeing 10x productivity gains?
Yet many companies struggle to realize these benefits. Why?
Blending Knowledge: My work example involved both my own knowledge and new research. That's challenging for AI. The approach to AI differs if you're dealing with "something we already know" versus "something entirely new." In the first case, AI helps with slight quality improvements; in the latter, it boosts speed significantly.
Categorizing Work: It's hard for individuals to categorize their activities into those pillars in a repeatable and systematic way—something essential to building effective tools and systems.
6. Why Companies Are Struggling to See the Potential
Everyone’s impressed by ChatGPT, Claude, and other AI tools, but many companies still aren’t seeing productivity gains.
A century ago, when a worker installed a lightbulb in a workshop, the productivity boost was obvious—it literally lit up the room. Today, most work happens inside our heads or in digital form. Harder to see direct, physical results.
Moreover, electricity wasn’t something you could just figure out—but AI is different. Its human-friendly interface—language—means anyone can use it, even if not very effectively.
This also means that everyone in a company can experiment with tools and prompts—but making these gains widely available and tracking them remains a challenge. As Ethan Mollick puts it: The gains are there; we just struggle to make them available and visible.
7. But Medium-Sized Companies, Really?
A question I get a lot is: "Why would a medium-sized company, with no IT department and no cloud infrastructure, be ready to adopt AI?"
It’s a fair question. But the key is understanding the difference between new technology and new general-purpose technologies.
Almost every medium-sized company uses the internet today. Many adopted electricity quickly. Smaller companies were among the first to get websites, even if they lacked IT departments or didn’t migrate to the cloud.
The difference is that the internet is a general-purpose technology, while the cloud is not. It’s just a new technology. All signs suggest that AI is a general-purpose technology that is transforming every industry and company size rapidly. That’s why I believe in making AI accessible to everyone.
Feel free to ask me again in 6 months about what I'm thinking. This is just a brain dump.