If, like me, you read Matt Shumer’s story “Something Big is Happening,” you’ve probably been thinking a lot about AI in the past few weeks.
For many leaders, it has led them to ask, “How do I get my people to use more AI?” I see ideas being tossed around like faster training, more tooling releases, or hiring a Head of AI to drive acquisitions. You’ve probably seen them too.
For some leaders in service businesses, like those I spend my time talking to every week, the penny has already dropped that thinking of AI as a productivity tool is a completely wrong framework.
We are on the brink of a business model change, not a productivity change.
Not all agents are created equal
Building an agent or skill to improve your daily workflow is one thing. Really building an agency business is another thing entirely. The first is real and precious. Many companies do. Automating the QA process, throwing LLM into the content workflow, using AI to accelerate analysis. Of course, these are real benefits but they are efficiency improvements layered over the existing operating model.
The agency business really looks very different. And startups right now aren’t just putting AI into their operations.
They are creating operational models where agents manage operations from day one, people are there only to make decisions that require true judgment. That’s a structurally different delivery model, not an instant version of what the service business is doing right now. The cost base looks different. The org structure looks different. The hiring model looks different. And once it’s built, you can’t close the gap by running training sessions.
I’m not yet aware of an enterprise business that has built a real agency operating system, at scale. Companies that say otherwise often describe a single person using a few agents with little domain knowledge. That’s a proof of concept, not a business model. The gap between those two things is where the real question of competition resides.
The threat is not what you think
What I believe to be true is that the pressure does not come from your existing competitors doing very well. It comes from newcomers who don’t have your overhead, your legacy systems, or your organizational inertia. And they don’t have your client relationship or your institutional knowledge either, and that’s important. If your costs are structurally higher than your competitor’s, that gap doesn’t last long. It becomes a price gap. In a service business with already thin margins, the price gap quickly becomes apparent.
At Brainlabs, our third hire was an engineer – we built a tech-enabled agency before it was a category. Many people will have first heard about us through our open source Google Ads scripts.
We are in a truly privileged position: we understand what it takes to build technology that is embedded at the heart of how a service business works. Despite our early days, I would be lying if I didn’t admit that we built the plane while it was flying. We also need to bet on investments and strategies against a landscape that is changing at an unbelievable pace. We are comfortable with this bet by sticking to key principles like LLM agnosticism and building a place, not pitching tents.
We currently use tools like Notion and Claude to rebuild our operating system. Not one-off agents for individual tasks but an enterprise ecosystem that leverages our existing context and combinations of intelligence. We are also completely rethinking our sourcing and recruiting strategy.
We know that the gap, between where most businesses are and where they are going, is actually a window. But it won’t stay open for long.
Companies keep making mistakes
Some organizations are responding by creating AI-specific roles such as the Head of AI or the AI ethics function. The instinct is understandable, but going wrong for some reason, you can’t outsource a business model change to one person.
If AI is going to reshape the way your business operates, it must be built into the core of how your business operates, distributed throughout your organization – not the function you provide.
So, for those who are asking, “How do I get my people to use more AI?”, I’d say I’ve asked some tough questions.
Say: “What does our business look like if agents handle most of the execution?”, “What do we need to build today to get there?” and “How do we price and sell this new business model?”
Businesses that get this right won’t be the ones to add more tools. They will be the ones who asked the hard questions early enough to act on them.
