Making AI Work

Most companies believe they are making progress with AI. A closer look suggests they are barely scratching the surface.

This week, we are digging into a framework that cuts through the confusion around AI adoption in organizations. Whether you run a team, manage a startup, or work independently, the three-part model below applies directly to how you compete and grow over the next few years.

Let's get into it.

Why Individual Wins Don't Add Up to Organizational Gains

Here is the tension that almost every company is sitting inside right now. Workers are quietly using AI tools, seeing real improvements in their own output, and saying nothing. Meanwhile, company-wide results look flat.

Four research-backed facts help explain what's going on:

They all run their businesses on beehiiv. Newsletters, websites, digital products, and more. beehiiv is the only platform you need to take your content business to the next level.

🚨Limited time offer: Get 30% off your first 3 months on beehiiv. Just use code JOIN30 at checkout.

Start building for 30% off today.

What closes that gap? Three forces working together: Leadership, the Lab, and the Crowd.

Leadership: The Part That Has to Come First

Two high-profile memos went viral this year. The CEO of Shopify and the CEO of Duolingo both sent clear signals to their organizations that this shift is serious and immediate. That kind of urgency matters.

But urgency alone doesn't move people. Research from Wharton professor Andrew Carton shows that workers are not motivated by abstract statements about efficiency gains. They respond to concrete, vivid pictures of what the future actually looks like. What will my job involve next year? Will the time I save get used to grow the team or reduce it? Will I be rewarded for finding better ways to work, or quietly resented?

Leaders don't have to have every answer. But they need a direction they're willing to share out loud.

The Crowd: Where Innovation Actually Happens

The employees inside your organization are not waiting for a playbook. They are already experimenting. The problem is that many of them are doing it quietly.

Studies suggest that about 20% of workers use officially sanctioned tools at work, while over 40% admit to using their own. The gap points to something significant. Workers who have found real performance gains are hiding them, either to protect their reputation, avoid extra work being loaded onto them, or stay out of range of cost-cutting decisions.

This is sometimes called the "secret cyborg" problem, and it represents a significant loss for any organization trying to learn faster than its competitors.

Experienced workers, specifically those who understand their domain well, are the best positioned to figure out where new tools are actually useful. They can tell when output is good and when it falls short. The goal for any organization is to make it safe and worthwhile for those people to share what they're learning.

The Lab: Building Things Before They're Ready

The third piece is a dedicated internal group focused entirely on figuring out what's possible and building it quickly. This isn't a traditional research and development function. It operates more like a startup inside the organization: rapid prototyping, constant testing, and a direct pipeline to the Crowd.

One of the Lab's most underrated responsibilities is building internal benchmarks. Almost every standard test for evaluating new tools measures things like math, trivia, or code. None of that tells you which tool is best at writing the kind of memos your team actually produces, or analyzing the financial models specific to your industry.

A useful example: the author of the source framework tested an agent called Manus, built on Claude, by giving it a fictional startup brief and a detailed financial model from a Wharton business simulation. The agent produced a 45-page business analysis, a pitch deck, a test of the financial assumptions, and a draft website, all from a short prompt and a spreadsheet. The output wasn't flawless, but it was more thorough and accurate than what most students produce in hours of work. That kind of test tells you something real about what a tool can and can't do. The Anthropic benchmarking guide is a practical starting point for building your own.

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

Learn more

The Bigger Rethink

There is a deeper question underneath all of this. Every process, structure, and workflow inside an organization was built around the assumption that intelligence is scarce and expensive. Research took weeks. Analysis required specialists. Content required production time.

None of those bottlenecks work the same way anymore. When research takes minutes, the constraint becomes knowing what to research. When code can be written quickly, the limitation is knowing what to build. When content can be produced instantly, what matters is knowing what will actually reach people.

The organizations that adapt fastest are the ones building feedback loops between all three layers: Leadership setting the direction, the Lab testing what's possible, and the Crowd discovering what actually works in practice. Together, they learn faster than any one piece could on its own.

Waiting for things to settle down before starting is not a strategy. The companies pulling ahead now are the ones willing to experiment while the picture is still blurry.

Worth Your Time This Week