The Tech Is Ready. Your Company Isn't.

Everyone is watching for the day AGI arrives. The collision that matters has already started.

AI-first companies evolve every day. New tools, new workflows, new ways of working, sometimes overnight. Most companies still change on an annual calendar. Their infrastructure, their approvals and their overhead were built for a slower world.

I've spent my career figuring out why people do what they do, then building for it. Right now, what people do has moved faster than the companies they work for. That's where this collision plays out.

Two speeds

Imagine standing on a street corner in Manhattan. Take 30 normal steps and you're barely down the block.

Now start with a one-yard stride and double it every step. After ten steps you're about a dozen blocks uptown. After twenty you're most of the way to Chicago. After thirty you've passed the moon.

AI moves like the second walk. Companies move like the first. Budgets get set once a year. New tools wait months for security review and procurement. A reorg takes a quarter to plan and a year to settle.

Both walks are happening at once, on the same street, inside the same company.

Inside the building

Walk into almost any office and you'll find people running real work through AI on their own. They research with it, draft with it, build small tools with it. Then they paste the result back into a process designed years before any of this existed.

The individual is on the fast walk. The company around them is still on the slow one. That gap is where the frustration lives, and where most of the wasted value sits.

There's plenty of pitch theater around AI right now. Big announcements, big licenses. Very little of it changes how a Tuesday afternoon actually works.

Where it breaks

I break AI work into three layers. Model, harness, data. The model is the engine. The data is what it knows about your business. The harness is everything wrapped around it: the tools, the context, the triggers, how work gets handed off.

Most companies buy the better model and expect the work to change. It doesn't. The gains show up in the harness, and in what you put into it.

I've seen the difference firsthand. I built a simple agent for one use case, and it couldn't reliably read its inputs. Then I swapped out its harness and kept the same model underneath. Instantly, my agent started getting it right. I hadn't changed the instructions. The platforms are improving the harness that fast, and they're doing it for you. What's yours to build is what runs on it: the skills, the context, the workflow.

Two kinds of slow

Companies are slow for two reasons, and AI only fixes one of them.

The first is coordination. Status meetings, handoffs, recaps of the recap, the process only one person understands. All of that is information moving from one person to another, which is exactly what AI is good at. I built an AI that remembers everyone I work with. Context that used to live in my head now lives in a system.

The second is accountability. Someone still approves the budget, signs off on the risk, and answers for it when something goes wrong. A better model doesn't change who that person is.

Coordination is about to shrink fast. Accountability becomes the whole game.

What to do about it

Think about it like a kitchen. The models are the ingredients, and everyone can buy the same ones now. The harness is the kitchen itself, and the best one is available to anyone. Making your own oven, pots, and utensils puts you back on the slow walk. Great restaurants win on what happens inside the kitchen.

Three things are yours to own.

Rework the workflow. Decide who works which station. In my workshops, teams take one workflow and break it into individual tasks. Then they sort every task. Keep it human, augment it, amplify it, or abandon it. I ask one question the whole time. What could AI own here? Agents can take the prep and the line. Tasting stays human, because someone has to answer for what leaves the kitchen.

Define the skills that matter. Skills are the recipes. They're how your company writes a brief, runs a review, builds a first draft. Right now most of those recipes live in a few people's heads, and every cook improvises. Write down the handful of dishes you serve every week. The people already on the fast walk have recipes of their own. Make those official, and give them time to teach them.

Centralize and own your data. That's the pantry. Client history, past work, decisions, what worked and what didn't. If it's scattered across a dozen systems, no recipe comes out the same way twice. Put it in one place you control, in formats no single vendor owns. When a better kitchen shows up next quarter, you carry the recipes and the pantry with you.

Small teams built on AI are already selling into companies that haven't changed. The companies that catch up to their own people first win the next wave of this.