Being AI-Native Is About Behavior, Not Budget

Your Company Has an AI Level & It's Probably Lower Than You Think.

Every company knows that they need to adopt AI across their operations, and every startup wants to convince others that they are already AI native. What this really looks like in practice is more nebulous and everyone is still trying to figure out what works best.

Beeing deep in the startup ecosystem and looking at how different teams are approaching this challenge I observe several themes. Most companies that were around before the recent AI acceleration are afraid of being left behind and know they need to give their employees a budget for tokens — but at the same time they they hear horror stories and are afraid of spending too much on tokens. While there is no right answer on what budget to set, I see those fears drive behavior that sorts a company into one of three levels.

Luckily for you and your organsation, these levels are not primarily defined by raw AI spend, but way more by what a company believes AI is, how much it trusts its people, and how it has re-organized the way that they approach and complete work.

Most people reading this will start out thinking that they are already at level 3, but this is a position only constant vigilance will help you keep. Similarly, if you feel like you are at level 2, don’t rest on your achievements, it’s easier to slip into a level 1 than to do the work necessary to become a level 3 AI company. The most important take-away is that these levels are not static and being at level 2 is the most dangerous place to be.

Level 3

AI as an entirely new way of working

You know you're at one of these companies because nobody talks about token spend. You have access to the best models, effectively without limits, and the only pressure from leadership is whether you're thinking big enough. Trust comes first, control second. The most ambitious people are never tied down to the pace of the lowest performers or the occasional token waster.

These companies understand that AI doesn't just make the old work faster. It makes entirely new work possible. Work that would never have been attempted before, because the price and complexity made it unthinkable.

The check to see if a company is operating at this level: look at their process, or the lack thereof. Small teams or single people iterate with true authority to ship. If something takes off, practically unlimited resources follow. Nobody has to make a business case before the artifact even exists.

When production is no longer measured in quarters but in days, the math has changed. Exploring a product idea used to cost a full team a quarter; a company could afford three or four bets a year, and at least one had to land. Now a single person can create a crude but working product in a week. The correct response is not to run the same bets cheaper. It's to run more of them and let reality vote.

And yes, some of that freedom will be wasted. Level 3 companies accept this, because they define waste precisely: spend that produces neither insight nor business direction. Failed experiments that taught something are the whole point. Accountability comes after, and it's proportional. Run by default. Only if your spend crosses a threshold without you ever having shared any progress do you get questions. Not a leash, but a conversation and better practices.

Level 2

AI as faster work

Here, everyone has access to the very best models, but leadership spends more energy reminding people to reduce spend and route to cheaper models than motivating ambition or new ways of thinking. Same work, same process, same ideas, just more code in less time. Sound familiar?

It feels responsible, logical, safe. It's actually the most dangerous place to be.

This is where the spending trap lives. Instead of punishing only the waste, everyone gets rationed pre-emptively. That prices the entire company at the level of its worst actor: the few people burning tokens on nothing become the ceiling for the most ambitious builders. Cost-reduction thinking has no discernment for output or potential. It caps everyone equally, and the best people feel it first.

From there, the slide is mechanical. When employees only use AI to do the same work faster, the visible output is efficiency. Efficiency gets measured. What gets measured gets optimized. What gets optimized gets budgeted. Within a few quarters, finance owns your AI strategy, and every conversation about AI is a conversation about cost. Thinking of AI as cost is what a Level 1 company does. And you got there without ever making a single big decision, just a thousand small routing choices that all seemed sensible in isolation.

The work that helps you escape this gravity probably already exists at your company. It's just not being encouraged or rewarded yet. The small tool your coworker built for themselves on a weekend. The skill you wrote but never told anyone about. Problems that sat unsolved for two years because they never justified a team, quietly fixed in an afternoon. These are hard to measure as efficiency because without AI, they simply wouldn't exist. This work currently lives in the margins, without funding, and employees won't tell you about it before you ask. That's what the cost frame does. The new way of working is already inside the building, but it’s kept at hobby scale. Thinking bigger means dragging it into the open and making this the new way to work.

Level 2 is not a place to rest. Your next actions here decide whether your company rises or falls a level. Entropy comes for us all.

Level 1

AI as an expense

Getting access takes a ticket, a justification, and about a month. And in the end all you get is the cheap model? You and half your colleagues have quietly resorted to paying for your own subscription, and leadership happily turns a blind eye. These companies focus on the issues with AI and would rather protect their downside than optimize for any upside. Deep down, they resent having to keep up, and resent having to train their people for a new reality.

I won't spend much time here, because Level 1 leadership isn't reading this post anyway. Only their hungriest employees are. If that's you: you already know. The question isn't whether the ship is sinking. It's whether you're still on board when it does.

And if your company isn't at this level yet, that doesn't mean your company is safe. Watch out for the process I described above!

Three signs your company is on the right path

Level 3 can't be bought with a big token budget. Quite the contrary: it's built on structure and trust.

The number of your live experiments, not your token bill, tells you which level you're at. If the cost of a bet dropped 10x, the number of bets should have gone up.

The unit of ambition is now the individual. Instead of endless alignment meetings, one person with conviction and a model can produce what used to take a team.

Your company uses human judgment, not a formula. ROI is hard to measure. Is your company searching for insight, or trying things that couldn’t have been done before? Even when results are still far off, they should be judging the direction of the work.

Images generated with ChatGPT Image 2.0 and inspired by General Atomic, Weather Control Poster by Erik Nitsche