When I talk about things like High-impact ICs and speed as a moat, everyone gets excited. But you don’t get those through ‘AI transformation initiatives.’ (Yuck)

Sure, having the right AI tools and skills is part of the equation. But even the best employees with all the most magical tools will get trapped… if your org structure and culture don’t adjust.

Right now, many companies already have the technology they need to go much faster. The blocker is company systems that are mostly designed to prevent things from happening. The power is centralized and all the team members are treated like a risk vector. Exhausting approval cycles, super tight boundaries on roles, unbreakable title-based hierarchies, and a whole tier of middle managers whose main job is to keep everyone in line.

All of this pretty much screams at employees, ‘We don’t trust you, stay in your lane.’

And it should be obvious: Rapid innovation cannot happen in that kind of environment. Super high-impact individuals and high-velocity teams will shrivel up and die.

Everyone’s talking about AI agents, but what you really need is employees with agency. And if you don’t figure out how to give them the access and autonomy to make decisions on behalf of your organization, all of your AI innovation efforts are screwed.

Sponsored by Optimizely’s Test + Learn Event (with me!!): On June 17, I’ll be speaking at this free, virtual event about how testing has shifted:
- Acquisition testing → Monetization
- Tiny conversion lifts → Bigger strategic swings
- “Growth at any cost” → Growth with margin discipline
- Endless optimization → Faster learning cycles
→ Register here.

Sponsored by Amplitude: Build better products with our AI analytics platform. With powerful AI Agents embedded across our platform, teams can analyze, test, and optimize user experiences faster than ever. → Learn more

‘Protecting’ employees by limiting access and autonomy

Most modern companies are pretty much built around a command-and-control, power-at-the-top model. Each employee is treated like an assembly-line worker with one specific function. ‘Do your job, don’t ask questions, we’ll do the thinking for you.’ The goal is to bring order to the chaos of big, broad organizations with lots of moving parts. If everyone thinks for themselves… who knows what could happen!

There are a couple of obvious ways that this gets applied:

1 - Access is gated by titles

One of the big reasons everyone is striving for bigger titles and team management positions is… that’s where you’ll be heard. That’s where you can actually get the required information. That’s where you get invited to the right meetings and channels.

Companies say they want high-impact ICs, then hide the information required to have impact behind manager titles.

Literally. When you get a manager promotion, you’re in a separate Slack channel where information is being shared with you that you had no access to, before.

We make assumptions that employees are not to be trusted unless they’ve earned their title - even if that was external, earned at another company.

2 - Only the bosses get to make decisions

Decisions are suuuper costly in organizations. It takes forever to make them and it’s almost impossible to reverse them, so you better get them right. Which means that everything is routed to the Very Important People who can make those decisions.

So decisions are slow because they’re expensive. They’re expensive because they’re slow.

In some ways, the whole system is trying to ‘protect’ the employees - because to be fair, the employees don’t have all the information that the managers do, and it would be easy to make the wrong call without the appropriate context.

The virtuous cycle of faster, cheaper decisions

The AI-native orgs strip out the status theater around the work. Anthropic is a good example: the visible structure is basically “member of technical staff” or “member of non-technical staff.” Not because titles are evil, but because titles often become permission structures. They decide who gets context, who gets heard, and who gets to make decisions. And that is exactly the kind of drag AI-native companies are trying to avoid. Which leads to:

A - Faster decisions.

If you ungate access to information and encourage autonomy, suddenly… the context doesn’t have to pass up-and-down the chain for every little thing.

What’s funny is, lots of companies are realizing how important it is for their AI programs to have sufficient context. ‘If the AI doesn’t have context, it can’t do a good job!’

Well, guess what? Employees also need context to make the right decisions and move forward quickly. Otherwise (just like an AI), they’ll make the wrong decisions or get stuck.

B - Cheaper decisions.

And when you make adjustments on the same day (rather than waiting until the next weekly/monthly/quarterly ‘strategy’ meeting or exec offsite), the cost of a wrong decision will also fall off. In the previous world, the cost of a bad choice was astronomical. It would take so much time to get out of it. But if everybody can run at a really quick pace, it’s okay to make a wrong turn! You can go a bit down that route, turn around, and be back on the right course, faster than other teams even start moving.

Management Bloat

In the traditional org structure, middle management is a feature, not a bug.

The primary job for this role is to travel decisions and information up and down the chain. Up and down, down and up, over and over. It’s a big game of telephone.

This is intentional. Companies don’t trust the junior employees with sensitive information or believe that they even need it.

Now, I will say that managers also do other things… but the structural need for those is being reduced, too.

• Manage cross-functional gridlock and unblocking: ‘Who’s going to do what’ and ‘Where can my team’s priorities fit into your team’s calendar’ and ‘When are we going to find time to support your initiatives,’ etc. But fewer teams and more centralized information make this less necessary.

• Career coaching and guidance: Always important, always necessary… but is the manager really the best career coach for all of their reports? From a personality fit and skills fit perspective, maybe not. This can be more democratized, too.

Companies need to realize that pushing AI tooling into their teams is not going to automatically fix the velocity problems. Yes, having those increased capabilities can empower them, but if they’re still blocked off by bureaucracy and their own limited context, they’ll never move fast.

How do you make this shift?

Okay, company leader. What do you do about this? You’re not going to fix it with a new AI tool, an all-hands announcement, or a slide that says “move faster.” This stuff is too baked into how companies make decisions.

And to be 100% honest - I’m not sure how you switch it. I can speak to the differences I’ve seen in how Lovable operates vs. lots of other companies I’ve worked with, but I don’t know how to make the shift.

But here’s my hypothesis: Don’t start by trying to rewrite the whole org. Prototype it. Try it out with some R&D or innovation teams to see what they can do. Give them aggressive targets and a flat structure, so that their full team has the autonomy and access to actually get things done. Put your sharpest, most high-agency people in the group to start and see how they do.

And be honest: you may need some new people in that group. Not because the current team is bad, but because organizational muscle memory is real. If people have spent years being trained to wait for permission, it’s hard to suddenly tell them, “Actually, now you’re allowed to act.”

The weight of responsibility

Employees need to listen up to this, too. Because of the design of many organizations, lots of employees haven’t experienced the downside of autonomy. Yeah, it sucks to just be treated like an assembly-line robot, being told exactly what to do… but that also spares you from the legitimately terrifying possibility of something going wrong and it really, truly being your fault.

At the end of the day, that’s why leaders get paid more: Accountability. They’re the one who is responsible for a thing getting done. If there’s no one telling you what to do, you have to figure it out for yourself. And if your launch flops or conversion craters or something else vital to the business falls apart… that’s on you.

Fortunately, with the right culture and a high-velocity environment, everything is figure-out-able. But as more companies figure this out and start handing out greater context and autonomy, people will need to consider whether they actually want it. High-impact roles come with status and comp, but there’s always a cost.

Agency > Agents

If your company is anywhere near tech, everyone is probably talking about agents. Are AI agents going to replace all employees?

I don’t think so.

Although I definitely believe that everyone’s current jobs are changing faster than we want to admit, it’s not agents we need to worry about.

At Lovable, we use AI agents internally to make departmental context easier for anyone to access. Each agent has a “parent,” usually the person with the deepest context in that area, who is responsible for keeping the agent accurate and up to date. That matters because it removes one of the biggest excuses for centralized decision-making: “People don’t have enough context.”

But the agent is not the unlock. The unlock is what happens when employees can actually act on that context. That’s the irony of the “AI agents will replace employees” conversation. Agents don’t have agency. They wait to be told what to do. High-agency employees do the opposite: they find the signal, make the call, and push the work forward.

So yes, build agents. Assign them parents. Keep them useful. But don’t confuse the tool with the operating model. The goal is not more agents. It’s more agency.

Trust-based growth starts with Trust-based teams.

A few months ago, I told you that Growth is now a trust problem. Software companies in particular are getting disrupted from multiple angles and the way to survive is to build a connection point with your current and prospective customers.

But that trust isn’t just something you have to build outside of your company. In fact, I don’t really know if it can exist outside of your company if it doesn’t exist inside of the company first. All of the techniques that make external trust-building possible start with you trusting your own employees.

Build in public? You’ve got to let people talk about what they’re actually doing. Sharing legit failures is scary, but that’s how you build credibility.

React quickly to user needs? You can’t have a centrally-managed, long-term roadmap that pushes urgent requests out by a quarter or two.

The old way of doing things was all built around maintaining a polished, fault-free exterior. Now, people want to see some humanity. Unique quirks and imperfections that show off an actual person on the other side of the screen. And that can’t be faked from the top. You’ve got to trust your employees enough to do that - and give them the access and autonomy to make it happen.

You can still adjust to the right internal and external approaches to make all of this work, but you’ve got to make that call, and quick.

How fast can your company make this decision?

Edited by Jonathan Yagel.