What it’s like to work at an AI-native company

Field notes from a millennial raised in pre-AI tech

Working at Lovable hits different.

I’ve worked and/or advised a lot of tech companies at this point (I think I’m becoming a pipeline for Bending Spoons acquisitions). Big ones, small ones, public ones, hypergrowth ones, slow ones, and everything in between.

But I’m also very much a millennial of the previous tech era. Most of my instincts about how companies are supposed to work were formed before AI: managers manage, people have clearly defined jobs, important decisions happen in meetings, information travels through org charts, and becoming more senior generally means managing more people.

And after more than a year at Lovable, I keep running into things that violate those instincts.

This is the first company I’ve worked at that I would describe as AI-native. And to be clear, I have exactly one data point here, so this is not a manifesto for how AI-native companies should work. I’m not even sure which parts are AI-native, which parts are Swedish, and which parts are just Lovable being Lovable.

But something is definitely different.

And I don’t mean ‘we use AI a lot.’ Most of the AI-at-work conversation is still pretty tactical: writing emails faster, generating code, summarizing meetings, making slides, researching customers, automating support. Those are productivity improvements. And when companies just slap them onto the same old organizational model, they still land kind of flat.

What’s been really strange is seeing the org itself change based on the assumption that AI is part of everything: what jobs look like, how people work together, who makes decisions, how information flows, and even what it means to be good at your job.

So consider these field notes from someone whose operating system was installed before AI. Some of it is awesome. Some of it is hard. And some of it still feels… weird.

---

Sponsored by Chargebee: I’ve said this before. I’ll say it again. Your monetization model isn’t the right one. And that’s true for every single AI company.

You need to rapidly test and evolve your pricing in lockstep with how your product is actually used. Sometimes that means offering top-up credits for bursty usage on top of subscriptions. Sometimes that means pricing outcomes in addition to volume.
To do this at the required pace and agility, you will need billing and monetization infrastructure purpose-built for this era. Chargebee is just that. Learn more →

---

Stuff I love

We have no titles.

Yeah, I advocated for this from the jump. We have ICs, Leads, and Heads, but titles don’t get more granular than that.

This was almost a necessity. We were growing so fast that we didn’t want to end up in a situation where an awesome Director today is still just as awesome next year, except the company has grown 10x and their scope now looks more like what another company might call a Manager. So now we need to have an awkward conversation about changing their title. It’s not a demotion. But it is. Except it isn’t, because the company just got bigger. But it definitely feels like one. Why would we do that to people? It would be demotivation central.

What I didn’t fully appreciate before is how much organizational machinery and politics are attached to titles. They’re not just labels for your job. They’re status, access, information, and even permission.

Depending on where you sit in your own company’s org chart, you may not realize how literal this can be. You get promoted and suddenly you’re added to different meetings, invited into different Slack channels, copied on different decisions, and given access to information you apparently weren’t senior enough to know 24 hours earlier.

Lovable’s org structure is almost uncomfortably flat by comparison. People opt themselves into conversations. And opt themselves out. Almost all Slack channels are public unless there’s a good reason for them not to be. Information doesn’t travel exclusively up and down a management chain. Try removing some of those traditional layers - you’ll start questioning why those layers even existed.

This is the part I still haven’t entirely figured out: what does all of this do to management?

A lot of the things I grew up thinking managers were supposed to do - distribute information, coordinate work, approve decisions, assign ownership - are things we seem to need less of here. Managers obviously still exist (we call them Leads and Heads) but the job seems less about controlling the flow of work and more about creating context, raising the quality bar, developing people, and making sure a bunch of highly autonomous humans are running (very very fast) in roughly the same direction.

I suspect management is one of the jobs AI-native companies are going to have to rethink the most. Again: Not that managers will disappear, but a surprising amount of what we historically called ‘management’ may have actually been the work required to make information and decisions travel through a hierarchy that is no longer needed.

Senior people are happy as ICs.

I’ve talked about this a lot at this point and it’s becoming more and more true. There are a surprising number of very senior people who are in IC roles at Lovable. Like, they had dozens of people under them in previous roles, but now they’re just crushing projects and tasks directly. We’re living in the High-Impact IC (HI-C!) era, yall.

I’ve definitely seen that at other companies, but it always had a bit of a cloud over it. Here, staying an IC or going back to IC does not seem to carry the same implication that you somehow stalled out halfway through your career.

Some of this comes from AI capabilities expanding what any one individual can do: One highly capable person can now research, build, analyze, write, prototype, launch, and iterate in ways that used to require a much larger group.

But some of this is just the natural rebalancing of a dumb system that constantly pushed people to transition into management, even if they didn’t want to. Even when their new role would completely remove their ability to deliver the value that made them stand out in the first place.

This one makes so much sense that I don’t think it’s going to stay within just AI-native companies. My own research study on this found that 71% of people (including managers) want to spend at least half their time on IC work. So I think it’s just a matter of time.

Everyone is very suspicious of meetings.

I’m not mad about this at all. I spend less than 20% of my time in meetings, which would have been straight-up unimaginable in some of the companies I worked at before. And it’s obviously not because we’re getting less done or coordinating less across the team. Meetings are just not the default way to share information.

Some of this is because our AI tools make it much easier to share context. You don’t actually need to be in the meeting if you have the ability to pull up the company’s shared knowledge base and see what happened, including why the decision was made.

But the biggest reason is that agency and autonomy are highly encouraged across the company. You don’t need a permission slip to move forward, or to send every decision traveling up and down the org for approval. We want to move fast, and when velocity is high and you can correct quickly, the cost of getting something wrong starts to really drop. In fact, it’s often lower than the cost of endless meetings trying to make the perfect decision (which doesn’t even exist). Yes, we get some things wrong. But that’s learning. It creates valuable context, we adjust quickly, and we move on.

Stuff that’s hard (for old-tech me)

The pace is insane.

I often say that I’ve *survived* at Lovable for over a year. I only half joke. Everyone in tech and especially in AI gets very hyped about the shipping velocity and how quickly innovation happens. But it is exhausting.

Not only is there this massive land-grab to ride the AI wave and pick up as many users as possible before they get locked into other platforms, everything can change overnight. If the big labs drop a new model, maybe the core capabilities of our product are very different. Or it allows competitors to do stuff that used to be impossible.

So much of Lovable has been crafted to create a super dynamic work environment that allows us to adapt to all of this, but the team inevitably takes a lot of the chaos straight to the face.

I took a 1-week vacation a few months ago and when I came back I literally felt like I was re-onboarding into a totally new company. Almost everything had changed, so I had to sprint to catch up… to everyone else who is also sprinting full speed.

Roles and responsibilities are blurry.

There are barely any boundaries between jobs at Lovable. People jump into things that are not technically theirs, start projects, build things, fix things, occasionally duplicate work, and step on each other’s toes. I’ve had so many moments where I thought I owned something only to discover that someone else had already took in a different direction. And I’ve certainly done it too, where I wandered into another team’s area because I saw a problem and wanted to take care of it.

Actually, this has been my MO for a lot of my career, so you’d think that I’d be 100% stoked about this.

At lots of companies (especially big companies), the role definitions and boundary-drawing are A Problem. And not even just the territorial pissing contests. Sometimes it’s just a ‘not my problem’ default thing: When you get your wrist slapped for fixing a bug that ‘belongs’ to some other team, well, you’re not going to go out of your way to fix bugs anymore, are you?

So part of me absolutely loves that everyone can kind of do anything at Lovable. But it can also be very draining. There are reasons why companies spend a lot of time defining ownership. Someone has to be accountable, duplicated work is annoying, and companies become unmanageable if nobody knows who is responsible for anything.

I’m glad the Lovable culture is encouraging and open, and that people don’t hold grudges, but I’ve definitely caught myself feeling territorial when someone is suddenly all up in my business and seemingly doing my job. But on the other hand, like the old saying goes, two minds are better than one. So I’ve had to learn to loosen my grip a little and accept that everyone is kind of doing my job, while I’m kind of doing everyone else’s.

I don’t actually know where the line is yet. There are definitely moments where ‘high agency’ is just a flattering way of describing duplicated work, unclear accountability, or someone stepping on somebody else’s toes. Sometimes the chaos produces a better outcome. Sometimes it’s just chaos.

We’re in a constant state of reorg.

At Lovable, the only constant is change. Teams change all the time. Responsibilities move, people shift areas, new teams appear, old ones disappear, and things that made perfect sense three months ago suddenly don’t anymore. A reorg isn’t really treated as an event here. It’s closer to a normal operating condition.

I’ve worked at companies where constantly changing the plan was mostly a symptom of confused leadership, and others where simply saying the word ‘reorg’ could blow up an entire quarter. Leadership would announce a change and actual work would grind to a halt while everyone tried to translate what it meant for their project, their role, their team, and possibly their entire life plan.

What feels different here is the assumption that the structure will change. The context around the company is changing insanely fast, and because we’re growing so quickly, the operational stages companies might normally move through over years are getting compressed into months. The structure that was right for the company six months ago probably shouldn’t be the structure that’s right today. So a reorg is less likely to be interpreted as ‘something went wrong’ and more as ‘we learned something new.’

That doesn’t make it easy. Constant change still raises the difficulty level of the work. Going fast is hard. Going fast when the stakes are high is stress city. Going fast with high stakes while your team, scope, priorities, and sometimes the entire foundation underneath you are changing is… definitely not for everybody. The unusual part isn’t that change is painless here. It’s that nobody seems to expect it to stop.

Stuff that’s weird

There are tons of cowboys.

Lovable has a lot of what I would call cowboys. These are the high-agency generalist types who see something that needs to happen and just yeehaw, go chase after it. This is the type who is not waiting for someone else to build them a plan, and maybe they’d just kind of ignore the plan if it were there, anyway. They think ambiguity is awesome and don’t have a problem when they have to get scrappy to piece something together at the last minute.

Lovable has maintained a really high ratio of cowboys throughout our scaling stages because… we kind of have to be. We’re on that product-market fit treadmill and innovation won’t stop (can’t stop).

This is… mostly great. I like to think of myself as a cowboy (this is not why I moved to Nashville, but it does fit kind of nicely). And this is very common in startups. But it’s not so common in scale-ups with over 500M+ in ARR. Usually, this is the realm of farmers - putting down roots and building the systems to grow things in predictable, plannable cycles. As our systems are growing and we need more reliable infrastructure, we need to figure out how to strike this balance.

People are automating themselves out of their jobs. And are excited about it.

One of the stranger things about Lovable is how normal it is for someone to join and immediately start asking, ‘How much of my own job can I automate with AI?’ And they’re actually happy about it. People seem less likely to be trying to protect a particular task just because that task happens to sit inside their current job description.

This seems pretty backwards for those of us who came up in a world where we were taught that job security comes from being irreplaceable. Most of us have been trained to build careers by accumulating responsibilities and becoming very good at performing them. You become the person who owns the process, knows how the system works, produces the report, answers the questions, or runs the workflow.

So this is another area where it’s been good to actually see it play out. ‘Automate your own job’ sounds terrifying if the implied second half of the sentence is ‘...and then we no longer need you.’ It feels completely different if the second half is ‘...so you can go figure out what your job should become next.’

I’m guessing that this fear is much more heavy in traditional orgs where AI capabilities are being presented as an existential threat to work. But at a company where people are expected to be getting the most out of this new tech, proving that you can automate a complex or difficult set of tasks just shows how valuable you are. There’s always a lot more to do on the other side anyway. Our team is definitely growing, not shrinking (we will be pushing 500 people by the end of the year… I started when we were just 20. WHUT.)

This kind of environment makes me grateful for my Growth background, where I’m quite used to jumping in, fixing something, handing it off to someone else, and then going to chase down some new opportunity to create value. But I know that’s not the norm.

AI agents are actual coworkers… with parents.

Yeah, this one has been weird.

We have agents all over the company. Marketing agents, analytics agents, codebase agents, a general Company Brain, and more and more small tools that people have built around their own work (you should see our swarm of GitHub agents that attack every PR!). At this point, I interact with AI agents almost as much as I interact with humans during a normal workday. That’s kinda crazy, right??

It’s also interesting how we manage them: Lovable does this through a role we call an ‘agent parent.’ This is the human who’s responsible for the quality of the agent’s output. They’re teaching it, updating it, correcting it when it starts giving bad answers, and generally making sure it stays useful. Very often, the parent is the person who previously held most of that knowledge in their own head and answered the same questions repeatedly for the rest of the company.

So, yeah, I’ll admit: it feels slightly dystopian.

But seeing how it plays out has been a lot less dramatic. If you’re the person who has answered the same question 50 times, you don’t want to be the one answering it a 51st time. Actually, at that point, you’ve probably already documented the answers and context somewhere, but the org just doesn’t have a good way to surface it.

From what I can tell, this structure makes people’s perspective and expertise even more valuable, because literally everyone in the company can benefit from it. If everyone depends on this info, it better be right.

The judgement check is just ‘Is it lovable?’

One of the phrases you hear constantly inside the company is some version of, ‘Is this lovable?’ Not exactly the basis for clean acceptance criteria. There is no checkbox for lovable. Which (are you seeing the trend here?) can feel a bit chaotic.

My experience has been that this means you have to rely on judgment a lot more. Everyone sort of defines lovability according to their own taste, which means you’ve gotta have people with good taste.

And ‘lovable’ is definitely a bar above ‘it works.’ Something can function perfectly, solve the problem it was supposed to solve, and still not feel lovable. The question forces you past functional correctness into something much harder to define: Is it delightful? Is it intuitive? Would you be excited to use it? Would you be proud to ship it? There’s no metric that neatly answers those questions, which is exactly why judgment matters so much.

It seems like everyone in tech is obsessed with the idea of taste right now, so I won’t go over it again, but with everyone producing more of everything, I can already see that companies are going to depend more and more on people who have this magical ‘judgment.’

This is NOT an all-encompassing recommendation

The weird paradox is that the more AI removes the mechanics of work, the more the company seems to depend on human judgment. When information is everywhere, execution is cheap, and everyone has enormous leverage, knowing what to do, when to interfere, what good looks like, and when something is actually done becomes disproportionately important.

So I don’t think you can take one of these things, drop it into another company, and expect the same result. ‘Let’s have fewer meetings!’ sounds great until nobody has context. ‘Let’s give everyone autonomy!’ sounds great until nobody knows what good looks like. ‘Let’s remove titles!’ sounds great until you realize that your entire status and decision-making system still totally depends on them. Whatever this is, it seems to work as a system.

And I’m definitely not saying other companies should work this way. I have exactly one data point here, so consider these field notes, not a scientific pattern.

That’s partly why I wanted to write it down now. I’m old enough to remember when the way we worked before all of this felt completely normal. Managers had the context. Jobs had boundaries. Seniority meant managing more people. Important decisions happened in meetings. Software was something most employees couldn’t just make for themselves. None of that seemed particularly strange at the time. But have a feeling some of it is about to (and maybe already is).

Oh and hey, we are hiring at Lovable. Come experience all of this together with me :)

---

Edited by Jonathan Yagel.

---