The most AI-forward company I've worked at automates boring stuff first.
I joined n8n three months ago expecting a somewhat chaotic scaleup with light process and a lot of "we'll automate that later." Instead, on my second week, I watched an agent assign PR reviewers by checking Linear tickets, Slack availability, and who was out of office in HiBob. No demo or tucked-away knowledge. It just works and it's one less thing someone had to remember.
I'd spent two years running my own startup, convinced I had AI figured out. I was wrong about almost all of it.
What I got wrong
When you're the only engineer in a company of fewer than ten people, you can't see what a 100+ person organization looks like when it takes automation seriously. I thought being AI-native meant engineers with good tools. Copilot, GPT wrappers, maybe a chatbot bolted onto the product. I also assumed scaleups had less process than big companies. Fewer meetings, lighter systems, move fast.
Both assumptions were wrong. n8n has process, but the process is automated. Maturity showed up not as bureaucracy but as agents and workflows handling the coordination that would otherwise eat everyone's day.
I'll admit something else: before I joined, I worried that workflow automation was becoming a dying category. Tools like Claude Code and similar agent frameworks seemed like they'd replace the need for visual workflow builders entirely. That fear lasted about a week.
Agents that matter look boring on paper
My second week, I opened a pull request. I didn't pick reviewers. I didn't check the team calendar. I didn't ping anyone on Slack.
An n8n workflow picked up the PR, read the Linear ticket to understand which team owned the work, checked HiBob for who was on leave, and assigned two available reviewers. Then it posted a summary in Slack with the PR link and what the change was about.
The whole thing took a few seconds. Without it, someone would have scanned the team page, cross-referenced the calendar, guessed who had capacity, and pinged them manually. It's a five-minute task. But it happens dozens of times a day across the engineering team, and five minutes times dozens adds up to hours of coordination that nobody had to think about.
This isn't the kind of AI that makes conference talks. Nobody is going to write a blog post about "we automated PR assignment." But it's the kind that actually changes how a team operates. The flashy AI features get the demos. The boring agents get the time back.
Everyone automates, not just engineers
At my startup, automation was an engineering concern. If something needed automating, I built it. That was the mental model: engineers automate, everyone else uses what engineers build.
At n8n, the sales team tracks their pipeline through workflows. Marketing runs content distribution automations. Someone built a weather reporting bot for the office Slack. During the World Cup, there was an agent posting match updates. These aren't side projects or hackathon demos. They run in production, every day.
The company has a name for this: Hunderfutter Zeit. Dogfooding. Every quarter, teams identify tasks that eat their time and try to automate them, using n8n's own product. Not because someone mandated it, but because people here use and care about what they build. When your product is a workflow automation tool and your sales team is building their own workflows, the feedback loop between users and builders gets very short.
I'd never seen non-technical people build their own automations unprompted. At most companies, "we use AI" means someone bought a Notion AI subscription or the marketing team uses ChatGPT for first drafts. At n8n, "we use AI" means the recruiter has a workflow that screens candidate applications, the sales rep has one that updates the CRM after every call, and the office manager has one that orders supplies when inventory drops below a threshold.
That's a different thing entirely. It's not top-down tooling. It's a culture where automation is the default response to repetitive work.
How engineering compounds the culture
The engineering side takes this further. We have custom agent skills that standardize how developers work: a create-pr skill that formats pull requests to pass CI validation, a diagnose skill for systematic bug hunting, a spec-driven development workflow that keeps implementation synced with design documents. Community-built skills like thermo-nuclear code review (yes, that's the actual name) and grill-me for stress-testing designs before writing code.
These aren't "AI magic." They're structured patterns that reduce variance in engineering output. A new hire follows the same PR process as a five-year veteran, because the skill handles the formatting and the CI checks. An engineer debugging a production issue follows the same reproduce-minimize-hypothesize-instrument loop, because the diagnose skill enforces it.
And because the engineering team lives inside automated workflows daily, building the product becomes a different experience. One of the interesting new features is Instance AI, n8n's built-in autonomous agent, and the AI Workflow Builder. When the tool you build all day is the same tool you use to automate your own job, you notice friction that might stay hidden if you were building for others.
What "AI-native" actually means
Before n8n, my definition of AI-native was: a company that ships AI features and uses AI tools internally. That's wrong, or at least incomplete.
Most companies buy AI tools. They give engineers Copilot, they add a chatbot to the product, they announce an "AI strategy" in the quarterly all-hands. That's AI-adjacent. It's adopting AI the way companies adopted cloud computing: as infrastructure someone else manages, bolted on from the outside.
AI-native is different. Automation is the default operating mode, not a project. People reach for the product when they have a repetitive problem, even if they're not customers. Agents handle unglamorous coordination work like PR assignment and pipeline tracking, and the demo-worthy features come second.
Three months in, if I had to write a checklist, it would be short. An AI-native company automates its own operations first. Everyone, not just engineers, builds and maintains automations. The product team dogfoods aggressively and treats internal pain as signal. And agents do the boring work, because that's where the hours are.
I couldn't have learned any of this as a solo founder with good tools. The culture is the thing. You can't simulate it.
Top comments (0)