At a glance
WHAT IT’S REALLY ABOUT
Laurel’s GitHub-based Company OS scales AI workflows across functions
- Laurel codifies how every function works in a GitHub-based “Company OS,” organizing playbooks and “skill files” so teams can apply consistent, high-quality workflows rather than relying on a few AI power users.
- The OS is operationalized through just-in-time delivery inside existing tools like Slack (e.g., automated feature-request intake/triage) and through AI tools (Claude, agent builders) that can call the right skills at the right moment.
- Zhang proposes a 3-step path to build a Company OS: start with one tedious workflow automation, evolve into playbooks that separate human vs. automatable steps, then scale via agents/skills and a “mega-agent” routing layer.
- Laurel’s operating model shifts PMs (and even non-technical roles) into “product builders” who can ship end-to-end changes using agentic coding tools, while governance is maintained via transparency, rules of engagement, code review, and lightweight review channels.
- Talent and org design shift toward leaner teams of highly senior, AI-fluent “orchestrators,” assessed via screen-share interviews and guided by an AI maturity model spanning chat usage to shared apps and production shipping.
IDEAS WORTH REMEMBERING
5 ideasTreat AI adoption as an organizational distribution problem, not a tooling problem.
Laurel’s core issue is the “1% vs 99%” gap—power users tinker while most don’t know what to use when—so they encode best workflows as shared skills and deliver them inside daily work streams.
A Company OS starts small: automate one painful, repeatable workflow first.
Zhang’s step-one recommendation is to pick a tedious motion (emails, CRM updates, request intake) and remove back-and-forth via simple automation before attempting a full OS.
Playbooks only matter if they’re executable—convert them into skills and agents.
A 50+ page playbook won’t be followed reliably; Laurel audits which steps require humans vs can be automated, then builds agents/skills per step and surfaces them just-in-time.
Build a routing layer (“mega-agent”) so people don’t need to remember dozens of agents.
Instead of expecting teams to call the right micro-agent, Laurel uses a higher-level agent that interprets intent and dispatches to the right sub-agent, reducing adoption friction.
Delivery beats dashboards: put skills where work happens (Slack/email), not in a separate AI interface.
Even “low friction” tools like agent-builder UIs create a context switch; Laurel prioritizes Slack-based workflows and integrated skills to drive consistent usage.
WORDS WORTH SAVING
5 quotesYou got these people who are these 1% AI users. They're, uh, tinkering with their workflows. They're highly AI-pilled. And then you have the, you know, 90 to 99% of the rest of the organization who isn't sure what to use when.
— Jiaona Zhang
The big learning that we've had is how do you create a wrap, like a, like a mega agent, something like the f- like a, um, a go-to-market agent that can be called by the sales team at any point, by the success team at any point, and then that agent is able to route the ask, the, the need or the help to whatever one of these sub-agents that is actually useful.
— Jiaona Zhang
Transparency is everything.
— Jiaona Zhang
I really don't believe in this. I think a lot of, quote-unquote, "AI-native companies" are just like, "Roadmaps are gone. Plannings are gone. Everything is gone." Um, and what I say is, well, if everyone's running in different directions, even if you're running incredibly fast, you're not really gonna get anywhere.
— Jiaona Zhang
The fundamentals and the principles have never changed. In fact, they're even more important than ever before. But the tools and the way you operate and the way you, um, can blast through the bureaucracy and feel empowered, that's radically changed.
— Jiaona Zhang
High quality AI-generated summary created from speaker-labeled transcript.
