Key Takeaways

  • The Grok Bot core team built and shipped the product in seven weeks, manually onboarding 200 to 300 users one by one.
  • Sitting on live calls where cloud virtual machines failed to spin up for 20 minutes created visceral urgency to fix onboarding friction within 24 hours.
  • Refusing to prompt or guide users revealed an organic user pattern: by week two, users began promoting a favorite bot to "Chief of Staff" to coordinate the rest.
  • Onboarding diverse non-tech users, including small business owners, forced the team out of the Silicon Valley bubble to build software that mainstream customers can actually use.

The 20-Minute Agony of Live Onboarding

When Roman Ugarte and his team built Grok Bot at SpaceXAI, they refused to hide behind self-serve sign-up flows. They spent live onboarding calls with 200 to 300 people, ranging from skeptical tech influencers to non-technical small business owners.

Sitting on a call while your product stalls is brutal. Ugarte explained why the core team had to experience that friction directly: “I think it was important for the core team to be in the room for those and to just sit on a call for 20 minutes when the computer isn't spinning up or when someone's in onboarding and they're just incredibly confused. So that immediately after you're like that can never happen again. we need to solve this tomorrow because tomorrow I'm onboarding this person and it needs to go better.”

When an automated analytics dashboard flags a drop-off rate, engineering teams schedule a sprint review next month. When an engineer sits in silence for 20 minutes while a real customer stares at a dead loading screen, the bug gets fixed that night. Manual onboarding turns cold metrics into acute social embarrassment.

Discovering Emergent Workflows Without Leading the Witness

When founders onboard users manually, their default instinct is to lecture. They demo the best features, explain the architecture, and tell the user exactly how to operate the software. Ugarte took the opposite approach. The Grok Bot team deliberately gave zero instructions on how to structure multiple AI agents.

By watching in silence without leading the witness, the team spotted emergent behaviors they never planned in a product roadmap. “Around the end of week two, we started to see these messages internally in Slack of people promoting one of their bots who is a bit of a standout performer,” Ugarte recalled. Users were “promoting that to their chief of staff. And then they would actually mostly talk to their chief of staff. And the chief of staff would fan out all of these tasks to the other bots and would kind of manage the team.”

Because the team stayed quiet, users invented an organizational hierarchy for agentic workflows on their own. Had the team forced a rigid multi-agent framework during onboarding, they would have blinded themselves to how real humans naturally delegate work.

Breaking Out of the Silicon Valley Bubble

AI founders often build for power users who already understand context windows, prompt chaining, and compute limits. Ugarte intentionally onboarded mainstream operators who knew nothing about agent architecture.

“We absolutely live in this kind of Silicon Valley AI bubble,” Ugarte noted. “Which I think is a useful place to be to kind of push the frontier and push the future of how these products are evolving. But we need to actively get out of that because I think a product like this has the chance of really being the way that the mainstream user and the mainstream kind of business customer can interact with AI in a way that's useful.”

If an AI tool requires a tutorial on prompt engineering to be useful, it is not a mainstream product. Manual onboarding with non-technical users exposes every hidden assumption baked into an interface.

What to Do With This

Block two hours tomorrow to run 1-on-1 onboarding calls with three new signups, including at least one person who does not work in tech. Keep your microphone muted for the first 15 minutes while they attempt to complete your core setup flow, and write down every step where they hesitate or click the wrong button. Fix the single worst point of friction before scheduling your next cohort.