Recipes / Grok Bot use cases
Grok BotRoutineDev & code

A Grok Bot chief design engineer for teams

Felix Lee set up a Grok Bot as his "chief design engineer". Each week it turns analytics drop-offs into designs, files the idea in Linear and waits for his pick before anything ships. He says one update from two flaws it found lifted engagement 22% and traffic 10%, and that he rejects about 60% of what it makes.

From Felix Lee on X · Oct 2026

Setup timeHe says a problem statement becomes designs by the next morning, where it used to take about two weeks
CostYour Grok Bot plan, plus Amplitude, Linear and Paper; he didn't give costs
RunsWeekly: it picks your sharpest analytics drop-off and turns it into designs
NeedsGrok Bot, His Chief Design Engineer template, Amplitude analytics, Linear, Paper, the design tool, Your design system (it helps, he says)
Last updated2026-10-30

What the source did

  • He set up a Grok Bot titled "chief design engineer". It doesn't ship for him; he approves every design.
  • Last week, he says, it dug through their data, found two design flaws and shipped an update. Since then, by his numbers, platform engagement is up 22% and traffic up 10%.
  • Every week it turns drop-offs in the analytics into one design for him to approve, in Paper.
  • It studies the best apps to see what's working, and writes up every idea in Linear with the problem and the fix.
  • It tells him when a design is good enough, so he doesn't overpolish, and waits for his pick before anything ships.
  • In replies he says he has shipped 100 design experiments with it, and that it cuts a two-week process, from problem statement to design, to the next morning. One reply links an 11-minute demo of the workflow.
  • He also says: "actually we reject 60% of it, but still a massive improvement in execution speed." First designs are bad, and you iterate on them in Paper with your own design system. He uses Paper, not a Figma connector.
  • His template, linked above, files a Linear ticket with three annotated Paper designs, recommends one, and checks the result 14 days after it ships.
“It's not magic. some designs are meh and I tell it so.”— Felix Lee (@felixleezd), X

Summary written by Agent Cookbook from the original post. See the original for the full context. How we source recipes

Our notes

What it takes
Grok Bot, Felix Lee's Chief Design Engineer template, Amplitude, Linear and Paper, and ideally your own design system.
Where it breaks
First designs that are bad until you iterate on them in Paper, a high reject rate (he rejects about 60%), no design system to work from, and skipping the 14-day check on what shipped. The +22% engagement and +10% traffic are his own numbers.
Human checkpoint
It files the ticket and recommends one of three designs. Nothing goes to engineering or ships until you pick it, and a human designer makes the final call.
Tools involved
Grok Bot, His Chief Design Engineer template, Amplitude analytics, Linear, Paper, the design tool, Your design system (it helps, he says)

✓ Verified Source link checked and still live. Notes by Thomas, editor. How we verify

Get their template

Made and published by Felix Lee. Start here; it is their original, not ours.

Approval rule

🛡
The bot finds the problem, files the ticket and recommends one of three designs. Nothing goes to engineering or ships until you approve it, and final judgement stays with a human designer. Sending, buying, posting, and deleting always need a human.

Set it up

  1. Add Felix's Chief Design Engineer template from the link above.
  2. Connect Amplitude, Linear and Paper, and give it your design system.
  3. Each week, read its Linear ticket: the drop-off it picked, the problem and the proposed fix.
  4. Look at the three Paper designs and its pick. Reject the weak ones and say why; he tells it when a design is meh.
  5. Iterate on the one you like directly in Paper until it fits your design system.
  6. Approve it before it goes to engineering. Nothing moves until you do.
  7. Check the result 14 days after it ships against the drop-off it was meant to fix.

Common failures

⚠
  • Expecting finished designs. He says they're bad at first and still need human iteration, even with a design system.
  • Shipping too much of it. He rejects about 60%; the gain he describes is speed, not hit rate.
  • No design system. He says it works okay if you have one; without one, expect more rework.
  • Skipping the 14-day check, so you never learn whether the fix moved the number.

Variations

  • Use it as a research pass only: have it study the best apps for one screen and write the Linear ticket, and design it yourself.

Related setups: Weekly Grok Bot team: trend scan to spec to build → · A Grok Bot chief marketing officer → · Grok Bot night shift: bugs to draft PRs →

Get the 10 best agent setups. Free.

Plus new recipes every week. One email, no spam.

Unsubscribe anytime. Privacy

Related recipes

Same job: Dev & code

Also uses Grok Bot

More Grok Bot recipes

In these guides: More Grok Bot use cases →

Get their template ↗Read original ↗