Loading post…
IntercomRussell NorrisTutorial
How to build a Content Design Agent
Summary
An overloaded solo content designer built VERBI, an AI content design agent, to scale standardized work without replacing human judgment or filling products with generic copy. The agent runs in Glean, using a curated collection of Intercom’s style guide, tone of voice, product and monetization glossaries, messaging guides, research, and official websites as retrieval-augmented generation sources. The system prompt defines its role, use cases, tone, and component rules, while a Glean Indexing API connection makes Storybook’s Surge design-system guidance available as a live source. Additional instructions tell VERBI to challenge users, avoid guessing when sources lack an answer, and keep responses concise. After 700+ triggers, the reported benefits are consistent content rules, faster drafts and iteration, fewer repetitive questions and errors, and more time for higher-impact design review.
Context
The agent was created for a solo content designer supporting multiple teams with a continuously full pipeline of requests, including reviewing flows, rewriting screens, and naming features. The stated goal was to extend content-design capacity while preserving critical thinking and maintaining a system that could continue to evolve.
Approach / What changed
The implementation combines a curated Glean collection of approved content and product sources, task prompts, a detailed system prompt, and access to component-level guidance in the Surge design system. Storybook guidance was first scraped into an HTML file, then connected as a live source through Glean’s Indexing API. Additional prompt rules make the agent challenge assumptions, avoid unsupported guesses, and keep responses concise.
Takeaways
A curated collection of approved sources lets VERBI ground responses in Intercom’s style guide, tone of voice, glossaries, messaging guides, research, and official websites rather than treating all internal information as equally reliable.
Component-level guidance in Surge is central to the agent’s precision. Connecting Storybook through Glean’s Indexing API replaced a manually refreshed HTML workaround with a live data source.
The system prompt explicitly counters common agent quirks: VERBI should challenge users instead of agreeing by default, say it does not know when sources lack an answer, and avoid unnecessary verbosity.
FAQ
Why did the agent need access to the design system, not just a system prompt?
Surge contains detailed content guidelines for individual product components, including accordions, banners, tabs, and tooltips. The post says this component-level guidance lets VERBI enforce specific rules rather than relying only on broad instructions.
How was Storybook made available to the agent?
Glean could not initially access the Storybook-hosted design system, so guidance was scraped into an HTML file and uploaded, requiring repeated re-scraping. IT later used the Glean Indexing API to make Storybook a live data source.
What safeguards were added to reduce unreliable or unhelpful responses?
The system prompt tells VERBI to challenge assumptions, avoid agreeing by default, and say it does not know when the required information is missing from its resources. It also instructs the agent to keep answers short and omit unnecessary social padding.
Source and freshness
- Source
- intercom.com
- Published
- Added to Yomu
Summaries are generated from the original by a language model. Check the original before relying on it.
Keep reading
Related reading
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
The best engineering reads, once a week
One useful email. Unsubscribe anytime.