Quora x Poe
Case Study: Turning a Q&A Platform Into an AI-Native Experience
When Quora launched in 2009, its mission was to build a library of human knowledge—organized through questions and answers, reputation, and curiosity. In 2023, that mission took a radical turn with the launch of Poe, a conversational AI platform designed to give people instant answers, summaries, and ideas from the world’s most advanced large language models.
I joined the team at the inflection point—where a trusted information platform needed to evolve into an AI-native product without losing its soul. My role was to help architect the visual, interaction, and editorial design systems that would guide Poe from an experiment into a full product.
The Challenge
Quora’s core strength has always been human knowledge. Poe had to extend that ethos into a new domain: machine-generated content.
But this wasn’t about creating a chatbot UI. It was about building trust, shaping expectations, and making Poe feel like an extension of Quora’s mission—not a novelty, not a toy, but a serious interface for serious knowledge-seekers.
We had to answer:
How do you design for a conversation that changes every time?
What does editorial tone mean when the author is an LLM?
How do you distinguish bots with different personalities and capabilities?
How do we monetize a new form of interaction?
Key Projects and Contributions
1. Bot Identity and Interface Architecture
We launched with multiple LLMs (Claude, GPT, LLaMA) accessible as separate “bots,” each with their own strengths. I worked on:
Bot profile design: Visual metaphors, avatars, tone calibration
Model selection UI: Helping users choose the right engine without needing technical knowledge
Navigation systems: Ensuring conversations with multiple bots stayed organized and legible
The result was a multi-agent interface that felt legible, intentional, and lightweight—even as the complexity of backend models increased.
2. Editorial Tools for AI Journalism
One of the most exciting initiatives was enabling real-time AI-generated journalism—summary bots, trending-topic explainers, and prompt-driven digests. I helped design:
Internal tools to create custom bots with prompt scaffolding
AI content workflows that mimicked editorial processes (headline, lede, narrative arc)
UI patterns for inline summaries, citations, and rephrasing
This work redefined what "publishing" meant—blending AI generation with human guidance into a new editorial format.
3. Monetization & Subscription Models
We explored how to make Poe sustainable while remaining accessible. I contributed to:
The upsell flows and premium tiers for advanced model access
Designing “bot-as-product” interfaces for creators to launch and promote their own bots
Experiments with in-conversation nudges and usage-based triggers
This helped frame Poe not just as a tool—but as an ecosystem.
4. Trust, Transparency, and Tone
LLMs hallucinate. They echo bias. They can’t be “edited” in the traditional sense. I worked on:
Framing systems to clarify AI authorship vs. human content
Prompt constraints and system messages to set expectations
Tone modulation for bots to avoid anthropomorphism while remaining helpful
The result was a UI that invited curiosity while maintaining caution—positioning Poe as an intelligent companion, not a sentient oracle.
Impact
Poe grew from an alpha experiment into one of the most recognizable public LLM platforms.
Thousands of custom bots were created within months of launch.
Quora transitioned from a static knowledge archive to a live, conversational knowledge engine.
Internal tooling enabled rapid experimentation with new models and editorial formats.
Reflections
Designing for Poe meant letting go of predictability and embracing emergence. Instead of guiding users down predefined flows, we had to design frameworks for exploration—where prompts became product features, and outputs became conversation starters.
It pushed me to think about designing language itself, and how tools can reshape not just what we say, but how we think.