WhatsApp Business Messaging
On WhatsApp Business, the buttons that started a conversation were lost among action CTAs, so customers couldn't easily begin one. I redesigned the in-thread experience and shipped a GenAI template that helps advertisers send better starters, lifting conversation initiation by 14%.
Paid messaging is Meta's fastest-growing business surface
Paid messaging (marketing messages) is WhatsApp Business's revenue-driving product — promotional messages businesses send to customers at scale on WhatsApp for re-engagement. After we launched GA for WhatsApp, I focused on conversation automation improvements, starting with Quick Replies. Check out our article & video:
Consumers can't tell which button starts the conversation, even with the intent
1 in 2 clicks on a business message is already a free-form reply. The intent is clear, but quick replies and action CTAs look the same, so the buttons that lead to a real back and forth get lost, and only 1% of messages ever reach one.

Conversation starters across the industry
Business messages with conversation threads drive 11% higher CTR than standard threads, so I looked into conversation starters across the industry: FoA (Click-to-Messenger, Click-to-Instagram Direct), WhatsApp's own AI chat surface, Cleo, Grok, and Siri. WhatsApp Business was the outlier.
Make it easy for consumers to start a conversation, lifting deep-conversation rate
Quick replies moved out of the XMA (rich media message) and now sit below it, left-aligned at lower opacity — discoverable but subordinate to the paid creative, so the action CTA keeps the click. The bet: surface conversational intent without eating CTR (the metric the marketing message is sold on).
Make it easy for advertisers to ship campaign-specific quick replies
Only 20% of business message campaigns use quick replies, and 90% are generic ("Learn more," "Yes," "Interested"). I shipped a GenAI template inside the Ads Manager chat builder that drafts campaign-specific replies, so advertisers ship better starters without writing from scratch. Since quick replies now sit outside the XMA, that interaction change also flowed through to the builder's chat preview.
Early results: +14% lift in conversation initiation, and a shift from clicks to conversations
A held-out A/B at 1% of marketing messages drove a +14% lift in conversation initiation rate. Beyond the number, I partnered with eng and PM to instrument conversation depth, reply rate, and completion alongside CTR — so the team now measures the conversation, not the click. That reframing is what earned leadership buy-in. The conversation-modal prototype became a reusable template that adjacent surfaces (Messenger, IG Direct) build on.
Adopted AI workflow to improve efficiency and quality of output
I adopted AI throughout the design process: • Leveraged internal AI analytics tool to come up with quick replies usage • Used multi-agents for PRD drafting and design explorations • AI-powered prototyping (the interactive prototype above was built this way)
Since this is still in experiment and under NDA, I can't reveal more information here. Please reach out to me via email (glimesong@gmail.com) for further details.





