AI Product Development: InstantAsk

01 | The Challenge

On DallasNews.com, over 98% of readers were anonymous, making personalization and retention nearly impossible. Readers often skimmed long-form articles, looking for quick answers but rarely converting or exploring further.

At the same time, I noticed a growing issue across the industry: site-wide AI chatbots were driving bounce rates up, giving users answers instantly without them ever engaging with the content.

The opportunity was clear: deliver instant answers without losing engagement.

02 | The Goal

Create an in-article AI experience that:

Gives readers fast, contextual answers
Encourages deeper scrolling and exploration
Converts anonymous readers into registered or subscribed users
Builds the foundation for personalized content recommendations

03 | Discovery & Analytics

Insight: True engagement comes from making information easier to access.

Analytics
Showed high drop-off mid-article and short session times.

User Research
Revealed readers wanted “bite-sized clarity” without losing trust.

Product Strategy
Prioritized increasing registrations to build first-party data and boost ad revenue.

04 | The Approach

1. Designing for Curiosity

I defined Instant Ask as an AI-powered Q&A widget embedded directly in the article:

Pre-generated questions appear as readers scroll (“When does this go into effect?”).
Clicking reveals a short AI-generated answer and a “Read More” jump to the exact section.
After 1–2 interactions, a registration prompt appears inviting users to “Ask your own.”
Subscribers unlock custom Q&A, related articles, and a personalized dashboard.

2. Collaborative Execution

I led cross-functional work across product, AI, and editorial:

Editorial: Built a CMS PowerUp for editors to approve or edit AI-generated questions.
AI/Engineering: Developed a retrieval-augmented generation (RAG) pipeline to ensure responses were grounded in owned content only.
Design: Created Figma prototypes showing the full “question → answer → engagement → conversion” flow.
Data: Defined KPIs for session duration, scroll depth, registration uplift, and clickthroughs.

3. Testing & Learning

We planned to launch MVP on one section:

A/B test widget placement and prompts.
Measure interaction-to-registration lift.
Use human-in-the-loop review to minimize hallucination risk.

05 | The Impact

For users:
✓ Quick, high-quality answers; no more endless scrolling.
✓ Clear pathways to deeper reading and learning.
✓ Personalized follow-ups and dashboard access for subscribers.

For the business:
✓ Increased session duration and article completion rates.
✓ Lift in registration and subscription conversion.
✓ More internal content discovery and repeat visits.
✓ Foundation for future personalization via user Q&A data.

Instant Ask turned passive readers into active participants, creating a smarter, more habit-forming reading experience.

06 | Tech Snapshot

LLM & RAG: GPT-based retrieval from internal article archives
CMS Integration: ArcXP (CMS) PowerUp for editorial oversight
Infrastructure: Azure + OpenAI embeddings
Analytics: Firebase + custom event tracking

07 | The Timeline

Phase 1 (MVP) | 6–8 weeks: Pre-generated Q&A, registration prompt, “Read More” jumps
Phase 2 | 4–6 weeks: Dashboard + related articles
Phase 3 (Ongoing) | Optimization, A/B testing, model tuning

08 | Transformational Impact

Instant Ask bridged AI and journalism in a way that strengthened editorial engagement. By embedding curiosity into the reading experience, we created a feature that earned attention, deepened trust, and unlocked a new layer of insight-driven personalization.