Deep: New Multi-Modal AI features Explored

2026-03-23 · Source: Department of Product · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Project & Product Management · Depth: Intermediate, quick

Summary

Pinterest's product teams are reportedly in a dispute with their CEO over the primary modality for the company's AI assistant, highlighting a broader challenge for product teams in 2026. While the CEO advocates for a voice-first approach to align with Gen Z expectations and create a "talking to a friend" shopping experience, designers and product leaders argue this could undermine Pinterest's core visual discovery value. This internal conflict underscores a significant shift in product design, where the interface layer is no longer fixed but a distinct decision across text, voice, image, video, or documents. The analysis will explore over 30 new multimodal AI features from companies like Google, Anthropic, DoorDash, and Lyft, and provide a framework for product teams to navigate these complex modality choices.

Key takeaway

For AI Product Managers designing new features, carefully consider the primary interaction modality. Your choice directly impacts user experience and product value, as seen in Pinterest's internal debate. Use a structured framework to evaluate whether voice, text, image, or a combination best serves your product's core purpose and user expectations, rather than adopting a modality simply for its novelty.

Key insights

Choosing the right AI modality is a critical product design decision that can make or break a product's core value.

Principles

• Modality choice impacts core product value.
• Multimodal AI combines interaction types.
• User expectations drive interface evolution.

Method

A five-test framework helps product teams decide on appropriate AI modalities by asking key questions about user interaction and product value proposition.

In practice

• Evaluate voice for conversational interfaces.
• Consider image for visual search and generation.
• Explore document understanding for structured data.

Topics

• Multimodal AI
• Product Design
• AI Assistant
• User Interface
• Voice AI

Best for: AI Product Manager, Product Manager, Director of AI/ML

Related on AIssential

• Meet the former Apple designer building a new AI interface at Hark — Hark aims to redefine human-AI interaction by deeply integrating AI models, hardware, and interfaces for a personalized, intelligent system.
• From Visual Question Answering to multimodal learning: an interview with Aishwarya Agrawal — VQA research has evolved from basic perception to complex multimodal challenges like cultural alignment and embodied AI.
• Mira Murati steps back into the spotlight, carefully — Thinking Machines Lab is developing "interaction models" for real-time, continuous human-AI communication beyond turn-based prompts.
• New AI Model SHOCKED Me — Thinking Machines Labs' new model offers real-time, multimodal, and context-aware conversational AI.
• Everyone Says “AI Is Everywhere.” Here’s What That Actually Means. — AI comprises distinct, specialized modalities, each with unique capabilities, failure modes, and applications.
• How AI Mode is changing the way people search in the U.S. — AI Mode is transforming search by enabling natural language, multimodal queries, and expanding the scope of what users can find.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Department of Product.