As corporate digital transformation developed, AI is shifting from being just an efficiency tool to a central player in business decision-making. It’s no longer just about streamlining tasks—it’s about driving core business choices.
At a recent offline seminar hosted by Huxiu Think Tank, Chen Ouyang, CTO of DeepZero|iPinYou, made a key point: companies need to move past the trap of focusing solely on piecemeal efficiency improvements and instead adopt AI-native applications that solve complex tasks end to end. He stressed that slapping “magic buttons” onto existing software rarely unlocks the kind of growth businesses truly need.
To unpack this shift in technology strategy, Huxiu Think Tank dug deep into the AI Agent partnership between DeepZero|iPinYou and Suntory (China). The Suntory case stands as a model for AI appilication: by building an AI-native intelligent agent to replicate consumer feedback, the company turned its new product development process—once heavily dependent on human experience—into a predictable, closed-loop industrial decision-making chain.
Core Paradigm Shift: From “Software + AI” to AI-Native Agents
Ouyang noted that the past practice has made one thing clear: retrofitting old business software with “magic buttons” or isolated AI features rarely delivers meaningful results.
- Prioritize business logic over functional add-ons: What business teams really care about isn’t faster clicks—it’s hitting end-to-end goals, like boosting new product success rates.
- Task breakdown and autonomous execution: AI-native agents can take on high-level business instructions, split them into manageable tasks, target specific audiences, create content, and roll out strategies—all on their own, with humans only stepping in to sign off at the end.
- Shift the value focus: Companies should stop fixating on cutting customer service costs and start focusing on growing marketing and sales. Ouyang believes a 10% lift in marketing performance holds enormous strategic value—and that’s where AI Agents shine brightest.
Redesigning the Decision-Making Pipeline: From 20-Character Searches to 1M-Token Conversations
Suntory’s China operations faced common pain points: scattered data, decisions based too much on experience, and slow response times to market changes. Its AI-powered new product development and launch platform—built on a multi-Agent collaborative framework—has completely reworked the entire R&D process.
1. “Replicating Consumers”: Validation Through Virtual Personas
The most groundbreaking part of the Suntory project is its virtual persona system.
- Simulated testing: The system creates “AI consumers” modeled after different demographic groups, letting Suntory quickly test product ideas and gather feedback without waiting for real-world trials.
- Data calibration: To make sure this virtual feedback is reliable, DeepZero|iPinYou calibrated the AI using historical survey data. AI consumers that deliver highly accurate insights, giving Suntory actionable market forecasts for new products.
2. Multi-Agent Collaborative Decision-Making System
The platform uses specialized Agents across different functions, turning siloed departmental work into a seamless digital process:
- New Product Insight Agent: Pulls together social trends and sales data to automatically suggest combinations of product elements with strong market potential.
- Packaging Design Agent: Uses audience preferences and product elements to generate packaging concepts on its own, cutting down design time significantly.
- Channel Insight Agent: Analyzes store data and consumption scenarios to create targeted launch channel and marketing strategies.
- Legal Review Agent: Checks all marketing and product content upfront to ensure it meets regulations, reducing compliance risks before they arise.
Think Tank Insights: Three Foundational Rules for AI Agent
In his talk, Ouyang shared key lessons for rolling out AI Agents in large enterprises—including Fortune 500 clients—that non-technical leaders need to know to judge project success:
- Start with the scenario, drive with the business: AI projects can’t be run in a vacuum by technical teams. Without clear business use cases—like marketing or sales—more than 95% of projects will fail.
- Combine large and small models: Standalone large language models (LLMs) can be unstable in recommendation scenarios. The Suntory solution pairs LLMs with proven traditional machine learning models (small models) to keep business results consistent and reliable.
- Dual human-AI collaboration: The platform offers two modes—AI Agent auto-recommendations and manual validation. This helps new employees get up to speed quickly while supporting senior experts, making it easier to roll out the system across the organization.
Conclusion: From Efficiency to Opportunity Creation
For Suntory, AI Agents do more than just speed up market research and design—they cut down on failure risk. By putting real (and realistically simulated) consumer feedback at the center of R&D, the company has drastically improved how well its products match what the market wants.
As Ouyang put it: Efficiency isn’t the end goal. The real power of AI is unlocking new business opportunities—doing things that were once too complex or hard to tackle, and doing them well.
