Original · Unique Research · 2026-03-14
Editor's note: This complete historical interview retains the original headline and promotional wording as source claims, not guarantees of commercial success. The headline's “below 1%” and the body's “below 30%” success rates have no stated samples, definitions or reconciliation in the source. Contract amounts and revenue ranges are retained without adding a currency, which the source does not specify. Career history, “world's first,” customer counts, research-time reductions, refund claims and competitive advantages are attributed to the interview or company and have not been independently audited. Zhang Keyi, Weirong Technology, Jirui Technology, Xiaobangbang, Yixuan Technology and Beita Data use romanized source names, not asserted official English legal names. Goals and predictions for 2026 reflect the interview date, not confirmed outcomes. The speaker's comments on data structuring, traceability and customer acceptance do not establish consent, privacy compliance or permission to use personal data; “effectiveness comes first” is his stated view, not this edition's privacy policy.
Unique Awards · Guest Interview
A First Contract Worth 230,000 Signed in 1 Week: Turning AI into a “Certainty Engine” for Product Innovation
“General-purpose large models can write poetry, but they cannot write a hit product that sells.”
Late at night, with Excel full of data on hundreds of failed new products, Zhang Keyi stared at the screen and suddenly realized: marketing can be powered by AI, but product innovation will always be a company's critical lifeline.
Former chief strategy officer at Anker Innovations, COO of Ziel Home Furnishing, and executive managing director for strategy at Fosun: behind these three roles lay the exhaustion of countless strategic reviews. When you oversee a business with billions in revenue but must rely on intuition to decide whether the next SKU is worth investing in, only those who have experienced it understand the helplessness of “data overload, difficulty generating insights and long innovation cycles.”
"
LynxAI is the world's first intelligent product-innovation platform, turning product success from “chance” into “certainty.”
This is not a complaint but everyday life for cross-border e-commerce businesses and consumer brands with annual revenue of 100 million–3 billion. New-product success rates persistently below 30%, difficulty retaining knowledge and inefficient cross-departmental collaboration: these problems confronting pet-supplies and home-furnishing businesses became the starting point for LynxAI.
The first paying customer was Weirong Technology: two conversations, one week and a contract worth 230,000. Seeing AI's ability to identify market opportunities precisely was enough to convince this pet-supplies company.
The research cycle fell from 2 weeks to 2 hours. To date, LynxAI has served 10+ major cross-border sellers ranging from hundreds of millions to around 10 billion in scale.
3 Parts of the Moat: Why ByteDance Cannot Do It
“If Baidu or ByteDance launched the same product tomorrow, what moat would you have left?”
Zhang Keyi's answer was calm: three elements, all indispensable.
① Data Barriers
Accumulated industry-specific data and knowledge graphs, and in-depth optimization of customer solutions. Proprietary algorithms have been iteratively optimized with real business data, while general-purpose large models struggle to adapt quickly to specialized vertical scenarios.
② Knowledge Accumulation
Integrating theories from consumer psychology, behavioral science, marketing communications and other disciplines to construct private knowledge graphs by specialty and industry.
③ A Team with Combined Backgrounds
The dual heritage of business and technology makes it possible to put the methodology into practice.
"
General-purpose large models can write poetry, but they cannot write a hit product that sells.
From the first customer to the current expansion at scale, every pitfall LynxAI encountered and every hypothesis it validated has become another building block of its moat.
The Reality of Commercialization: Selling Ongoing Support, Not SaaS
"
Customers are not buying tools; they are buying results. We are not selling SaaS but the certainty of staying alongside them all the way.
The monetization model is pragmatic: SaaS subscriptions + expert services + a share of results.
But Zhang Keyi is candid: customer churn is not caused by price but by “results falling short of expectations.” His response is direct: rapid improvement and iteration, plus “hands-on support”—aligning goals in advance and frequently iterating the quality of delivery. With this mechanism, LynxAI has so far had no customer request a refund over results.
Has your new-product success rate also remained below 30%?
LynxAI's first product, the “JAXX Market Insights Expert,” has launched to validate PMF. Requirements confirmation has the highest failure rate along the journey from acquisition to renewal: customers often confuse expectations of “buying a tool” with those of “delivery of business results.”
That discovery in turn shaped LynxAI's service approach: selling results rather than tools, and providing continuous support rather than a one-off delivery.
3 Goals for 2026: The Make-or-Break Challenge of Scaling
Zhang Keyi set three must-achieve goals for LynxAI in 2026:
① Growth at Scale
Shift from being “expert-service-driven” to “product-driven,” signing 80–100 enterprise customers.
② Lower the Entry Barrier
Enable small and medium-sized businesses with limited budgets to use AI for product innovation through a “self-service subscription” model.
③ Sharing in Results
Enter result-sharing partnerships only with existing customers. The greatest risks are trust and the probability of success, but high returns come only with high risk.
If the first goal is not met by year-end, it will “affect our industry standing.” That statement leaves no retreat.
The End-State Prediction: Human–AI Collaboration, Not Replacement
“Do you think AI will eventually replace product managers, or make product managers more human?”
Zhang Keyi's answer: human–AI collaboration and joint evolution.
Platform-based product design, following a technical approach of “structured templates + dynamic adaptation”: standardize the underlying framework, but adapt the specific content to the customer's industry and customer segments.
In the AI product-innovation value chain, LynxAI operates upstream, providing “support and recommendations from insights through product-innovation decisions.” It forms partnerships rather than competitive relationships with fellow event participants Jirui Technology, Xiaobangbang, Yixuan Technology and Beita Data.
Asked about the position at the end of 2026, Zhang Keyi offered a confident prediction:
"
LynxAI will become consumer brands' preferred intelligent product-innovation platform, and there will be a breakout innovative product recognized by the market.
When asked “Which metric are you least confident about?”, his answer was straightforward: “I am relatively confident.”
This article is based on Unique Research's in-depth interview with Zhang Keyi.
Selected Interview Q&A
A Snapshot of the Interviewee
Q: In one sentence, how would you make investors remember LynxAI?
A: LynxAI is the world's first intelligent product-innovation platform, turning product success from “chance” into “certainty.” The words that cannot be removed are “product innovation,” our core distinction from other AI tools.
Q: What was the moment of realization that led you to found LynxAI?
A: Reviewing missed market opportunities and failed product decisions, and making strategic plans and decisions on key product investments. I was organizing reviews, overseeing planning and making decision recommendations. Yet we always lacked credible structured data and professionally grounded methodologies tested in practice. We could only rely on experience and gut decisions.
Positioning and Moat
Q: What specific pain points does LynxAI address?
A: Product managers' and founders' problems with “data overload, difficulty generating insights, long innovation cycles and uncertainty in commercialization.” Customers previously relied on scattered Excel analyses, third-party research reports or consultancies. These were costly, inefficient and made it difficult to accumulate knowledge.
Q: Who is the typical customer, and how did you acquire the first paying customer?
A: Cross-border e-commerce businesses and consumer brands in pet supplies and home furnishings with annual revenue of 100 million–3 billion, facing “low new-product success rates and difficulty retaining knowledge.” The first customer was Weirong Technology, in pet supplies: two conversations, a contract signed in 1 week, worth 230,000. The key was demonstrating AI's ability to identify market opportunities precisely.
Q: If ByteDance built the same product tomorrow, what would your moat be?
A: Accumulated industry-specific data and knowledge graphs, and in-depth optimization of customer solutions. General-purpose large models struggle to adapt quickly to specialized vertical scenarios.
Commercialization in Practice
Q: What is the monetization model? Do customers leave because of price?
A: SaaS subscriptions + expert services + a share of results. They do not leave because of price but because results fall short of expectations. With hands-on support, we align goals in advance and iterate frequently. So far, no customer has sought a refund over results.
Q: Which stage from customer acquisition to renewal has the highest failure rate?
A: Requirements confirmation. Customers often confuse expectations of “buying a tool” with those of “delivery of business results.”
Q: What is the one breakthrough you must achieve in 2026?
A: Growth at scale: shifting from “expert-service-driven” to “product-driven” and signing 80–100 enterprise customers. Failing to achieve this by year-end would affect our industry standing.
Industry Insights
Q: Will AI replace product managers, or make product managers more human?
A: Human–AI collaboration, evolving and iterating together. Product design is moving toward a platform-based approach.
Q: How do you resolve the tension between data privacy and marketing efficiency?
A: Data structuring + traceability. No customer has abandoned cooperation over data-security concerns so far. Effectiveness comes first.
Q: Small and medium-sized businesses vs. large enterprises: whose needs are more urgent but harder to meet?
A: Small and medium-sized businesses have more urgent needs that are harder to meet, with limited budgets and weak data foundations. We currently mainly serve enterprises above the designated size threshold. In the future, “self-service subscriptions” will lower the entry barrier.
Q: Would you accept a pay-for-results model? What is the greatest risk?
A: We enter result-sharing partnerships only with existing customers. The greatest risks are trust and the probability of success.
Q: Where do you sit in the AI marketing value chain, and who would you like to integrate your offering?
A: Upstream, providing “support and recommendations from insights through product-innovation decisions.” We hope to collaborate with data providers and marketing-advertising partners.
Q: What do you predict for the end of 2026? Which metric are you least certain about?
A: LynxAI will become consumer brands' preferred intelligent product-innovation platform, and there will be a breakout innovative product recognized by the market. I am relatively confident.
This document is original content by Unique Research.