JPMorgan Dialogues with Zhipu AutoClaw: Why are agents booming, does model quality still matter, and how can commercial monetization be achieved?

JPMorgan Dialogues with Zhipu AutoClaw: Why are agents booming, does model quality still matter, and how can commercial monetization be achieved?

```

Why has "raising lobsters" gone viral? Which industries might be disrupted by "AI lobsters"? How can it achieve commercial monetization?

According to the China Securities Research Report released by J.P. Morgan on March 12, its analysts, XU Wentao and YAO Cheng, recently communicated with the project manager of Zhipu AutoClaw, deeply analyzing the reasons behind the popularity of products such as AutoClaw and OpenClaw, and the subsequent application implementation, monetization pathways and logic.

J.P. Morgan believes, "The importance of products such as AutoClaw and OpenClaw lies not in their making autonomous AI commercially viable overnight, but in their significant lowering of the threshold for non-technical users to experience agent workflows."

For the market, the core impact is: Although agent adoption is expected to expand model usage and infrastructure demand, short-term monetization remains in its early stages. Actual deployment will first land in relatively structured workflows, not broad, fully autonomous human replacement.

Agent ("Lobster") Explosion: A Victory of Product Design, Not Model Mutation

The recent craze for OpenClaw-type products: is it a leap in model capabilities or an optimization of product design? The interviewee gave a clear answer. This trend "reflects improvements in product design and usability, rather than a mutation in model intelligence."

The interviewee especially emphasized three key factors: "Integration with existing communication tools, the durability of agents building user profiles over time, and broader system permissions to extend the agents' practical working scope."

J.P. Morgan pointed out that this distinction is crucial. The current popularity is driven by better productization and accessibility, meaning user engagement can be expanded before true enterprise-grade monetization is realized.

Quality of Foundation Models: Key to Commercial Ceiling

With endless new agents, will foundation models soon become commoditized? The clearest point from the discussion: "The commercial ceiling for agents still largely depends on the quality of the foundation model."

The interviewee repeatedly emphasized, "Agents are essentially containers or mediums; models remain the core factor that determines whether tasks can be completed accurately and consistently, and whether reasoning depth suffices to work in high-value contexts."

J.P. Morgan believes this rebuts the view that, in the short term, the agent layer will completely commoditize the model layer. "Better models should still result in better task completion, stronger instruction following, more stable long-context performance, and superior handling of open-ended workflows." Therefore, the rising adoption of agents is still a major positive for leading model providers.

Commercial Monetization: Still in Early Stages, Focusing on Structured Tasks

Despite the heat of the agent concept, the interview hints that "in the short term, the agent market may still be in an exploratory stage without sufficient monetization." Current products are still helping users discover use cases. For significant expansion in commercial scenarios, "6 to 12 more months of model improvement, workflow training data, and product iteration may be needed."

J.P. Morgan sees this as consistent with the status quo of enterprise AI. "Coding and technical workflows remain the clearest early monetization path, as tasks are structured, objectives are clearer, and execution trajectories are easier to define." Beyond coding, lack of standardized "trajectory" data is a key constraint on agent execution for real-world multi-step tasks.

Market Landing: Technical Engineering and Structured Workflows First

In entering the market, which domains will adopt agents first? The interviewees highlighted three main categories:

Technical engineering: "From coding extending to testing, deployment, operations, configuration, and debugging. This seems to be the most commercially credible category today."Information and content workflows: "Including research, report generation, document processing, office file operation, and internal content production."Personal productivity: "Such as email, calendar, and message management." While attractive to consumers, sustained monetization may take a longer time.

J.P. Morgan advises investors, "Short-term expectations should be based on technical and structured enterprise tasks rather than overly aggressive assumptions from the early adoption by consumers."

Open Architecture and Moats: Quickly Replicable Features Are Not Key

Another important perspective in the discussion is that the agent layer may not be a winner-takes-all proprietary model channel. AutoClaw supports multiple model vendors, with management explicitly supporting an open architecture rather than forcing exclusive use of Zhipu models.

J.P. Morgan believes this broadens the potential market for the product and increases the opportunity for the agent platform to become a model ecosystem aggregator. But for model vendors, this means "the agent interface itself may not guarantee exclusive downstream value capture, unless the provider also leads in model performance, agent tool invocation, and workflow integration."

Regarding the moat, management believes that feature comparisons are less important, as many visible features can be quickly copied.

They believe genuine defensibility is built on three aspects: "Speed of product insight, quality of foundational models, and cumulative agent features (such as browser tools, memory systems, and workflow handling)."

J.P. Morgan agrees, noting that investors should focus on whether providers can "improve task completion rates over time, reduce friction, and comprehensively use data to improve agent performance."

Industrial Chain Reshaping: Who Benefits, Who Gets Disrupted?

Wider adoption of agents will benefit several parts of the AI stack:

First, wider agent adoption should benefit model providers, as more autonomous workflows mean more token consumption and sustained usage.Second, it should benefit inference infrastructure, cloud, and related compute providers, especially as demand growth continues to outpace supply.Third, open API and controlled-integrated collaboration and workflow platforms can also benefit by serving as the interface that embeds agents into daily work.

Conversely, companies whose value proposition is "shallow intermediation or low-threshold information processing" may be at risk. For "roles or services with limited moats, open information, and relatively easy-to-automate workflows, AI is more likely to apply pressure."

Additionally, security and regulation are real constraints for enterprise deployment. Management believes that "prompt injection, permission errors, malicious third-party skills, and software vulnerabilities are real constraints." This may slow down monetization in the short term, but will raise the importance of trusted suppliers and compliance-grade architectures.

J.P. Morgan gives Zhipu an "Overweight" rating, with a target price of 800 HKD by December 2026 based on a 30x estimated 2030 P/E ratio, expecting annual revenue CAGR of over 100% in 2026-2030.

 

~~~~~~~~~~~~~~~~~~~~~~~~

The above wonderful content comes from Zhui Feng Trading Desk.

For more in-depth analysis, including real-time interpretations and frontline research, please join Zhui Feng Trading Desk·Annual Membership

Risk Warning and DisclaimerThe market has risks, and investment needs caution. This article does not constitute personal investment advice and does not consider the individual investment objectives, financial situations, or needs of any specific user. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific situation. Invest accordingly and at your own risk. ```