Honor CEO James Li: The Battle for the AI Entry Point Needs a New Approach; AHI Framework Adds TQ Trustworthiness

Honor CEO Li Jian: The battle for the AI entry point needs a new playbook. Phoenix Tech Download the app for exclusive early access. November 06, 2026 12:32:41 From Beijing. Updated November 06, 2026 12:32:32

Author / Liu Yukun

However, technical capability can take over applications, but it is hard for it to become an entry point universally accepted by the ecosystem. The underlying logic of the battle for the entry point is changing.

Today, at the 2026 Mobile Network Conference, Honor CEO Li Jian gave a different answer. He said the entry point does not need to be fought over; trust will gather on its own; the entry point of the future does not belong to any single vendor, but to every user.

Speaking of the battle for the AI entry point, he mentioned that over the past decade, building walls was the industry's moat—the higher whose wall, the larger whose territory. But users' lives have never been divided into inside and outside the wall. On the surface this is a statement of position, but in reality Honor is already switching to a different competitive logic.

A Goldman Sachs research report in January this year proposed that the super entry point of the AI era may be monopolistic. Users need only one default entry point, and other participants may degrade into functions that are called upon. For terminal vendors and model vendors, no one is willing to be just a screen without decision-making power. So seizing the entry point, doing proxy operations, and taking over across applications have become a seemingly natural choice.

"The measure of technology is ultimately the measure of people." In Li Jian's view, what determines a business paradigm has never been technology, but scarcity. From the channel economy to the attention economy to the intent economy, every shift in scarcity redefines where the entry point is.

And the closed loop from intent to result must cross three thresholds: intelligence, context, and trust. The more proactive the intelligent service, the more severe the challenge to trust.

This is also why Honor added TQ to the AHI framework this year. In March this year, when Honor talked about AHI, the core was still IQ and EQ: intelligence and sense of life. Eight months later, the framework has gained TQ, that is, trustworthiness. The logic is not complicated: the more proactive the agent, the clearer the boundaries need to be; the stronger IQ and EQ are, without TQ constraints, the greater the risk of overstepping and losing control.

This is not just an upgrade in product philosophy.

In May this year, three departments jointly issued the "Implementation Opinions on the Standardized Application and Innovative Development of Agents," setting safety and trustworthiness as the industrial bottom line; the mandatory national standard plan for agent security issued in July was also drafted under the leadership of China Mobile and other units. The requirements of regulation and industrial security are already very clear. Honor's talk of trust at this point in time is responding to changes in the rules and also paving the way for subsequent cooperation.

This also makes the terminal's position more special. Li Jian's logic is that data permissions stay local and are controlled by the user, and the terminal is visible and tangible, so the terminal is both the entry point of intent and the exit point of trust.

This is the background against which Honor proposed the user-symbiotic Agent. Honor defines it as a digital companion symbiotic with the user, whose mission is to understand the user, empower the user, and protect the user. The ownership of data and the boundaries of its use are written into the product definition first. Any later discussion of who decides for whom is derived from this definition.

In terms of technical implementation, this requires more systematic security design. For example, isolating sensitive intent memories and authorization credentials in a chip-level secure area, using on-device local models to perform dynamic filtering at runtime, and building isolated computing fences at the memory layer, to ensure as much as possible that data does not leave the local device. Trust here is not an adjective in a brand brochure, but must become verifiable machine logic.

This also explains why Honor wants to write trusted settlement into its cooperation list with China Mobile. The two sides want to jointly build a complete trust chain covering trusted networks, trusted identity, trusted settlement, trusted operations, and trusted terminals. Trusted settlement is one link in this chain; it is not a clearing system that Honor proposed on its own and that is already running.

Operators hold identity and settlement systems, which is the foundation for building a trusted chain. This time China Mobile released the MoMA mobile model service platform and the liquid-cooled intelligent computing foundation at the conference, connecting on the cloud side with computing power networks, cloud models, and the service ecosystem; on the other end, Honor uses on-device local shields and the AHI framework for intent scheduling. What the two sides form is a device-cloud division of labor, not simple channel cooperation.

Honor does not intend to build all services itself; what it needs is an alliance that can bring together services, models, and settlement. After connecting Lingxi, Qwen, Doubao, and Gemini into a network, Honor still needs China Mobile's trusted settlement chain to explore rules for charging by result.

This also concerns Honor's own transformation.

The core of the five-year target of more than 10 billion US dollars is not selling a few more phones; the key is whether it can transform from a phone company into an AI terminal ecosystem company. Selling hardware is one-time, while the value of an ecosystem company comes from every time a service is invoked. If billing by result can hold, the terminal will no longer be just the sales endpoint of the industrial chain, but may become the starting point for billing and secondary revenue sharing.

Of course, all of this is still very early. A large number of apps have no intention of signing access agreements yet, and if the boundaries of proactive services are not well controlled, they can easily become high-frequency interruptions. After users authorize proxy operations to agents, the validity of transactions and legal liability also need supporting rules to be further clarified. The game of interests at the commercial level will continue for a long time.

But the direction is already fairly clear: continuing to fight over the entry point will hardly solve the problem of delivering results. What really needs to be discussed is how to co-build the entry point and how to jointly build trust.

To put it more plainly, the entry point does not need to be grabbed, but trust needs to be won bit by bit.