In the previous article titled "What are the good paths for Web3 entrepreneurship in China? (Part Four)", Portal Labs first discussed three types of Web3 teams that are more focused on infrastructure and how they can migrate their capabilities towards AI.

Data teams can look at the AI data layer to address issues related to authorized data, verifiable data, and compliant calls; identity and account teams can consider Agent permissions, accounts, and execution records; payment and wallet teams can focus on Agent automatic settlement, API micropayments, and financial auditing. These three paths share a common point: they all take the foundational capabilities that Web3 has developed over the past few years and apply them to the new demands arising from AI Agents.

However, the directions that Chinese Web3 teams can migrate to are not limited to data, identity, and payments. There are two additional categories of teams worth examining individually.

One category is security and risk control teams. In the past, they served contracts, wallets, cash flows, and on-chain risks; as we enter the AI Agent stage, new security issues will arise regarding permissions, tool calls, automatic payments, data access, and execution records. The more an Agent can do for users, the more boundaries need to be set, anomalies need to be checked, and records need to be kept.

The other category is application-layer and community-oriented teams. They may not need to transform into AI infrastructure companies but can integrate AI into their existing products and operational processes to enhance investment research, content, customer service, community engagement, education, and user conversion efficiency. For this type of team, AI acts more like a layer of capability enhancement rather than a complete career shift.

Therefore, this article will continue along the logic of the last one: how security and risk control teams and application-layer and community-oriented teams should migrate towards AI.

At the same time, Portal Labs also needs to clarify another point. Not all AI directions are suitable for Chinese Web3 teams to enter. Some directions may seem very hot, such as general large models, general Agent platforms, AI traders, and automatic yield products; however, they actually have high thresholds, intense competition, and may even encounter very sensitive compliance boundaries.

Whether migration is possible cannot be solely based on the popularity of AI. More importantly, it is about what capabilities the team originally has, whether these capabilities can be applied in real scenarios, and whether they can find a clear payer.

Security and Risk Control Teams: From On-chain Security to Agent Behavior Auditing

Security and risk control have always been a direction within Chinese Web3 teams that can traverse cycles more effectively.

Regardless of market conditions, Web3 projects need to conduct contract audits before going live, wallets need to be protected against theft, cash flows need monitoring, attack incidents need tracking, and there has always been a demand for KYT and anti-money laundering tools. Many security teams have survived due to these actual needs.

In the past, these teams primarily focused on smart contract vulnerabilities, private key risks, wallet security, on-chain attacks, cash flow directions, and suspicious transactions. After the development of AI Agents, security issues will extend from on-chain assets to a broader range of automated behaviors.

Because Agents are no longer just answering questions; they will begin to call tools, access data, execute processes, and even initiate payments and on-chain operations.

For example, if a company integrates AI Agents into their CRM, email, contract database, internal knowledge base, and ticketing system to organize customer information, generate meeting minutes, draft reply emails, inquire about contract terms, and even automatically create tasks and follow up with customers, this scenario may appear to enhance efficiency but will involve a significant amount of permissions and data flow. Can the Agent read all customer profiles? Can it send contract contents to external tools? Can it access employee emails? Can it automatically send emails to customers? If it is induced by an attack on prompt words, will it leak internal information?

These will all become new security issues.

If a company begins to adopt AI workflows on a large scale, security needs will extend from model security to behavioral security. Companies will not only care about whether the model's answers are correct, but also about what the Agent has done, which systems it has called, which files it has accessed, to whom it has transmitted data, and whether they comply with internal permissions and regulatory requirements.

This is precisely the direction in which security and risk control teams can migrate.

Teams that previously conducted on-chain monitoring, auditing, risk control, and cash tracking can transfer their capabilities to Agent behavior auditing, permission anomaly identification, data call monitoring, automated payment risk control, and corporate AI security governance.

For instance, providing companies with Agent operation logs so that each tool call can be traced; setting permissions boundaries for AI workflows to prevent unauthorized access; establishing risk control rules for automated payments to identify anomalous calls; and generating audit reports for internal data calls to help companies meet compliance requirements.

This direction may not necessarily have strong viral potential but has a clear B2B attribute.

The more AI a company adopts, the greater the need for security, permissions, and auditing, especially in industries such as finance, healthcare, government and enterprises, legal, and education where AI cannot prioritize efficiency alone but must also be controllable, traceable, and accountable.

For Chinese teams, the direction of security and risk control also allows them to more easily avoid high-risk narratives. It does not need to directly handle tokens, manage user funds, or promise returns. As long as it addresses real risks encountered in the AI usage process within enterprises, there is an opportunity to form sustainable service revenue.

However, this direction also has a threshold.

Agent behavior auditing cannot be simply understood as “rebranding on-chain monitoring.” It requires an understanding of corporate permission systems, AI tool calls, data security, log analysis, and business processes. If Web3 security teams want to enter this realm, they need to supplement their knowledge of AI engineering and corporate security and cannot just rely on their previous contract auditing methods.

But in the long term, this path is worth attention. The more AI integrates into real businesses, the more security issues will transcend the model itself. Those who can help enterprises understand what the Agent has done, which behaviors pose risks, and how to trace issues will likely become important service providers in the AI infrastructure space.

Application Layer and Community-Oriented Teams: From Web3 Products to AI-Enhanced Products

This type of team includes content platforms, investment research tools, trading tools, educational products, community products, growth tools, and user operation products. They may not be suitable for directly becoming AI infrastructure but are well-suited to embed AI into their existing businesses.

The most common mistake application layer teams make is to rush to transform themselves into AI companies as soon as they see the AI trend. Those who used to focus on community are now saying they want to do AI social; those who used to focus on content are now claiming they want to become an AI content platform; those who used to focus on research are now saying they want to be AI investment advisors. It sounds like a significant change, but without real scenarios and paying demand, it can easily turn into a mere rebranding.

A more realistic approach is to embed AI into existing products to solve already existing user problems.

Some references already exist for this direction. For instance, products like Kaito essentially do not simply create an “AI chat tool” but organize project dynamics, social media, narrative popularity, content dissemination, and user attention around the problem of information overload in Crypto, enabling researchers and project parties to quickly see what the market is discussing. It inspires application-layer teams to realize that AI does not necessarily need to be a standalone product; instead, it can serve as a layer of capability for information filtering, semantic organization, and signal discovery.

Similarly, some Crypto Copilot and research assistants do not decide on the quality of a project for users but rather compile announcements, white papers, governance proposals, on-chain data, financing information, and market trends into more easily understandable content. For investment research tools, this has more value than simply making a “Q&A robot.” Because the real pain point for users is not the inability to ask questions but the overwhelming amount of information to handle daily, the scattered sources of information, and the high judgment costs.

Community and operational tools follow the same logic. Project parties need to manage user inquiries, event feedback, community content, KOL data, and growth leads every day. If AI is only placed in Telegram or Discord to answer a few frequently asked questions, its value is very limited. However, if it can help project parties organize frequently asked community questions, label users, identify active contributors, categorize event feedback, and generate operational summaries, it becomes a genuinely integrated tool in operational processes.

Educational products can also be viewed this way. The most challenging aspect for new Web3 users is not necessarily the lack of content but rather the vast amount of content, high entry barriers, and the difficulty of distinguishing truth from misinformation. AI can generate learning paths according to user levels, explain terminology, organize case studies, provide Q&A practice, and even simplify complex content into versions suitable for beginners.

Therefore, for application-layer teams, AI is better suited as an amplifier of product capabilities and operational capabilities.

Content platforms can use AI for information filtering, summarization, recommendations, and multilingual distribution; investment research tools can utilize AI for on-chain data interpretation, project monitoring, market information collation, and risk alerts; community products can use AI for automated Q&A, user stratification, event operations, and content review; educational platforms can implement AI for personalized learning paths, course generation, and inquiry support; trading tools can leverage AI for data analysis, risk reminders, and strategy assistance.

These directions may not seem as grand as “Agent economy,” but they are easier to implement. Because application-layer teams already have users, content, scenarios, and operational experiences. Adding AI addresses problems that already exist in the original products, such as too much information, user confusion, high customer service costs, slow content production, low research efficiency, and heavy community operations.

The key to such migration is not to depart from the original user scenarios.

If a Web3 investment research tool primarily serves traders and researchers, then AI can help users understand announcements, white papers, on-chain data, and market changes more quickly. If a Web3 education platform caters to beginner users, AI can assist in personalized Q&A and learning paths. If a community product serves project parties, then AI can help project parties stratify users, maintain communities, and reach out for events.

These are all actual demands.

Application-layer teams often do not need to pursue “transformation.” Integrating AI as a new capability into their original products enables them to capitalize on existing user bases, content, and business foundations while also avoiding entering a completely unfamiliar AI red ocean.

Of course, this path should not stop at merely adding a chat robot.

Many products that claim to be AI-enabled currently only have an added Q&A window. The user experience has not significantly improved, nor has business efficiency changed noticeably. This type of AI-enabled integration is unlikely to create long-term value.

Truly effective AI enhancements should be embedded within users' original workflows. They should either save users time, enhance decision-making quality, lower operational costs, or increase conversion and retention. If these objectives are not achieved, AI functionality will quickly become a mere ornament.

Thus, for application layer and community-oriented teams, the most pragmatic migration approach is to first use AI to make the existing products and operations more efficient. Whether users find the information easier to understand, whether project parties can manage communities more easily, whether researchers can complete judgments more quickly, and whether customer service and growth can operate with fewer resources—these factors are more critical than “whether to transform into AI.”

Which Directions Should Be Avoided?

After discussing teams suited for migration, it is important to clarify which directions should be approached with caution.

The first type involves building a general large model from scratch.

This direction requires modeling capabilities, computational resources, training data, research teams, and long-term capital investment, already placing it in a highly competitive market. Large model companies, major internet firms, and AI-native startups are all competing in this space. If Chinese Web3 teams do not have particularly strong technological and resource accumulation, it is difficult to gain an advantage by directly diving into this area.

A more practical issue is that the advantages accumulated by Web3 teams in the past typically do not lie in model training. Many teams excel in protocols, data, wallets, payments, security, communities, and overseas markets. If they directly pivot to general large models, it means discarding their prior accumulations and entering a more burdensome, competitive, and costly track.

The second type involves immediately starting a general AI Agent platform.

Many Agent platforms sound grand and appear capable of performing any task. However, once implemented, users often care less about how large the platform is and more about whether specific tasks can be completed consistently. Can it integrate into real workflows? Can it reduce labor costs? Can it ensure result quality? Are there customers willing to pay? These questions are more important than “platform narratives.”

Without clear tasks, delivery standards, and paying customers, Agent platforms can easily stall at the demo stage. They may seem advanced, but it's challenging to make them part of users' daily routines.

The third type involves AI traders, automatic yield, and smart advisory directions.

These products can easily gain traction in the Web3 circle as they naturally align with users' expectations for returns. AI automated trading, AI helping you make money, AI making investment decisions for you—all sound very appealing.

However, the issues in this direction are also the most complex. They can easily touch on user funds, yield promises, asset management, advisory compliance, and trading risk control. Once the product expression is slightly aggressive, it may shift from "tool assistance" to "yield promise." For Chinese teams, these directions are particularly sensitive and challenging to pursue as a long-term stable entrepreneurial path.

The fourth type involves simply placing an AI shell around existing projects.

Originally focusing on NFTs, now adding AI generated images; originally working on GameFi, now including AI NPCs; originally managing wallets, now featuring an AI chat assistant; originally engaging communities, now adding an AI bot. Such transformations may create short-term buzz, but if they do not improve product value, it is difficult to retain users or convince actual payers.

AI can serve as an entry point for capability migration but cannot fundamentally resolve issues for a project with no real demand.

If the original business lacks users, revenue, and scenarios, simply changing the narrative to AI will most likely lead back to the same questions. Why do users need it? Who will continue to pay? What problem has the team truly solved?

Thus, for Chinese Web3 teams, determining whether an AI direction is worth pursuing cannot be based solely on its popularity. More crucially, it must have a real scenario, clear payers, reusable capabilities, and relatively clear compliance boundaries.

In Closing

The AI cycle has arrived, and Chinese Web3 teams certainly need to observe and should observe.

However, what truly deserves attention is not which concept is trending again, but whether the capabilities accumulated over the past few years have new applications.

From data, identity, payments, to security, risk control, and application-layer products, what Web3 teams can migrate are essentially the foundational aspects already rooted in their original businesses. AI offers new scenarios for these capabilities but will not substitute for a project with no genuine demand.

Thus, for Chinese Web3 entrepreneurs, transitioning to AI is not the key; the ability to migrate is.

If the past accumulation includes data, accounts, payments, security, operations, and user scenarios, then AI could represent a new pathway.

If there was only narrative and packaging in the past, merely switching to AI will only be a change to a hotter term.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。