AI companies are racing to automate everything from writing code and creating images to scheduling advertisements, summarizing meetings, and much more. However, as these systems become increasingly intelligent, the impact on human jobs is becoming more apparent. Some experts warn that AI-generated jobs could trigger a wave of large-scale job losses that are faster and more widespread than current economies are prepared for. XEM details here .
Instead of fighting against the future, a platform originating from the crypto world is betting on a different path. If automation is inevitable, then so should ownership.
Action Model has just launched an invite-only Chrome extension that allows users to train AI systems by Chia real-world browser activity such as clicks, page navigation, data entry, and task execution. The platform calls this a Large Action Model (LAM) – not only creating content but also learning how to handle digital tasks. In return, contributors receive points, which can be converted into Governance Token – representing the right to participate in deciding the system's development.
"If AI is going to replace digital labor, then workers should be the owners of these 'machines'," Chia Sina Yamani, founder of Action Model.
Training AI to perform tasks
Unlike chatbots that only generate content, LAMs are designed to run software directly. The idea is simple: If humans can perform a digital task using a mouse and keyboard, then trained AI should be able to do the same.
“A few years ago it was the era of chatbots. Now it’s automation,” Yamani said. “There are now about a billion people working with computers. If a company has a tool that helps automate tasks continuously, saving a significant amount of money compared to hiring staff, they will definitely use it.”
Action Model's utility collects real-world operation data (which users choose to Chia ) to train the AI. Tasks such as payroll processing, CRM management, or basic operations can be recorded once and repeated by the AI later. Contributors can also Chia their automation processes on a public marketplace, where all activities are tracked and rewarded according to the platform's incentive model.
Agent AI is becoming increasingly prevalent in the industry, as models gradually shift from content creation to automated real-world task processing. These systems – as explained in this article – are constantly learning how to interact with and navigate the digital environment using real user data.
The platform has now attracted over 40,000 registered users through a waiting list, friend referrals, and a partner community. Access remains invitation-only to ensure the quality of contributors and to reward early participants.
How is this different from existing automation tools?
Most current automation tools rely on APIs or rigid integrations. However, the vast majority of real-world digital tasks take place on legacy systems, internal dashboards, or tools that were never thought possible to automate.
“Zapier automates software. We automate the work,” Yamani said. “Only about 2% of internet data is actually accessible through APIs. The remaining 98% still requires human intervention.”
With Action Model, users don't need to know code or complex integration. They simply record their work process as usual. The AI will learn that real-world process and then replicate it independently.
As a result, Action Models are flexible in capturing special cases and processes that have never been documented – something that traditional automation systems cannot reach.
What about privacy?
All training data is user-selected. Sensitive pages such as email, health, or banking are blocked by default. Users can pause training, block specific websites, or completely delete their contributions.
“The first principle is very simple: We don’t need your personal data, just a working model,” Yamani said. “The data is processed locally and anonymized before being used to train the AI.”
Once deleted, data is completely lost; even the company cannot recover it. Your contributions will be aggregated with data from many other users, applying k-anonymity to prevent individual retrieval. The dashboard allows you to track and manage your entire training history and rewards at any time.
"While Big Tech often collects this type of data without asking for consent, we are transparent, users have active control, and those who provide data to the AI will also receive real rewards," Yamani added.
So, can bots manipulate the system?
To avoid the common problems of previous crypto reward systems, the Action Model uses behavioral analysis to verify actual user actions. The system looks at the structure of tasks, timing, changes, and decisions – elements that automated bots or click farms would find difficult to fake.
"Just clicking randomly doesn't make any sense," Yamani Chia . "The actual process is always purposeful, involving stopping, correcting mistakes, trying again, and deliberating. You can't fake this on a large scale."
Several projects that previously rewarded interactions and posts have been banned from major platforms due to the massive amount of AI spam, bot replies, and fake interactions. As a consequence, APIs were locked and the Token ecosystem collapsed due to poor quality performance.
ActionFi – the platform's reward system – is designed to completely eliminate this risk. ActionFi doesn't reward tweets or clicks, but only real, structured workflows performed by real people.
"We don't reward noise. We only reward links that are genuinely helpful," Yamani added.
Who actually owns the system?
Currently, Action Model controls the utility, training logic, and reward system. However, the project is committed to gradually transferring this ownership to the Token holder community over time. A DAO (decentralized autonomous organization) structure will be implemented, allowing all participants to contribute to the platform's direction, reward mechanism, and model implementation.
"Coordination is always necessary in the initial stages, but the important thing is whether the design itself focuses on concentrating power," Yamani said.
If implemented as advertised, ownership will allow Token holder to truly influence decision-making regarding the infrastructure connected to the very data they helped build.
If AI is inevitable, will ownership also become so?
The next generation of AI will not only build on language but also on human labor. From office work to operations, a wide range of behind-the-scenes tasks are now within the reach of intelligent agents.
“You’ve heard that millions of screen jobs will be automated. That’s not in a few decades – it’s starting now,” Yamani said. “If your data contributes to training AI, you deserve to own the results.”
Whether Action Model can scale, maintain transparency, and build a sustainable economy remains to be seen in the coming months. But the project's direction is clear. The big question with AI isn't just what it can do, but who will benefit from it.
As AI is changing the way we work, will the future belong to large platforms, or to humans?
