Philosophy

First

In June 2025, Stanford University's "SALT Lab" released a report titled "Future of Work with AI Agents: Auditing Automation and Augmentation Potential across the U.S. Workforce".

With the rapid development of "composite AI systems" (i.e. AI Agents), more and more professional tasks are facing the possibility of being automated or assisted by AI. However, the alignment between academia and industry on "what workers want AI to do" and "what current technology can do" is still unclear.

For 104 occupations and 1,500 working Americans, they collected their "automation/assistance preferences" for various tasks and their expectations for human involvement. The results were divided into four areas according to "expectation-ability":

*Green Light Zone: automation is desired and the technology is mature;

* Red Light Zone: automation is not desired, but the technology can do it;

* R&D Opportunity Zone: automation is desired, but the technology is not yet mature;

* Low Priority Zone: automation is not a concern and less investment is made in technology development.

The results are summarized in the following 4 points:

1. There is a major mismatch

41.0% of Y Combinator startups/projects focus on the "low priority zone" and "automation red light zone", which is significantly out of touch with the real needs of workers.

2. Willingness for human-machine collaboration

Most professional tasks are not simply "automated or not", but require equal cooperation with AI Agents. Workers as a whole tend to have higher human participation, and are worried that leaving it entirely to AI will bring about issues of control and responsibility.

3. Task impact scope

The study estimates that about 80% of American workers will have at least 10% of their work tasks affected by large language models (LLMs); 19% of workers may have more than half of their tasks potentially changed..

4. Core competency transformation

As technology takes on more "information processing" tasks, soft skills such as interpersonal communication, creativity, and ecosystem thinking will become competitive advantages in the future workplace.

Against the background of this research result, what role the start-ups play, what track they focus on, what practical problems they solve, and what role they play in the problem-solving process will be the key to how far the start-ups can go in this AI revolution.

Against the background revealed by the results of this study, how far the start-ups can go in this AI revolution depends on how they position their roles, choose to focus on the track, and actually solve real problems. Specifically, if companies can take the role of "human-oriented technology enabler" and focus on the "R&D opportunity zone" first, they will not only have differentiated competitiveness, but also be more likely to win user trust and market recognition.

This means that start-ups need to deeply understand the expectations of the workforce in various industries for AI, and focus on those areas with great potential for human-machine collaboration but have not yet been fully met by technology, such as professional decision-making assistance, copywriting draft generation, and customer dialogue assistance. At the same time, companies should also avoid blindly pursuing the automation of "red-light district" tasks, that is, application scenarios that are technically feasible but seriously conflict with workers' wishes.

In this process, start-ups will play the role of "reconciler" between technical capabilities and actual needs. On the one hand, they will promote the breakthrough of AI agent capabilities, and on the other hand, they will also take on the responsibility of educating the market, building user trust, and shaping a human-machine symbiotic working model.

Ultimately, the AI ​​start-ups that can truly cross the cycle will not be the "most powerful in technology", but the technology implementers who can best understand, respect, and serve people.