How will our work evolve and even exist with the advent of AI agents than it is now? But let me pay up front that AI tools will not change the basic work of PM. This is to identify key issues to solve and guide the best ideas for implementation. AI agents definitely augment and in some cases you can exchange certain activities, but that's a good thing.
Don't succumb to the vigilante stories of how your work can be adversely affected. Each PM's role is unique. Although we share general aspects, the creation of product concepts, definition of requirements, iteration with customers, GTM and social media PM's day-to-day operations are very different from the work of cloud infrastructure PM, and many aspects need to be automated. As a minitheo for your product, you alone will decide what you need for success. Therefore, you must be the one who decides how your work will evolve and make your product successful. You're in the driver's seat and choose something to augment or automate with an AI agent to do the job better. A recent Stanford research paper defines a useful framework for making these decisions and reveals that workers' desire for automation is a more critical factor in success than mere technical feasibility.
A human-centered framework for AI adoption
Stanford's research sheds light on how AI agents can help them with their work. We introduce a human-centric automation matrix, the desire of 2×2 plot workers for AI capabilities, which helps prioritize AI automation for PM tasks. It emphasizes that workers want to automate boring and repetitive tasks, but are deeply concerned about losing control and agency. The vast majority of workers in the study were concerned about the accuracy and reliability of AI, fearing unemployment and fearing lack of surveillance as another concern. Case studies in highlighting the risks of full autonomy are the recent issues of wiping away the entire company's database, creating data that hides bugs and ultimately apologises, and apologizing (see FastCompany).
This trust deficit logically excludes completely autonomous AI for high-stakes communication with customer and vendor communication. It is to clarify that AI plays a partnership or supportive role. This paper introduces the size (has) of human institutions to measure the degree of automation (see Autonomous Levels of Self-Driving Vehicles):
- H1 (no human involvement): AI agents operate completely autonomously.
- H2 (High Automation): AI requires minimal human surveillance.
- H3 (Equal Partner): Humans and AI have equal involvement.
- H4 (Partial Automation): AI is a tool that requires important human orientation.
- H5 (essential for human involvement): AI is a component that cannot function without continuous human input.
Most workers are fairly comfortable in the H3-H5 range and prefer that AI is a partner or tool rather than an alternative. PM decisions are not just about what to automate, but also about giving up control of AI agents.
This concept is better explained using the 2×2 matrix with automation capabilities on the X-axis and the desire for automation on the Y-axis. The four quadrants are classified as follows:
- Green Light Zone: High automation desire and high ability
- Red light zone: low desires and high abilities
- R&D Opportunity Zone: High desire, but low ability
- Low priority zone: low desires and low ability
This framework will help you determine which jobs are possible and will likely be hired in the workplace.
Run the framework
Instead of blindly following the mission of “using AI agents,” PMS needs to do what they do best. Think strategically about what's best for your business. Use this 2×2 to identify the ripe areas of automation that impact the most, and keep your team happy and productive.
- Green Light Zone: These are top priorities. Automating market insights, integrating customer feedback, and generating the first draft of PRD is technically feasible and highly desirable task. They save time, reduce cognitive load and free you to do higher levels of strategic work.
- Red Light Zone: Carefully proceed. AI can automatically generate marketing collateral, manage customer communications, and handle vendor contracts, but PMS is not ready to give up control of these high-stakes tasks. Errors can have serious consequences, and enhancement (H3-H4 is scale H3-H4) can be the correct option.
- R&D Zone: You need to innovate to prepare the technology to automate jobs. There is a high desire for automation, but technology is not ready, but it requires more investment to get there.
The most important thing is to take charge. The PM to engineer ratio won't improve anytime soon. Adding agent functionality to Toolkit is your best bet to scale impact. But drive with caution. To thrive and make yourself essential, you must be the one who shapes the future of your role.
Key takeout
- Prioritize desires over feasibility: Human-centered automation matrices are powerful tools. Enhance traditional tools (e.g. impact/effort, rice, kano) by considering recruitment and trust rather than just ability. The real success is building AI tools that your team actually uses.
- Think about agents, not just automation. The human agency scale (H1-H5) is used to determine the level of automation. Data-rich, repetitive PM tasks (e.g., discovering market insights, prioritizing databases) are categorized into the “green light” zone for high workers' desires and preparation for AI. These are also inputs into decision making, so necessary checks and balances have already been introduced in subsequent steps. Others may be categorized as H4 only because they are simply tools. This approach helps in managing risk and building trust.
- Focus on augmentation in high stakes areas: creative, strategy, or customer-facing tasks (aka “red light” opportunities) are in good agreement with augmentation strategies. AI generates options, analyzes data and provides insights, but ultimate decisions and communication must remain human.
- Core PM skills are more valuable than ever. AI agents handle more information-focused activities. We need to further develop unique human skills such as strategic thinking, empathy, stakeholder management, and organizational leadership.
The future of product management will be shaped not only by AI capabilities but also by advanced PMS choices. The most successful and adopted approach is human-centric and focuses on what PMS really needs to be superior. Those who acquire this strategic partnership with AI will not only survive, but will define the future of their role.
reference
[1] Y. Shao, H. Zope, et al. (2025). “The Future of Working with AI Agents: Auditing the Potential of Automation and Augmentation across the US Workforce.” arxiv preprint arxiv: 2506.06576v2. https://arxiv.org/abs/2506.06576
[2] S. Lynch (2025). “What workers really want from AI,” Stanford Report. https://news.stanford.edu/stories/2025/07/what-workers-really-want-from-ai
