AI Agent Invasion? What Experts Expect for 2025

AI
RESEARCH
AGENTS
ADOPTION
The AI Agent Invasion has started. We are in the early stages. Experts share their expectations and how you can prepare yourself.
Authors

Christina Garcia

Matthew Rothenberg

Published

January 19, 2025

Keywords

AI Agent Invasion, AI Invasion, Jobs and AI Agents, Prepare youreself for AI Agents, What are AI Agents, Sims vs Agents, AI Agents What Experts Expect for 2025, YouGot.us

Agents: The Kind of AI That Does Things Independently

If you haven’t heard about AI agents, that’s fine. Here’s your chance to catch up—and we’ve got news for you.

If you run a business and choose to ignore AI agents, that’s probably still okay—as long as your competition is equally lagging.

The short of the story is, that agents are coming, whether you like it or not. The best way to prepare is by educating yourself.

In November 2024, we decided to just do that. YouGot.us, in partnership with the tech community group MLOps Community and the tech company Prosus, brought together dozens of AI experts to discuss AI agents. More than 300 participants—mostly AI engineers, startup founders, data scientists, and even product managers, over half of them in senior or executive roles—shared their experiences and expectations with us.

What is Agentic AI?

But before we dive into our experts’ answers, let’s establish agentic AI really means. We like the definition from Thomas H. Davenport and Harry Bean from their article in January’s MIT Sloan report: “…the kind of AI that does tasks independently.” By that they mean “autonomous and collaborative AI programs” that are focused on “generative AI bots that will perform specific tasks.” But before we dive into our experts’ answers, let’s establish agentic AI really means. 

In other words, instead of disappearing down a rabbit hole in pursuit of answers—as large language models like ChatGPT 2, 3, and 4 often do—agents tackle a given task collaboratively. They use checks and balances, reassess their feedback, and take on different roles throughout the response process. Imagine two or three ChatGPTs, each assigned a different role, collectively figuring out how best to answer a question. Essentially, agentic AI attempts to be self-correcting by being collaborative.

The technology company Anthropic distinguishes between Agents and Workflows:

  • Workflows follow code paths and orchestration is predetermined.

  • Agents are autonomous, self-controlled and self-orchestrated, giving them greater independence.

Self-Orchestrated vs Self-Organized

We want to point out that agents are not yet self-organized, like living systems or autopoietic systems described by biologist and philosopher Francisco Varela. Autopoiesis refers to an organism’s capacity to renew and sustain itself continually, producing the very components that comprise its structure and processes. In contrast, AI agents still rely on external programming, maintenance, and updates from human developers or other automated systems. However, agents’ current ability to create code and thereby structure and function advances the technology to the verge of self-organized AI Agents.

The path to autonomous agents can be traced back to the Agent-Based-Modeling (ABM) approaches of the 1990s and early 2000s. Fabien Michel, Jacques Ferber and Alexis Drogoul describe the four key concepts of ABM in their 2008 article “Multi-Agent Systems and Simulation” as

  • “Autonomous activity of an agent, i.e. its ability to carry out an action on its own initiative”

  • “Sociability of agents, i.e. their ability to act with other agents (from a social point of view)”

  • “Interaction is what connect the two preceding concepts,” and

  • “Situatedness of clients, the fact that agents are placed into an environment which defines the conditions in which the agents act.”

Agents versus Sims

These principles lead to a crucial distinction: from a modern agentic AI perspective, the “clients” in many ABM scenarios behave more like “sims” than true agents. Sims are simulations of individuals with particular personalities, interests, and goals. By contrast, we employ LLM-powered agents to create these simulations—allowing us to study behavior and decision-making in a controlled, research-oriented setting.

So, let’s recap before we take a look into the future. We define agents as tools or workers rather than subjects of behavioral research, like Sims. According to Anthropic, agents are “tool-based LLMs” that comprehend complex inputs, engage in reasoning and planning, use these tools reliably, and recover from errors.

What Scientists Expect from Agentic AI in 2025

Agentic AI is a hot topic at Stanford’s Institute for Human-Centered AI (HAI), and its leading researchers are closely watching its systemic advance: “There’s some API with Anthropic’s Claude that can actually operate your computer—put a meeting on your calendar or help you buy a plane ticket,” says James Landay, HAI Co-Director and Professor of Computer Science. While Landay finds the current models useful, he also notes that fundamental breakthroughs may require more time.

Russ Altman, HAI Associate Director and Professor of Bioengineering, anticipates “general contractor”-style systems for problem-solving: “They’re made up of a bunch of AI systems that talk to each other. In some cases, they may negotiate with one another; in others, they hand off tasks to ‘expert LLMs,’ which then return answers.”

Meanwhile, James Zou, Associate Professor of Biomedical Data Science, sees “a significant shift from relying on individual AI models to using systems where multiple AI agents, each with diverse expertise, work together.”

3 Key Takeaways from our 300 AI Experts

Our group of AI experts is overwhelmingly composed of AI practitioners—only about 6 percent identify as researchers. Their takeaways focus on the current state of organizational interest, technical feasibility, and value generation.

1. Tech company workers have to adopt the quickest

We asked our group of AI practitioners—segmented by the size of the companies they work for—about their expectations for AI agent adoption in 2025. Their responses were clear: 60% of early-stage technology startups anticipate extensive or full adoption by then. While this may not be surprising—many startups have only just begun leveraging LLMs—the competition is moving quickly.

If you work at a large technology company, chances are AI agent adoption is already well under way. As a tech worker, you’ll need to adopt faster than the rest of the industry. Ignoring AI agents is not an option. It may also be time to reconsider your role: if your work can be easily replicated by AI agents, you’ll need to “level up” and become a manager of agents, rather than their competitor.

Our advice: If you’re in tech, now is the time to adopt AI agents. Stop competing—start managing.

2. The Second Wave will Hit Vertical Industries

If you think other sectors are lagging, think again. The pace of AI agent adoption is only accelerating—particularly in healthcareretail/e-commerce, and telecommunications, which are now catching up to the tech industry. Vertical adoption of AI agents looks poised to be especially strong.

Anshika Mathews highlights the rise of “vertically-focused agents” and how “industries burdened by repetitive, manpower-intensive tasks” will see increased adoption by 2025:

“Vertical AI Agents excel where general-purpose tools fall short. Designed to automate entire processes, they’re bringing precision and efficiency to tasks that have long relied on human effort. From optimizing debt collection strategies to revolutionizing customer support, these systems are engineered for impact.”

Startups like AprioraSalient, and Sweetspot are already riding this wave.

Our advice: Be part of the wave. Learn how to leverage vertical agents for greater productivity.

3. Focus on Efficiency and Cost Savings

AI agents can be used for lots of tasks, especially repetitive ones. But AI competition is not stopping here. Agents will be expected to make data-driven decisions, to enhance customer experience, to help gain competitive advantages, to build better products and offer better services and most importantly to cut costs and improve efficiency.

Our advice: Leverage AI agents and work with internal agentic teams to help reduce operational costs and drive efficiency.

Lessons for the Rest of Us

The rise of AI agents presents new opportunities for the human beings in your enterprise. Instead of competing with agents, your team can leverage its knowledge and skills in the precise areas where AI agents come up short. Here is how you can start riding the wave:

Embrace Automation for Routine Work

  • Why? Offloading mundane or repetitive tasks to AI agents frees you to focus on higher-value activities.

  • How? Identify small tasks first (e.g., data entry or basic customer queries). Learn how to delegate these to AI tools so you can invest more energy in creative or strategic projects.

Focus on Unique Human Strengths

  • Why? Skills like critical thinking, empathy, leadership, and complex problem-solving are still difficult for AI to replicate.

  • How? Hone these abilities through ongoing education, mentorship, and real-world experience. The more you can do what AI cannot, the more indispensable you become.

Learn to Work Alongside AI Agents

  • Why? AI agents can be collaborators—not just tools you “use.”

  • How? Treat them as part of the team. Ask for input, experiment with their suggestions, and refine their outputs to improve both your workflow and their performance.

Broaden Your Skill Set

  • Why? As AI agents grow more capable, your future roles will likely involve overseeing or guiding them.

  • How? Pick up at least a foundational understanding of how AI tools work. This could mean learning prompt engineering, data analysis, or basic programming concepts.

Track and Demonstrate Value

  • Why? Showing measurable impact keeps you relevant.

  • How? Whether you save time, improve customer satisfaction, or boost sales with AI assistance, gather data on these achievements and share them with your team or management.

Develop a Growth Mindset

  • Why? AI automation can be unsettling if you feel left behind. Approaching new technology as an opportunity rather than a threat will serve you well.

  • How? Maintain curiosity, welcome feedback, and treat mistakes as learning experiences—both for you and for the AI agents you train or manage.

Maintain Ethical and Responsible Practices

  • Why? As AI’s presence grows, ethical use and data integrity become even more critical.

  • How? Familiarize yourself with your organization’s policies on data usage and AI governance. Ensure you handle customer or company data responsibly.

Key takeaway: Don’t see AI agents as competitors; see them as catalysts for your own professional growth. By automating repetitive tasks, focusing on human-centric skills, and actively learning to collaborate with AI, you can future-proof your career and increase your value within any organization.

Citation

BibTeX citation:
@online{garcia2025,
  author = {Garcia, Christina and Rothenberg, Matthew},
  title = {AI {Agent} {Invasion?} {What} {Experts} {Expect} for 2025},
  date = {2025-01-19},
  url = {https://yougot.us/news/2025-01-19-Agent-Invasion-2025/},
  langid = {en}
}
For attribution, please cite this work as:
Garcia, Christina, and Matthew Rothenberg. 2025. “AI Agent Invasion? What Experts Expect for 2025.” January 19, 2025. https://yougot.us/news/2025-01-19-Agent-Invasion-2025/.