AI agents are taking over complex enterprise tasks

New credentials from Perplexity reveal how AI agents are driving workflow efficiency gains by taking on complex enterprise tasks.

Over the past year, the technology sector has operated on the assumption that the next evolution of generative AI will advance beyond conversation to action. While large language models (LLMs) act as reasoning engines, “agents” act as hands capable of executing complex, multi-step workflows with minimal supervision.

However, until now, the view on how these tools might actually be used in wildlife has been vague, largely based on speculative frameworks or limited surveys.

New data from Perplexity, which analyzes hundreds of millions of interactions with the Comet browser and Assistant, provides the first large-scale field study of general-purpose AI agents. Data suggests that agentic AI is already being deployed by high-value knowledge workers to streamline productivity and research tasks.

Understanding who is using these tools is essential for forecasting internal demand and identifying potential shadow IT vectors. The study reveals notable heterogeneity in adoption. Users in countries with high per capita GDP and educational attainment are more likely to interact with proxy tools.

Most important for institutional planning is professional collapse. Adoption is highly concentrated in digital and knowledge-intensive sectors. The “Digital Technology” group represents the largest share, representing 28 percent of users and 30 percent of inquiries. This is closely followed by academia, finance, marketing and entrepreneurship.

Together, these groups represent more than 70 percent of all adopters. This suggests that the individuals most likely to benefit from a proxy workflow are the most valuable assets within an organization: software engineers, financial analysts, and market strategists. These early adopters don’t work. Data shows that “power users” (those with early access) make nine times as many proxy queries as regular users, suggesting that once technology is integrated into their workflow, technology becomes indispensable.

AI agents: Partners in enterprise tasks, not servants

To move beyond marketing narratives, companies must understand the benefit these agents provide. A common view suggests that agents will primarily act as “digital gatekeepers” doing routine administrative work. However, data challenges this view: 57% of all agent activities focus on cognitive work.

Perplexity researchers developed a “hierarchical taxonomy of agents” to classify user intent, revealing that the use of AI agents is practical rather than experimental. The predominant use case is “productivity and workflow,” which accounts for 36 percent of all agent queries. Followed by “learning and research” at 21 percent.

Specific anecdotes from the study illustrate how this translates into enterprise value. For example, a procurement professional used Assistant to scan customer case studies and identify relevant use cases before engaging with the vendor. Likewise, a financial agent delegated the tasks of liquidating stock options and analyzing investment information. In these scenarios, the agent handles the information gathering and initial synthesis independently to allow the human to focus on the final judgment.

This breakdown provides a specific signal to operations leaders: The immediate ROI for agentic AI is to scale human capabilities rather than simply automate low-level friction. The study defines these agents as systems that “automatically cycle between three iterative stages to achieve the ultimate goal: thinking, acting, and observing.” This ability allows them to support “deep cognitive work,” acting as a thought partner rather than a simple servant.

Cognitive adhesion and migration

One of the key insights for IT leaders is to “stick” AI agents into an organization’s workflow. The data shows that in the short term, users show strong persistence within the topic. If a user engages an agent for a production task, their subsequent queries are very likely to remain in that domain.

However, the user journey often evolves. New users often “test the waters” with low-stakes inquiries, such as asking for movie recommendations or general information. Over time, transformation occurs. The study suggests that although users may access across different use cases, query posts tend to move toward cognitively oriented areas such as productivity, learning, and career development.

Once a user assigns an agent to debug code or summarize a financial report, they rarely return to lower-value tasks. The “Productivity” and “Workflow” categories show the highest employee retention rates. This behavior suggests that early pilots should anticipate a learning curve as use matures from simple information retrieval to complex task delegation.

The “where” of agentic AI is just as important as the “what.” The Perplexity study tracked the environments – specific locations and platforms – where these AI agents operate. The focus of activity varies by mission, but senior environments are the staples of the modern enterprise suite.

Google Docs is a primary environment for editing documents and spreadsheets, while LinkedIn dominates professional networking tasks. For Learning and Research, activity is split between course platforms such as Coursera and research repositories.

For CISOs and compliance officers, this represents a new risk profile. AI agents don’t just read data; They are actively dealing with it within core enterprise applications. The study explicitly defines proxy queries as those that involve “browser control” or actions on external applications via APIs. When an employee assigns an agent to “summarize customer case studies,” the agent interacts directly with the private data.

The focus of the environments also highlights the potential for platform-specific improvements. For example, the top five environments account for 96 percent of inquiries in professional networks, primarily on LinkedIn. This high focus indicates that companies can achieve immediate efficiency gains by developing specific governance policies or API connectors for these high-traffic platforms.

Business planning for agentic AI that follows Perplexity data

The proliferation of capable AI agents calls for new lines of research for business planning. Data from Perplexity confirms that we are past the speculative phase. Agents are currently used to plan and execute multi-step actions, modifying their environments rather than simply exchanging information.

Executive leaders should consider three immediate actions:

1. Productivity and workflow audit Friction points Within high-value teams: Data shows that this is where agents naturally find their footing. If software engineers and financial analysts are already using these tools to edit documents or manage accounts, formalizing this workflow can lead to consolidated efficiency gains.

2. Getting ready for augmented reality: The researchers note that although agents are autonomous, users often break tasks into smaller parts, delegating only subtasks. This suggests that the near future of work will be collaborative, requiring employees to hone their skills in how to effectively “manage” their AI counterparts.

3. Infrastructure processing and security layer: With agents operating in “open-world web environments” and interacting with sites like GitHub and corporate email, the DLP landscape is expanding. Policies should distinguish between a chatbot that gives advice and an agent that executes code or sends messages.

As the agent AI market is expected to grow from $8 billion in 2025 to $199 billion by 2034, early evidence from Perplexity is leadership. The shift to enterprise workflows driven by AI agents is underway, driven by more digitally capable segments of the workforce. The challenge for the organization is to harness this momentum without losing control of the governance needed to scale it safely.

See also: Accenture and Anthropic partner to advance enterprise AI integration

Want to learn more about AI and Big Data from industry leaders? Check out the Artificial Intelligence and Big Data Expo taking place in Amsterdam, California and London. This comprehensive event is part of TechEx and is co-located with other leading technology events. Click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.