Perplexity Teammates: Workplace Agents That Run on Goals, Not Instructions

Perplexity has introduced Teammates, continuously running workplace agents that take goals rather than instructions, available for now on the web to the most expensive Max subscribers. The agents run in an isolated environment, take over existing projects, skills and memory, and also carry out Perplexity Computer tasks. Chief executive Aravind Srinivas says about ten thousand such agents already work on the company's internal workspace, twenty times the number of employees. The news was reported by eYssen AI news on November 5, 2026.

Goals, Not Instructions

The distinction between goals and instructions is the heart of the Teammates idea. A traditional assistant waits to be told what to do, step by step, and then does exactly that. A Teammate is given an outcome to achieve and is expected to work out the steps itself, keep going over time, and keep the project moving without being prompted at every turn. That is a different kind of tool, and it asks a different kind of trust from the user. Instead of checking each action, the user defines the destination and lets the agent find the route.

For workplaces, the shift matters because so much office work is made of ongoing projects rather than one-off tasks. A research project, a competitive monitoring brief, or a content pipeline does not end when a single question is answered; it continues for weeks and months, with new information arriving all the time. An agent that runs continuously and pursues a goal fits that shape of work far better than a chatbot that forgets everything when the conversation ends. The question for teams will be whether the agent's judgement is good enough to be left running, and how much oversight it still needs.

An Isolated Workspace That Inherits Your Work

Teammates run in an isolated environment, which is an important detail for anyone worried about an agent wandering through their files. Isolation means the agent works in its own space rather than directly inside a person's live systems, which gives the user a layer of separation between the agent's activity and their own work. At the same time, Teammates take over existing projects, skills and memory, so they are not starting from scratch. They inherit the context a user has already built up, including the projects they are working on and the skills and memory attached to them, and they can also carry out Perplexity Computer tasks.

That combination of inheritance and isolation is what makes the product interesting and also what makes it worth watching carefully. An agent that knows your projects and your history can do far more useful work than one that starts cold every time, but it also has more to get wrong. The fact that Teammates are available first to the most expensive Max subscribers suggests Perplexity is rolling the feature out to its most committed users before widening access. Those early users will effectively be testing whether continuously running agents can be trusted with real workplace projects.

What It Means for Search and Content Teams

For SEO and content teams, Teammates point to a future in which research and monitoring are delegated to agents rather than done by hand. A Teammate given the goal of tracking a topic, a competitor, or a set of keywords could work through the material continuously, surfacing changes as they happen rather than waiting for someone to run a report. That could change the rhythm of search work, moving it from periodic checks to constant awareness. It could also change who does the work, as junior research tasks are handed to agents and humans focus on strategy and judgement.

The internal numbers underline how seriously Perplexity takes this direction. About ten thousand such agents already work on the company's internal workspace, twenty times the number of employees, which means the company is already running an agent workforce far larger than its human one. That is a striking claim, and it suggests Perplexity sees agents not as a side feature but as the main way work gets done inside the company. Whether other workplaces follow that path will depend on what these early Teammates actually deliver, but the direction of travel is now clear: the agent is becoming a colleague, not just a tool.

Sources

Sources: eYssen AI news, November 5, 2026.