The world of AI is moving beyond simple chatbots that answer one-off questions. The next frontier is autonomous agents that manage ongoing work. Perplexity's new 'Teammates' aim to do just that, offering an always-on assistant for complex projects.
What Are 'Always-On' AI Agents?
Imagine
an assistant who doesn't just answer a question and then forget the context, but one who takes on a long-term assignment and works on it 24/7 in the background. That's the core idea behind Perplexity's new feature, called 'Teammates'. Unlike conventional AI tools that require a new prompt for every task, these agents are designed to be persistent. You can assign them a role, like 'Market Researcher' or 'Sales Analyst', and a goal. The agent will then independently plan and execute the necessary steps, remembering the project's history, files, and your feedback across multiple sessions. It's a shift from AI as a reactive tool to AI as a proactive team member that manages workflows and only notifies you when a decision is needed.
From Single Queries to Complex Projects
The true purpose of these agents is to tackle 'long-running tasks'—the kind of work that spans days or weeks. For example, you could assign an agent to constantly monitor your top three competitors for any price changes, product launches, or new executive hires, and deliver a summary report every Friday. In a sales context, an agent could be tasked with identifying customers whose engagement has dropped, reviewing their order history, flagging unresolved support tickets, and then drafting a personalised follow-up email for your approval. Other potential roles include managing a social media campaign, triaging engineering bug reports, or even assisting with legal research by comparing a new contract against previous agreements and flagging discrepancies.
How Does It All Work?
These agents operate within Perplexity's existing cloud-based platform, called Perplexity Computer. Users set up a 'Teammate' by giving it a name, a role, and a specific job. You then connect it to the necessary tools it needs to perform its work, such as Slack, Gmail, or GitHub. The system is designed for multi-agent collaboration, meaning multiple Teammates can work together, coordinate, and even assign tasks to one another to complete a larger project. Throughout the process, each agent maintains its own thread where a user can monitor its progress, see what it has completed, and provide new instructions or feedback. The system is built to learn from these interactions, improving its performance on future tasks.
A Glimpse into the Future of Work?
This move by Perplexity, along with similar developments from rivals like OpenAI and Anthropic, signals a significant evolution in workplace AI. The focus is shifting from simply accelerating individual tasks to automating entire workflows. By delegating recurring, process-driven work to AI, the goal is to free up human employees to concentrate on strategy, critical decision_making, and final approvals. For many professionals, this could mean less time spent on manual data gathering and report generation, and more time focused on higher_value analysis and creativity. The introduction of agents that can operate 24/7 also brings the potential for increased efficiency and continuous optimisation of business processes.
The High Cost and Lingering Questions
While the promise is significant, this advanced capability comes at a price. Access to Teammates is currently exclusive to Perplexity's top_tier 'Max' subscription, which costs around $200 per month or $2,000 annually. This positions it as a premium tool for enterprise teams and power users, rather than a casual productivity aid. Furthermore, the rise of autonomous agents introduces critical questions about data security, privacy, and governance. If an agent has persistent access to your email, company files, and internal communications, ensuring it acts only within its designated boundaries is paramount. As with any powerful new technology, establishing trust and clear oversight will be just as important as the agent's ability to perform its tasks correctly.
