Perplexity is giving its Computer AI agent a more persistent role in ongoing workflows. The company is expanding Computer’s automation capabilities so recurring tasks can continue working from the context built up during previous sessions rather than treating every run as an entirely new assignment.

That’s an important distinction for AI agents. A scheduled task is useful when it can run on its own, but it becomes considerably more useful when it understands what happened the last time it ran.

Perplexity’s Computer already supports background and recurring workflows, but the company’s current Projects setup is designed to give those workflows a persistent home. Projects combine conversations, files, connected applications, automations and Perplexity’s Brain memory system. Brain can study project files and previous sessions between tasks, allowing subsequent Computer tasks to start with accumulated context.

In practical terms, that means a recurring workflow can keep moving instead of repeatedly asking the user to explain the same project.

For example, a user could set up an automation to monitor developments in a particular area, prepare a recurring report, maintain a status document or gather information from connected services. Computer can run that work on a schedule and use the surrounding Project context to understand what it has already been doing.

Perplexity says Projects can run recurring tasks hourly, daily or weekly. Computer can monitor sources, publish summaries and maintain status documents in the background.

The persistent context is particularly important here. Projects have a shared file system where Computer can work with uploaded documents, imported folders and information pulled from connected sources. The Project’s Brain then maintains knowledge accumulated from files and sessions.

That creates a workflow closer to an ongoing research assistant than a traditional chatbot.

Imagine using Computer for a weekly competitive-intelligence report. On the first run, it might gather the relevant information and create a report. On subsequent runs, the automation can return to the same work, check for new developments and update the existing material rather than starting over from scratch.

The same approach can apply to project updates, research monitoring and other repetitive knowledge-work tasks. Perplexity specifically positions Projects around use cases such as recurring status updates, monitoring sources and preparing drafts for review.

This also fits with Perplexity’s broader push to turn Computer into an agent platform rather than simply an AI interface. In recent months, Computer has gained persistent project memory, connected applications, reusable Skills, background workflows and access through services such as Slack and Microsoft Teams.

Perplexity has also continued expanding the underlying Computer system. Its September updates added effort controls, additional model options and other ways to configure how Computer handles different tasks. The company’s changelog says Computer can now use GPT-6 Astra for eligible subscribers and lets users adjust the effort level from Light through Ultra.

The bigger shift, though, is persistence.

AI assistants have traditionally been good at completing individual requests. The harder problem is getting an agent to work on something over days or weeks without forcing the user to reconstruct the entire context every time.

Perplexity’s combination of Projects, Brain and scheduled automations is an attempt to solve exactly that problem. Instead of “run this task again tomorrow,” the idea is closer to “keep working on this.”

That could make recurring AI workflows considerably more useful, particularly for research, reporting, monitoring and other work where the latest result depends on everything that happened previously.

And as AI agents become more autonomous, remembering previous work may ultimately matter just as much as being able to perform the work in the first place.