Perplexity turned AI search into an agent that builds Excel files and checks data on its own

Perplexity Computer demonstrated how an AI tool can move beyond the role of "smart search" to become an executor. In a test on a real research task, the system assembled a four-sheet Excel file in 7 minutes 7 seconds, cited 33 sources, and separately flagged ten discrepancies in benchmarks.

How the Test Went

The user asked the system to collect all AI models released from March 7 to 14, 2026, find benchmark results for each from at least three independent sources, mark discrepancies in the numbers, and then calculate the price-to-performance ratio.

The output was not simply a list of links, but a working file with a ranking of seven models, a list of disputed metrics, and explanations for why different values appear in different publications.

The most interesting part was not the Excel itself, but the assembly method. Computer did not copy text from the browser into cells manually or condense the result into one long answer. It wrote a Python script build_spreadsheet.py, ran it, encountered errors, fixed them itself, and reassembled the file. In other words, the system went beyond search and summarization: it created an artifact that can be opened, verified, and immediately used in work.

Why This Isn't Search

According to Perplexity's description, Computer is an agentic layer launched on February 25, 2026, on top of a set of 19 models. Claude Opus 4.6 handles central orchestration and coding tasks, deep research goes to Gemini, quick factual checks go to Grok, and specific media tasks are routed to other models.

The user does not give a sequence of commands but a goal, after which the system itself breaks it down into subtasks, distributes the work, and assembles the answer into a final result.

"Perplexity is no longer a search engine.

It's a different category of tool."

The test demonstrated this well in its details. For example, Computer found a gap in GPT-5.4's GPQA Diamond rating: 92.8% in OpenAI's blog versus 74.8% from NxCode, and linked the discrepancy to different reasoning effort settings. Even more useful was the cost conclusion: NVIDIA's Nemotron 3 Super at $0.10/$0.50 per million tokens delivered a cost-to-performance ratio of $0.37 compared to $9.17 for GPT-5.4. And the system calculated this itself, without a separate request.

What Was Announced Next

On March 11, 2026, at the Ask 2026 conference, the company demonstrated that Computer is not a one-off demo feature, but part of a broader product. In the personal version, the agent can run on a dedicated Mac mini, have persistent access to local files and connected services, monitor triggers, and launch tasks in the background. Sensitive actions require user confirmation, and an emergency switch provides quick manual control.

In parallel, the company showed how it plans to monetize and expand this approach. Access to Personal Computer is open through a waitlist for the $200/month Max plan, and the enterprise version can already connect to Snowflake, Databricks, Salesforce, and HubSpot. For developers, the company released the same internal components that power Computer, and added paid databases and mobile synchronization to the research loop.

Key points of the announcement are:

• Personal Computer with persistent access to files, applications, and sessions
• Enterprise version with connectors to Snowflake, Databricks, Salesforce, and HubSpot
• Four APIs for developers: Search, Agent, Embeddings, and Sandbox
• Direct access to Statista, CB Insights, and PitchBook data within research
• Launch on iOS with cross-device synchronization and subsequent Android release

What This Means

The main shift here is that the value of AI is increasingly measured not by a beautiful chat answer but by a ready-to-use work result: a table, a report, a calculation, a list of disputed data. Perplexity still has weaknesses—an opaque credit system, dependence on Apple's ecosystem for the personal version, and unverified enterprise figures. But the trajectory itself is clear: AI search is turning into a research agent that not only finds but also delivers.