Author: Damian Player; Compiler: Peggy, BlockBeats

Editor's note: While most people still think of AI as "a more efficient search tool," Perplexity is getting to work.

This article focuses on a difference that has been repeatedly ignored-why, when using AI, some people only get a paragraph of answers, while others directly get deliverable results. The key is not the model capability, but the way of using it: whether to use the tool as a dialogue window or as an execution system that can be commanded and scheduled.

The new class of tools represented by Perplexity Computer replaces "questions" with "tasks" as the core interaction method. From contract review, competitive product analysis, to data cleaning and report generation, users no longer describe problems, but directly define the final deliverables. Coupled with connecting enterprise tools and solidifying personal background and style samples, this ability has further evolved from one-time output to a reusable and automatically run workflow.

More importantly, the boundaries of automation are being redefined. It no longer just assists in completing a certain step, but can run continuously, perform across tools, and even proactively propose supplementary tasks. This means that the relationship between people and tools is shifting from "use" to "management and delegation."

Under this change, the real watershed is no longer whether to use AI, but whether to start using it to "deliver results."

The following is the original text:

Those who figure this out will gain an asymmetric advantage. Soon everyone will learn how. But before it all becomes obvious, here's how you can get started.

In the past year, developers have been running autonomous AI agents (such as Claude Code, OpenClaw, etc.) in the background. They can do research, build products, and deliver complete results directly without the need for repeated stares or prompts. But you have never been able to use this set - unless you can use the terminal and write code.

Perplexity Computer changes that. For the first time, non-developers can use the same capabilities. All you need is a browser and a task to give it.

Most people open Perplexity, type a question, get the answer, and then close the page. They missed the point. Perplexity Computer is not designed to answer questions, it is designed to perform tasks.

Stop asking questions and start giving it the real work.

Why most people fail

CFOs, lawyers, consultants... they open the tool, type in a question, get a decent answer, and think, "Oh, a more advanced Google." Then they spend the next 90 minutes cleaning up the same spreadsheet they just cleaned up last Monday.

The problem is not the tool, but the usage. They think of it as a chatbot.

How to ask: "What are the risks in this contract?"

Task method: "Review this contract. Check all statements one by one to see if they are supported by public sources; highlight ambiguous wording, missing clauses, and parts that may bring legal liability; list the 5 most critical risk points, with references to specific clauses; output a Word document with traces of revisions."

Same contract. One method only gives you a list to read by yourself; the other method directly gives you a finished product that can be sent to the customer.

It only takes 10 minutes to set up this system

Connect the tools first. Click connectors in the sidebar. Perplexity can connect to more than 400 apps: Gmail, Google Drive, Slack, Salesforce, Notion, SharePoint... connect everything you actually use.

Then let it know who you are. Just type it once: "I work for a certain type of company in a certain position. I regularly produce X, Y, and Z content. Please remember this context in each session." It will retain this information for a long time.

Tell it "what is good" again. Find 2–3 results that you’re most satisfied with, upload them and type: “These are some of my best examples of work. Please study their format and tone and use them as a reference when generating content in the future.”

In this way, it is not guessing your style, but reversely deconstructing your proven path to success.

10 minutes, do this first.

A real-life example: That Monday that didn’t take 90 minutes anymore

A financial analyst receives a data export every Monday, 150 lines in a confusing format: duplicate data, three date formats, ratings written in words instead of numbers. She spends 90 minutes each week cleaning the data before starting her analysis. The same question, repeated every week.

She entered only one command: clean the document, remove duplicates, unify the date format, convert text ratings into numbers; perform analysis on the cleaned data; generate an interactive dashboard with filtering function and provide a sharing link; output a PDF report comparing before and after cleaning; save all files to the "Monday Report" folder of Drive.

4 minutes later: clean data sets, interactive dashboards, sharing links, PDF reports—all in her Drive.

Then she asked again: "Are there any improvements that I haven't asked yet that would make this more useful?"

The system suggested two things: first, set this task to run automatically every Monday at 7 a.m.; second, add a new task to generate Tuesday's management briefing based on underperforming sectors.

She set both items and closed the page.

Every Monday after that, it would run automatically—whether her computer was on or off.

This is exactly the set of capabilities developers have been using for the past year. Now you can use it in your browser.

What people are already doing with it

@gregisenberg does a live test on the @startupideaspod podcast.

He was given just one task: Find the companies advertising on competing podcasts, identify the people actually responsible for sponsoring them, and write a personalized email for each one.

The system found Ramp’s VP of Growth, grabbed the content of a podcast he participated in two weeks ago, wrote a cold email, quoted his specific speech in the show, and then sent it directly. Greg didn't say "send", the system judged that the task was completed and executed it on its own.

Then it proactively suggested: Monitor competing product podcasts, and once a new brand starts advertising, immediately remind them and attach the corresponding contact person - "Contact them when the budget is first launched."

Ultimately, the process resulted in 96 prospects being surveyed in parallel, with follow-up emails scheduled for Days 3 and 7.

On Marketing Against the Grain, the team used it to audit the entire HubSpot product page: automatically crawling the entire site, scoring by custom criteria, sorting issues, and generating a shareable site report. What originally took the team a week's work was completed during the recording of the show.

These are done live and are not demos or pre-scripted.

Usage for specific tasks

In finance, a portfolio analyst gave Nvidia just one assignment before its earnings release.

The result returned: a real-time interactive dashboard with $130.5 billion in revenue, 75% gross margin, 114.2% growth, a full income statement, and projected margin trends from fiscal 2021 to 2028, all with filtering and sharing links.

No Excel, no manual data search, completed in 5 minutes.

Perplexity can directly call SEC disclosures, FactSet, S&P Global, PitchBook and other data sources - no API key or additional authorization is required, and the system is built-in.

Legal Scenario:
"Review this contract. Verify, clause by clause, that all statements are supported by public sources; highlight ambiguous wording, missing standard clauses, and content that may create legal liability under [specific state] contract law; list the 5 most critical risk points with reference to specific clauses; output a Word document with traces of revisions."

A reviewer uploaded a proposal claiming that a market had grown 43% year over year. Perplexity Computer found only 4% of the real numbers, stopping problems before contracts were signed.

Marketing scenario:
"Analyze the best-performing content of [Competitive Product 1], [Competitive Product 2], [Competitive Product 3] in the past 30 days; find the most interactive content formats and topics; identify content gaps; generate a 30-day content calendar based on these gaps and save it as a Google Doc."

Set it as a scheduled task. The latest competitive product analysis is automatically generated every Monday, eliminating the need for manual research.

Operation scenario:
"This is our CSV data for Q1. Please clean the data; analyze revenue by region and product line; identify the three biggest problems; generate one-page action recommendations; create a one-page reporting PPT; save all files to the project folder."

Five deliverables, one instruction. While you're in the meeting, it's already done.

Model Council: Get three judgments in 60 seconds

When you're faced with a decision with real consequences, enter the question once. Perplexity will call Claude, ChatGPT and Gemini at the same time, and a "synthesizer" will summarize their consensus and disagreement.

·The consistent part among the three: high confidence conclusion

·Parts where there are differences: further judgment is needed

Someone asked whether the product should be priced at $297 or $497. The three models gave different answers, but the synthesizer found that the only conclusion they all agreed on was: Don't go below $297. The decision is complete.

Many companies will pay consulting firms to lock analysts in conference rooms to draw conclusions.

Here, only one instruction is needed.

Real core competencies

Getting real value from Perplexity Computer depends 80% on one thing: your ability to clearly describe the "final output".

Not a technical configuration. It’s about whether you know enough about what you want to deliver. Don’t describe the steps, describe the results.

After each task is completed, remember to ask again: "Is there anything I haven't asked yet that can make this result more useful?"

It will point out the blind spots almost every time. Use it every time.

Start here

Open Perplexity (pro version $20/month). Enter the Computer page, click connectors, and connect Gmail and Google Drive first.

Enter three sentences of your background (only once). Upload 2–3 examples of your best work and let it learn your style. Then choose a task that you spent more than 2 hours on last week, and the output is similar every time: describe it in terms of "final deliverable" and send it. Observe the execution process. If it is a recurring task, set it to run automatically before closing the page.

The developers have been using this for a year. The output gap between them and others is real.

This is the way to narrow the gap.