Original title: how non-developers automate work like engineers (without writing code)…
Original author: Damian Player
Compiled by: Peggy, BlockBeats

Editor’s note: While most people still see AI as a “more efficient search tool,” Perplexity is starting to execute real work.

This article focuses on a repeatedly overlooked difference—why, when people use the same AI, some only get an answer, while others directly get deliverables they can actually ship. The key isn’t model capability; it’s how you use it: do you treat it like a chat window, or like an execution system that can be directed, scheduled, and run?

A new class of tools represented by Perplexity Computer replaces “asking questions” with “tasks” as the core interaction model. From contract review and competitive analysis to data cleaning and report generation, users no longer describe the problem—they directly define the final deliverable. Combined with connecting enterprise tools and locking in personal background and style examples, this capability evolves from one-off outputs into reusable, automatically runnable workflows.

More importantly, the boundary of automation is being redefined. It’s no longer just for helping you complete a single step—it can run continuously, execute across tools, and even proactively propose additional tasks. That means the relationship between humans and tools is shifting from “using” to “managing and delegating.”

In this shift, the real dividing line is no longer whether you’re using AI—it’s whether you’ve started using it to “deliver results.”

Below is the original article:

Those who figure this out will gain an asymmetric advantage. Soon, everyone will learn how to do it. But before all of this becomes obvious, here’s a way you can start early.

Over the past year, developers have already been running autonomous AI agents in the background (like Claude Code, OpenClaw, etc.). They can research on their own, build products, and deliver complete outcomes without people constantly watching or repeatedly prompting back and forth. But you’ve probably never needed this—unless you can use a terminal and write code.

And Perplexity Computer changes that. This is the first time non-developers can use the same kind of capability. All you need is a browser—and a task you can hand it to complete.

Most people open Perplexity, type a question, get an answer, and then close the page. They miss the key point. Perplexity Computer isn’t for answering questions—it’s for executing tasks.

Stop asking questions. Start giving it the real work.

Why most people fail

Chief financial officers, attorneys, consultants… They open the tool, type in a question, get a pretty decent answer, then think: “Oh, a more advanced Google.” Then they spend another 90 minutes cleaning the spreadsheet they cleaned last Monday.

The problem isn’t the tool—it’s the way they use it. They treat it like a chatbot.

Question mode: “What are the risks in this contract?”

Task mode: “Review this contract. For every statement, verify whether there’s a published source to support it; flag vague wording, missing clauses, and sections that could lead to legal liability; list the five most critical risk points, with specific clause citations; output a Word document that includes tracked revision marks.”

Same contract. One approach gives you a list and makes you read it yourself; the other gives you a finished product you can send to your client.

Set up the system in just 10 minutes

First, connect the tools. Click connectors in the sidebar. Perplexity can connect to 400+ applications: Gmail, Google Drive, Slack, Salesforce, Notion, SharePoint… Connect everything you actually use.

Then tell it who you are. Input it once: “I’m in a certain role at a certain type of company. I regularly produce content like X, Y, and Z. Please remember this background in every session.” It will keep these details long-term.

Next, tell it what “good” looks like. Find 2–3 deliverables you’re most satisfied with, upload them, and enter: “These are my best work samples. Learn their format and tone; use them as reference when generating content going forward.”

In that case, it won’t guess your style—it will reverse-engineer the successful path you’ve already validated.

10 minutes—do this first.

A real example: that Monday that no longer costs 90 minutes

A financial analyst receives a weekly data export every Monday: 150 rows with a messy format—duplicate data, three different date formats, and ratings written as words instead of numbers. Before she can analyze, she spends 90 minutes cleaning the data every week. Same question, repeated every week.

She entered a single instruction: clean the file, remove duplicates, standardize the date formats, convert text-based ratings into numbers; analyze on the cleaned data; generate an interactive dashboard with filtering and provide a share link; output a PDF report comparing before and after cleaning; save all files to the “Monday Reports” folder in Drive.

After 4 minutes: a clean dataset, an interactive dashboard, a share link, and a PDF report—everything appears in her Drive.

Then she asked one more question: “Are there improvements I haven’t asked about, but that could make this work even more useful?”

The system suggested two things: first, set this task to run automatically every Monday at 7:00 AM; second, add a task to generate Tuesday’s management briefing based on the segments that perform poorly.

She set both, and then closed the page.

From then on, every Monday it runs automatically—whether her computer is on or not.

This is exactly the same capability developers have been using over the past year. Now you can use it right in your browser.

What people are using it for

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

He gave it only one task: find the companies running ad campaigns on competitor podcasts, identify the actual sponsors responsible, and write a personalized email to each person.

The system found Ramp’s VP of Growth, pulled podcast content he participated in two weeks ago, wrote a cold email citing his specific remarks on the show, and sent it directly. Greg didn’t say “send”—the system judged the task was done and executed it on its own.

Then it proactively suggested: monitor competitor podcasts—once a new brand starts running ads, immediately notify and include the corresponding contact—“reach out as soon as the budget kicks in.”

In the end, this workflow completed research in parallel for 96 potential leads and scheduled follow-up emails for day 3 and day 7.

On the Marketing Against the Grain show, the team used it to audit the entire HubSpot product pages: automatically crawl the whole site, score based on custom criteria, rank issues, and generate a shareable website report. What would normally take the team a week was completed while recording the episode.

All of these were done live—not a demo, not a pre-script.

How to use it for specific work

In the finance domain: a portfolio analyst issued just one task before Nvidia’s earnings release.

The result was: a real-time interactive dashboard containing $13.05 billion in revenue, a 75% gross margin, a 114.2% growth rate, a complete profit and loss statement, and a profit margin trend forecast from fiscal year 2021 to 2028—fully supporting filtering and share links.

No Excel, no manually hunting for data—done in 5 minutes.

Perplexity can directly pull from data sources like SEC filings, FactSet, S&P Global, PitchBook, etc.—no API key and no additional authorization required; everything is built into the system.

Legal scenario:
“Review this contract. For every statement, verify whether it’s supported by a publicly available source; flag vague wording, missing standard clauses, and any sections that could lead to legal liability under [specific state] contract law; list the five most critical risk points and include specific clause citations; output a Word document with tracked revision marks.”

A reviewer had uploaded a proposal claiming that a certain market’s year-over-year growth was 43%. Perplexity Computer found the real data was only 4%, stopping the issue before signing.

Marketing scenario:
“Analyze [competitor 1], [competitor 2], and [competitor 3]—identify the best-performing content from the past 30 days; find the content formats and themes with the highest engagement; identify content gaps; based on these gaps, generate a 30-day content calendar and save it as a Google Doc.”

Set it as a scheduled task. Every Monday, it automatically generates the latest competitor analysis—no manual research needed.

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

Five deliverables, one instruction. By the time you’re in the meeting, it’s already done.

Model Council: three judgments in 60 seconds

When you face a decision with real consequences, just input one question. Perplexity will call Claude, ChatGPT, and Gemini at the same time, and a “synthesizer” will summarize their consensus and differences.

· Areas where all three agree: high-confidence conclusions

· Areas with disagreement: need further judgment

Someone asked whether product pricing should be $297 or $497. The three models produced different answers, but the synthesizer found their only shared conclusion: don’t go below $297. The decision is done there.

Many companies spend money hiring consulting firms to lock analysts in a conference room and come to conclusions.

Here, you just need one instruction.

The real core capability

To get real value from Perplexity Computer, 80% depends on one thing: whether you can clearly describe the “final output.”

Not the technical setup—whether you’re clear enough about what you’re delivering. Don’t describe steps; describe the result.

After every task is completed, remember to ask again: “Is there something I haven’t asked about that could make this result more useful?”

It almost always points out blind spots. Every time. Every single time.

Start from here

Open Perplexity (Pro plan $20/month). Go to the Computer page, click connectors, and first connect Gmail and Google Drive.

Input your three-sentence background introduction (just once). Upload 2–3 of your best work samples so it can learn your style. Then choose a task you did last week that took you more than 2 hours and has similar outputs every time: describe it using “final deliverable” wording, send it. Observe the execution process. If it’s a recurring task, set it to run automatically before closing the page.

Developers have been using this setup for a year. The output gap between them and everyone else is real.

That’s how to close the gap.

[Original link]

Click to learn more—BlockBeats is hiring for open positions

Welcome to join the official BlockBeats community:

Telegram subscription group: https://t.me/theblockbeats

Telegram group chat: https://t.me/BlockBeats_App

Twitter official account: https://twitter.com/BlockBeatsAsia