Perplexity Computer: 6 Advanced Features That Changed Everything

In my last video, I showed you the  basics of Perplexity Computer. Today,   I'm going to break it. One prompt, five  deliverables, zero effort. Let me show you how.

Quick recap for anyone new here. Perplexity  Computer is an AI digital worker. It's not a   chatbot. It's a system that does the work for you  using over 20 AI models. If you haven't seen the   beginner video, the link is right up here and down  below in the description. Today, we're going to   go deeper. six advanced features that completely  change how I work. And every single demo I show   you, you can do it yourself. All the prompts are  down below in the description. Most AI tools give   you text. You ask a question, you get an answer,  then you have to go do that thing. Computer is   different. It just doesn't tell you what to do.   It actually does it in your real apps. So, let

me show you. Over here in the sidebar, you've got  this connectors panel. This is where a computer   plugs into your actual tools. Gmail, Google  Calendar, Slack, Notion, GitHub, Salesforce,   over 400 apps and counting. I'm going to connect  Gmail right now. All I need to do is click enable.    Go through standard Google authorization. Give it  permission. Done. Now, watch what happens when I   give it a task. We're going to go back to new task  here. And I'm just going to tell it to do this.    Search my Gmail for the most recent email from  a newsletter, a subscription service. Summarize   what it was in two sentence. Then draft a plate.   unsubscribed request. Email back to the sender.    Show me the draft before sending. Right now,  it's searching my Gmail. Now, this didn't take   very long at all. I can see it found from Lenovo  a uh newsletter email. So, it went and drafted or   crafted the message here in a very polite tone.   And then it went and asked me, would I like to   change any of this or send it to a draft? I just  said create draft. And this is what happened next.    I went over to my Gmail and this draft was right  here. All I have to do is hit send on it. I want   to point out computer will always show you the  draft and ask for confirmation before sending.    It won't fire off emails without your approval. So  the flow would be computer drafts the email, shows   you the draft or review, you approve it, and then  it sends it out through the connected Gmail. Now,   here's what makes this powerful. Computer connects  to over 400 apps. It just doesn't read your data.    It can take actions. Send the emails, create the  calendar events, post to Slack, update your CRM.    The transferable concept here. Think about what  repetitive app switching you do every single day.    That's what connectors eliminate. Okay, this next  one is a game changer for anyone who does the same

type of work over and over. Custom skills. Think  of it like this. You can program a computer's   behavior without writing a single line of code.   create a reusable skill that changes how it works   permanently. In skills panel, you can see that  there's built-in skills for slides, research, data   analysis. These come preloaded, but the real power  is creating your own. Just look at the top right   hand corner, create skill. I'm going to create it  with Perplexity. Now, the neat thing with this,   I can have a conversation with Perplexity. So, I  could give it a little bit of information and then   Perplexity will walk me through, ask me questions  to help clarifying things. But in this case,   I know specifically what I want. So I'm just going  to go and put that into the prompt. So this is   what I want. Help me create a skil called YouTube  script helper. When it's activated, it should   ask me for the video topic and target length.   Research the topic thoroughly using web search.    Create a script outline with three to five main  points. Format everything as a clean document.    Include three title options. And always write in  a conversational energetic tone suitable   for YouTube. So, I'm just going to go ahead and  submit this. And when you get really specific on   what you're asking, then it's going to have less  questions for you as well. But take advantage of   perplexity, having that conversation with you  when you're creating a skill if you're not sure   what you want. So, right now, it's building the  skill, setting up the instructions, defining the   workflow, and it's going to go and save it. All  right, the skill is all done. Take a look at the   workflow really quickly. I can see how it's going  to gather input, research the topic, build the   full script, adds estimated timestamps. Okay, so  how do we use it? So at the very bottom to use it,   just ask me to help you write a YouTube script.   So let's go back to a new task. And if I just go

and ask it, help me write a YouTube script for an  AI tools replacing Google search. Target length 10   minutes. And let's send this off. Look at this.   Right away, I'll load the YouTube script skull.    Okay, it's all done. took about seven minutes in  total. I can see where it was researching uh the   topic I gave it. I went through everything started  the document creation. I can see a check off about   building the script, the title options, generating  timestamps as it works through all of this. Now,   here's the document down below. It tells me  what's inside uh with everything. I'm going   to click on the document that it created. So, I do  have some options. I can download it or I can just   uh open it in Google Drive here. So if I just  click on it and here it is. So here are the title   options. AI is place in Google search. Here's what  you need to know. Um my thumbnail concept. Maybe   I could put this into just the image creator and  try it. We have our time stamp. So we have about   a 9minute video. Now it knows me a bit too and the  length of videos that I tend to do. So maybe that   could be why it uh picked that. And it I can see  uh a little bit of information in here. So yeah,   I should go ahead and try to make this video and  see if it uh how it comes out. Then you guys will   know if it did a good job researching. So if  I take a look, I can see where it's sourcing   everything as well as I go through it. So it's  pretty detailed uh with all of this information   with the timestamps. I think my first glance at it  uh maybe the timestamps wouldn't take that long,   but we'll see. And here's all the sources. So that  was seven minutes worth of work. Complete script,   title ideas, thumbnail idea, timestamps. Skills  are like installing apps on your phone. They   make computers specialized for your workflow.   Once created, they activate automatically when   relevant. You can also upload skills from GitHub  or community sources. And here's the really cool   part. Skills can chain together. Your research  skill feeds into your slide skill automatically.    One prompt goes from research to finished  presentation. This one is for anyone who's ever

stared at a messy spreadsheet and wanted to throw  their laptop right out the window. We're going   to upload a real messy data set and watch what  computer can do with it. How it can clean it up,   analyze, build charts, and deliver a full report.   No formulas, no pivot tables, no code. And here's   what we're working with. It's 50 rows of customer  feedback data. And it's a disaster. duplicate   entries, misspelled product names, ratings that  say four instead of the number four, dates in   three different formats, broken email addresses in  Excel. This would take me a bit of time to clean   up. Let's see what computer does. Okay, let's go  grab the file that we need. It's just this CSV   file right here, messy customer feedback. And this  is the prompt I'm going to give it. So, I've just   uploaded this messy customer feedback. This is a  customer feedback data. I'm telling it where. And   this is the actions I'm telling it to do. So,  clean the entire data set, fix the duplicates,   standardize product names, fix the date, show  me a before and after summary of what you did,   analyze the clean data, what's our average rating  by product, create a visual dashboard with charts   with charts showing each key findings, and export  the clean data as a new CSV file. Give me a PDF   report of the analysis. So, let's send this  off. Now, that took approximately four or maybe   slightly under five minutes to do. I just really  enjoy watching it go through. I can see it where   it was running the Python on it where it cleaned  the data. Uh, it keeps going through. Uh, let's   keep going down to the bottom here. Generating  the PDF here and towards the bottom. So, it says   what was fixed. Here was the before and after  issue. All the different issues and items fixed   here. The key findings are right in here. So now  I have this right here. So I'll just click on this   and open it up. Here we have a clean one which  I can download. So like I showed you before, we   let's download this as a CSV this time time. And  I'll just open it right back up in Excel. We'll   just fix the columns real quick. And there we go.   Everything has been cleaned. Right? remember down   below or down this one where we had some numbers  in here before and now those ones are fixed out   of there. Let's just hop over to the report now.   That's right here. And we get our key metrics,   our data cleaning summary, flagged invalid emails,  product performance, customer ratings here,   complaint analysis. So all of this took four  minutes and just went through that series of   tasks that I gave it. That's one thing to know  about when working with perplexity computer.    The task you're telling it what you want to get  done. Computer runs real code in a sandbox. Python   data libraries, visualization tools, everything.   It's not guessing at the analysis. It's actually   computing it. And it shows you the code it ran so  you can verify everything. All right, we're not

going to leave it here. We're going to go one step  further. I'm going to say create an interactive   dashboard that I could share on the web that could  help analyze these numbers. As this is building   the dashboard, I just want to point out that  you can get Perplexity computer on your mobile   device as well. So, this is the update. And if I  go over to computer, you can see all the different   things even right now. Cleaning and analyzing the  customer feedback is happening right now on my   phone. So, I just wanted to point that out to you.   All right, it's all done. Let's check it out. Now,   I'm just going to go to full screen and we'll  give this a test. So rather than just sharing   a PDF like I was going to before, now I can go and  share this. So this is going to be an interactive   dashboard of all the different things. So if I go  ahead and pick something else, you can see how it   changes it uh based on uh what I'm choosing here.   So if I go back to all products, everything is   there. I can go and share at this point. So I can  go anyone with the link and copy the link and send   it over to somebody. And I can always go back and  forth and tell it to adjust it. I do notice they   even put a uh dark mode and light mode on this  so you can switch it over. This is the one that

made my jaw drop the first time I tried it. We're  going to upload a business document and computer   won't just read it. It will fact check the claims,  identify the risk, and suggest improvements with   track changes like a senior adviser marked up your  document. I've got a draft business proposal here,   a partnership deal between two companies. Tech  Pulse Electronics proposing a deal to Global   Retail Corp. I want computer to review it like a  senior adviser would. Catch the bad math. Flag the   risky terms. Improve the writing. Let's go grab  that report. And this is the prompt I'm going to   give it right here. Now, I want to make a point.   If you're doing this over and over again, this   type of thing, this is something you could make a  skill as well. Right now, I'm just using it as a   task. So, I just uploaded draft partnership. what  I mentioned. I wanted to do these main things.    Fact check everything. Identify the deal terms.   Flag any vague language. Suggest improvements.    Give me a summary of your top five concerns.   Deliver it in a new Word document to track changes   and comments. So, I'm going to submit this. All  right. It's the moment of truth. This one took   a little bit longer at 9 minutes. It did all that  research and was able to put together the report.    There's a few things I want to check to see what  it found. and it caught it right here. So, uh,   this part, this number was inflated. You can see  the smartwatch market grew by 4% year-over-year   in 20125, not the 43% as stated. We have some  other numbers here that it caught. And this is   all based on different research to gathered like  it went out and found that uh this one they said   it looked like maybe uh the stat appears to  be inverted but it still found the mistake. And look at this. It also flagged this as a risky  thing. The net 90 payment here 180 180day return   at 85% credit. So it gives through all this  information exactly like a human adviser would.    um how it how they would mark up your proposal.   And we have the final draft proposal at the end   now that uh it will be corrected and that we can  go ahead and just download it and share it from   there. Okay, this is the one I've been wanting to  show you. Everything we've done so far separates

task, research, analysis, document review, all  powerful but all individual. Computer's real power   is chaining everything together. One prompt,  multiple deliverables. Watch this. But first,   computer needs to know about our company. Tech  pulse is fictional. Computer can't Google it. So,   we need to give it our data. And this is  actually a key lesson for viewers. If you   want computer to work with your business, you  upload files. That's the context. I have this   spreadsheet right here. 150 rows of sales data.   Project regions, revenue, profit, sales, reps,   real company data. Although this is all fake. I'm  just going to go and drag and drop it right into   here. So now computer has our internal data. I'll  use this for our company info and live web search   for everything about competitors. So that combo,  your private data plus real time research is what   makes this so powerful. Now let's give it the  big prompt. Now take a look at all the things I'm   going to get it to do. I wanted to research, find  our top eight competitors, a comparison table, an   executive brief, a visual dashboard kind of just  like what I showed you before, a slide deck. So,   use the sample data I uploaded here as the context  to perform the data. So, I'm going to submit this.    Look at all the different skills that are being  uploaded for this. All right, it's all done. That   took about 10 minutes. uh look they even use  claude code for some of these here. If I go   down the deliverables will be at the bottom. But  the other place if I just click up here you can   see we have the leadership deck here. We have  the executive brief. We have the competitive   uh comparison and we have the app. This is the  live dashboard that I showed you before. So if   I go ahead and click on this it will open up. We  have this dashboard that I can share with others   here. And as I hover over, you can see how  it's interactive at any of these points. So,   this was just part of the prompt. Let's close  this out. Let's go and check the uh competitive   comparison. This is an Excel spreadsheet that I  can download as uh Excel. So, the different tabs   at the bottom that we can go through and this was  all put together from that multi-chain prompt that   I gave it. And we'll just quickly go back and  let's check on one more thing that can be the   uh leadership deck. So now we have this quick  deck put together. If you need to do a quick   presentation on it, we have it right here. This  is what they mean by composition. Research flows   into analysis. Analysis flows into documents.   Documents flow into visuals. Computer figures out   the dependencies and runs things in parallel where  it can. This single task would have taken a team   of people an entire day. Computer did it while I  was talking to you. Here's where it gets really

powerful. Everything we've done so far, one-time  task. You give a prompt, you get a result.    But what if computer could do the work for you  every single day automatically? Take a look at   this prompt. Set up a daily task that runs every  morning at 8 am. Each day, I want you to search   for the latest news about AI. Find the top three  most important stories. Write a brief summary each   why and why it matters. Send me the briefing as a  notification. Keep it concise. I want to be able   to read it in under two minutes. Let's send this.   Now, look at this. It's setting up a reoccurring   task every morning 8 am. This is all going to run  in the background. I don't need to be in the app.    remember. It's going to search  the news, find the top stories, write the summary,   and it's going to send it to me. Email and  mobile, and I'm just going to go ahead and   click create. And now I have this scheduled daily  task. Now, sometimes it doesn't pop up with that   other window that just popped up. Sometimes it  just will go ahead and create the task and runs   the normal way. That time it popped open in the  task creator. Think of this as hiring a part-time   assistant who works 24/7 and never forgets news  monitoring, price tracking, competitor watching,   inbox summarizing, anything you normally have to  remember to do every day. Computer just does it.

Tip one, start with the outcomes, not the steps.   Don't tell a computer how to do it. Tell it what   you want done. It's smart enough to figure out the  steps. Tip two, use memory. Computer remembers you   across sessions. The more you use it, the better  it gets. It already knows my name, my channel,   my preferences. Tip three, run multiple computers  at once. You can have up to five different tasks   running in parallel. Start one, switch to  another thread, start another. They all run   simultaneously. Tip four, upload your own files.   The more context you give it, the better the   output. Brand guidelines, templates, past reports,  upload everything. Tip five, skill stack. Your   custom skills combined with your built-in skills  automatically. Research skills plus slide skills   means one prompt goes from research to finish  presentation. Perplexity computer isn't just a   chatbot upgrade. It's the closest thing we have to  an actual AI employee. It connects to your tools,   remembers your preferences, runs on a schedule,  and chains entire workflows together. Every demo   I showed you today, you can do it right now.   All the prompts are in the description below,   and the sample files are linked so you can  follow along exactly the same way I did. Thanks   for watching this week on Teachers Tech. I'll see  you next time with more tech tips and tutorials.