# My $10M Company is Run by Claude Cowork - Here's How.

Channel: The Next New Thing
Video: https://www.youtube.com/watch?v=gWZwP_8UEGM
Duration: 12 min
Language: English
Words: 2491
Transcript page: https://viewrankai.com/tools/youtube-transcript/gWZwP_8UEGM

---

[0:00] Everyone talks about Hermes agent, open claw, all these other cutting edge tools. Meanwhile, Claude Cork is right on your desk and it's more powerful than you realize. Let's see how. Presented by Zapier, the AI automation company. So, I'm the CEO of Grain. We've got tens of thousands of users. We are doing millions of dollars a year in revenue. We're growing fast and I use a lot of AI tools. There's a lot to do every single day and how to get a productivity edge using AI. I use Claude Code, Codex, you know, you name it. But, I honestly find myself coming back to Cork a lot. It's

[0:33] just like a very easy way to prototype automations and see what's working. And then if there's a way to sort of make it more expansive, make it more gigantic, I'll then move it into a more architected flow, you might say. But, I honestly just find it such an easy way to to try ideas and riff on them. Let's take a look at a couple of your ideas. Why don't we start with email triage? That's something everyone can

[0:55] relate to. Yeah, I mean, it's awesome. It's a simple flow, but it's something that every single day we've got to stay on top of our email. You're constantly getting emails from customers, from uh you know, vendors, etc. And so, as much as in my mind, like as much as I'm wasting time, you know, on non-high signal email, that's stuff I could be doing that's higher value. So, wanted to kind of think through how could AI help me with this? I'm a I've been a big

[1:21] superhuman user for a really long time. I love superhuman, but it still is kind of in the very like gooey-centric [snorts] software design. And so, I I built this daily email triage flow to try to push further along that. The core idea of this flow is and I'll show it I'll show you kind of one of the outcomes in a second is that it uses not just the email itself, but also the meeting context. It uses Grain transcript data to understand who these people are, what previous commitments have you had, what conversations have you had, what open threads. And it uses that context to not only better triage and prioritize the

[1:57] emails, but also draft smarter responses. And I And I've been As I've been kind of loading that meeting context in addition to the email context, it's just much, much better outcomes. Okay. So, it's going through. It's saying, "These are the emails you need to pay attention to, and I'm actually writing follow-ups for you. And the follow-ups I'm writing are based on what I've read in the email and what you have said in past conversations." Let's take a look and see what one of these reports

[2:20] looks like. Yeah, so what it's doing is it's giving me a DM with flagging urgent emails that require my attention. And over time I've tuned the instructions to be quite smart about what truly is urgent and needs my attention. And then it buckets also other emails that are important, but not urgent. It also will uh flag things that are, you know, possible product signals that I could be looking at. These are oftentimes things that like Intercom chats are getting forwarded to my inbox. And then it also just kind of gives me spam unfiltered suggestion. And actually, I've started having it just automat- As it's gotten better at detecting spam, just

[2:57] automatically marking them as done. So, I don't even have to ever see them. And then it just suggests like things I should maybe unsubscribe from cuz they're just cluttering my inbox. So, it just kind of gives me like a overview of the urgent. And you can see kind of summarize each of these and then just flags other things. If you open this Slack canvas, there's a breakdown of all these things in more detail. And what I love One of the things that I love really I love about this is that it drafts responses if a draft is needed for me. And I can just open those

[3:25] directly in Superhuman. And so, for example, if I switch to Superhuman, I'll just show you an example of kind of all these drafts that are already drafted for me so that there's something just immediately me just take a look at the at the top for a moment there. Yep. Yep. So, this is Canvas. And the cool thing that you've done is you found a way to put full HTML interactive pages inside of a chat because of these uh canvases. I thought that was really big takeaway for me. I never use canvas through any kind of automation. So, this is helpful. And

[3:56] then you can see the very top ones. Those are the four things that need the most. All right. What happens if you click open in Superhuman on the message for me for today? Yeah, so this one is because we'd reschedule this podcast and I had needed to accept your invite. So, if I click this, it just opened to the calendar invite and uh reminded me to accept it

[4:15] so I knew I was coming. Ah, okay. And then can I see how it creates a draft for you? Yeah, so you can see for example, it drafted two emails for me uh previously. One is a question I got from Zapier wanting to about a case study cuz we've been building a lot of stuff with uh

[4:30] with Zapier. And I've also had some conversations with them. So, it was able to pull some of that grain context. Uh and then it so sort of just drafted this reply to Rob from Zapier. Yeah. Hey Rob, thanks for the clarification and for flagging this. We've been actively building with Zapier SDK recently for some internal automation. So, this is likely us generating that volume. I'll check with

[4:51] our engineering team. Got it. Oh, wow. So, it knows what's go This is it writing it and it knows that you did all that because it's listening it's using grain transcripts and sees that you've all been doing that in meetings talking about this. Wow. Yeah, it's pretty it's pretty awesome because like there is uh there's no way that from this thread it knows all this context. It just It knows this because of the context that this automation I've instructed it to pull from its grain the

[5:18] grain conversations. Okay. Um well, from my team cuz my team has been talking about this and from conversations I've had with like the folks doing the Zapier SDK. Um by the way, can we give people this recipe so that they could share it and use it themselves? Yeah, I'd be happy to do that. All right, we'll have it in the description. Um speaking of Zapier, they're a sponsor. What are you doing with the Zapier SDK? Is there one thing

[5:40] that you can bring up? Yeah, I've been using their SDK to build a lot of automations as a way to not have to you know, I'm I'm using Claude code plus their SDK to build automations. I find it a really nice way to not have to deal with a lot of the key management because I've already got all these Zapier offs already in my Zapier account. They just recently launched triggers, which I'm really really excited about so that I don't have to worry about setting up webhooks either, so um just a lot of automations like for example, we sent we have like this uh sales brief where it kind of pulls all

[6:13] the context of every single deal, puts it into Slack, tags people who need to follow up on their deals. We do some Intercom automations with it. There's just there's a lot of stuff we're kind of a lot of internal ops and rev ops stuff we're automating. I love it. I love it because um like you said, you get to connect all your tools in. You have one key essentially that you give to whatever app you want to work with and you know it can only open up the things and do the things that you want it to do. Okay, let's see what else

[6:37] you're doing with uh Co-work. While most of my attention has been on offense and and trying to understand, you know, how do we create new things and and and take bold steps forward as a product, I have created some automations to help me I call it a defensive product radar to help me make sure that I'm still prioritizing and coming up with solutions to the things that are active paper cuts of our existing user base cuz those are also critical to ensure you're

[6:59] building a delightful product. Okay, can we see that? I'll show you the output again in Slack. Okay, everything for you goes into Slack, huh? Or a large part of it? That's that's one of my main ways of doing these Co-work automations. As I said, it's it's a really nice way to to just ensure that these are getting inserted into my daily workflow. I'm really enjoying Co-work cuz it's just like a really quick way to set up these things and then and see what works, see

[7:22] what doesn't work and then iterate it. Okay. Okay, here you go. So, this is the defensive product triage. It flags the top defensive issues that based on Intercom chats, sales calls, customer success calls, and Amplitude signal things that I should be focusing on. And then it uh also looks at linear and at the existing issues and kind of suggests issues that perhaps are under prioritized based on the signal it's getting from the market and from our

[7:54] customers. Mhm. Uh suggests new issues to create and then and then kind of puts that all into again a canvas report with the links to the tickets that are either there or that it created that I need to triage and prioritize. Can I see that previous screen one more time? Yeah. So I love how it says, "Hey, look, here top issues." It apparently somebody thinks that they had their meeting notes shared even though it was turned off. So, that's a big priority for you and you're taking

[8:22] Absolutely. Critical for trust. that. The next thing below that, so that's top defensive issues. Then underneath it, I really like how a linear issues are being flagged based on conversations that you and your team are having. So, if a lot of people keep bringing up the same issue and it's in linear, now this automation says, "Hey, pay attention to it. I know it's just another item in your linear, but this is one that a lot of people are bringing up for you." Got

[8:46] it. 100%. Okay. And then when you open it and then it creates linear tasks for anything it thinks needs a task and only when you move it into like let's start taking action does the team start to take action? Yeah, we have we use linear triage the linear triage bucket. Mhm. And so as things as things are in triage, I then review them and then I put them into the relevant project and assign them to people. So, I still kind of maintain that final decision-making on what our team is spending time on cuz

[9:19] that's critical. Okay. Can we just see Coda work one more time? Just kind of click into it and let's see if we can see the recipe within Coda work. But of course, we'll have it below for anyone who wants to use it. Yeah, of course. So, right here product defense radar, you can see uh you can see here's the instructions

[9:37] right here. Uh-huh. Uh, okay. Review recent customer product and support analytics and linear signals to identify defensive uh product issues Grain should not miss. Create high confidence linear issues in ENG triage. Got it. Here's the goal. Help Jeff answer what product friction, customer pain, turn risk, bugs, or recurring complaints should we notice and potentially fix. Um, and then it's using call notes that Grain is taking and what

[10:07] else? It's using Intercom chats, Grain calls, Amplitude, and then existing Linear tickets. And it's doing it every day at 8:00 a.m. How do you put this together? You chatting with it and then every time you have a problem, you chat with it and tell it to improve? Correct. Yes, exactly. Yep. The one of the things that I find very easy to iterate on is in every time a task runs, it it runs in a chat and it has all that context from the run, so you can just in Claude say, "Hey, can you change these instructions? Can you improve this?" And it'll just automatically improve. So,

[10:39] it's just so easy to give feedback. I think of it as like almost like an employee that it's just like really easy like, "Hey, you did this for me. Can you please do it this way differently every time?" So, the iteration cycles seem to be just really lightweight. Also is you can because it has all that context, if you needed to actually do some work, like for example, just when I was working on one of these, I can

[11:01] actually I let me actually show this. So, you can see when it on its last run, I there was an issue that I'd remembered that a customer had had uh had asked for help with and had given some feedback on and it was related to a different issue that was flagged by this run, so I I sent I I basically said like, "What issue is associated with this this customer's issue?" Um, and then can you find any Slack uh Slack message that I'd like to understand the current status of of diagnosing the issue cuz we weren't really sure what was causing it. So, with that context, it it basically

[11:32] found everything related to the problem. If flag one part of the issue, it reminded me another user's because it had that full context, it was able to connect these two separate issues and help create a plan to get that with the full the full problem resolved. All right. This is automation using Claude co-work easy mode for people who want to take it to the next level and say I want a full-blown AI agent, we've got an Hermes agent walk-through and we'll have a link for it right there. If you want to take

[11:59] that to the next step.

---

About this transcript
Read from YouTube's own caption track and laid out by ViewRank AI (https://viewrankai.com).
ViewRank AI finds the videos already beating a creator's own average on Instagram, TikTok and YouTube Shorts, transcribes them from the audio itself in more than 60 languages, and turns what worked into new ideas and scripts.
Free transcript tools, no account needed: https://viewrankai.com/tools
How to read any video this way: https://viewrankai.com/llms.txt