How I Use Loom + Ai To Automate Everything I Do - Steph Smith

All right, welcome, everybody. I'm your host, Kit Bodnar. This is the one and only Steph Smith. Steph, I think you are one of our most reoccurring guests. I think one of my favorite humans in the world. You are so smart and you know so much about everything, and we love to have you on the show. And then we're joined by Kieran Flanagan.

Who I'm here as well.

We know he's fine, but we'll just do. We'll just do this show. Kieran, what are we talking about today, my friend?

Actually talk about practical use cases, things that we are actually doing with AI. Too much of it is abstract. We're kind of talking about the future of AI. We're talking about the big picture of AI. We're actually going to get into the day to day. What do we do day to day? How have we integrated it into our workflows? What are the core use cases? And I'm actually really curious to learn from you, Steph. You're one of the best content creators on the Internet. I know you've done a bunch of new courses, you're on Internet pipes, you've done a bunch of great content. And so let's start with you. What's a use case you can share with our audience where you find it pretty beneficial to how you work?

Yeah. So actually, in thinking about this, I tweeted this out this morning and basically it's this idea of in the last few months, we've had a few new models come out, from code to writing, imagery, et cetera. And this is one that I think anyone can implement. What is stopping you from. From going and recording your processes in Loom. Loom then generates sops for you.

Tell people what sops are Standard operating procedures.

God, this is already so good.

I'm telling you, Kieran's like, I'm walking off the show steps way better than me.

And then you feed that sop into OpenAI's new Zero1 model and you say, reason through how this can be automated or improved. And then you take the automation suggestions and you put it. You use cursor to actually automate those few things. Or, you know, you can use a zapier or where things can be done without code, but anyone can do this. And also you can tell your teams to do this, because I feel like for a long time you're like, oh, okay, how can we be more efficient? And your team's like, I don't really know how. Just implement what I'm calling async week, you go off on your own, you go those four steps, and then hopefully at the end of that week, you have, like, a very significant improvement in how every single person on your team operates.

Can we talk about what an overachiever you are? Because that was actually two things in one. It was. You kind of, I think, set it up to automate your workflow, right?

Like, where you're going to basically figure.

Out what you can automate, document that. But you said something interesting there that I want you to share with everybody, which is, like, get feedback on your work.

Because.

Because, like, what you're saying is you can also say, hey, can I. This thing that I'm doing, the way I'm doing, or could it be better? Could it be more efficient? Right?

Yeah, yeah. Because I think a lot of people actually, when I tweeted this out this morning, someone said, well, that requires that people actually have a process. So there's a. There's actually a precursor to that.

This is really good.

But then in order for you to have a process that actually yields someone being able to provide feedback. And by the way, this can be cyclical, where you now have a process that not just humans can interpret, but LLMs, right? So your whole organization is actually codified in some way.

So I love this for. We just got off stage. One of the things we talked about in terms of how you make AI much, much better is actually onboarding. And so, like, we talked about the example of Klarna Sebastian, the CEO, said the first time they tried AI for customer support, it didn't work because they didn't have great onboarding docs. And this is a really good example of how you can create onboarding docs to tasks at scale. It's not that dissimilar, actually. The reason I was like, this is a better. I'm kind of frustrated because this is a better version of the use case that I was going to talk about. And so, like, one of the things that I am obsessed by is, like, me and Kip talk a lot about craftspeople, right? Craftspeople are the people you want in your company, people who have deep domain expertise. But a craftsperson, to have scale, doesn't need to have operational skill sets. Like, they have to build operating models to be able to do a bunch of work at scale. And I think if you could actually have a chief of staff paired with a craftsperson, you would get way much more operational scale. So a chief of staff is someone who can actually scale your operating model. Kind of make sure that all of the things you need to get done across teams are being done. And so one of the things I did last night, I told the story a little bit just on a different stage, is I had the option yesterday of going to sit in a VIP box and watch Pro Jam or build an agent for this show. So I decided to choose.

What did you choose? What do you think you chose?

Did you make that decision instantly? Because I was like, one of them involves talking of people that I don't really know and one of them is sitting in my room building an agent. It's always going to be in the.

Room building the agent literally a hundred times out of 100.

And so what I did was I wanted to have like a chief of staff who could build an operator model. And so one of the things I believe I'd operate a model is they should solve for a couple of things. They should solve for speed of speed of execution, momentum and accountability. Which means you should have weekly sprints, you should have follow ups, you should have templates of what people need to fill out. You should have mostly done async and then live live discussions on most important topics. Like you have a really succinct operating model. And I basically taught the LLM through uploading all of the docs that we've done, the actual meeting docs, the templates, the schedule. I uploaded everything that we've done as part of one of our operating models to recreate that and tell me how it could get better. And it did it perfectly. It created the templates, it told me the follow ups, it told me it built me the calendar of all of the meetings to put into people's schedules. And I think you could go one step further. I think I could build an agent that when the meeting's done, I could actually send the transcript of that meeting to the agent, that the agent could actually pull out the follow ups and send them to the DRI.

And grade it.

Yeah, and grade it.

Grade the meeting for sure.

The takeaway from our audience here is I think your one is really great, which is the more you can teach it the exact thing you want done and it can actually help you grade that thing. It can get onboarded to that task and start to replicate for you at scale.

Yeah. And I think some people might say, okay, well I can't do step four, I can't code whatever. But honestly, a lot of the value comes in that first step. Because most people never actually take the time to document a process because they think, oh well, it's all in my head, there's all these nuances. And it's like, actually, no, can you systematize it? And most likely if you can't, you're not thinking really critically about the inputs and outputs.

That kind of goes with one of the use cases I was going to bring up, which is like, basically automation apps. You're like, hey, I don't. Some people are just gonna be like, hey, I can't code. I can't. I can. It's gonna give me recommendations, it's gonna tell me what to automate. But then like, what am I gonna do with it? Well, now, like, we. We just did a show on the replic coding agent, and even if you can't code, you can go to the coding agent and say, hey, here's a prompt of all the things I want to automate. Can you give me ideas for the best ways to build applications to solve this problem? And it will give you three, three to five different options. And one of the most amazing things about that agent is that gives you stuff you would never thought of. Like, sometimes I know what I want to build and I go. And then it's like, do you want to add all these other features that you didn't ask for? And it's always like, yes, that's really smart. I should have thought of those. Please do those.

Kick was taking me through something which I thought was like, really? I hadn't thought of witches. You can just give the replit agent a bunch of API docs and say, what shall I build? And then it will give you things and you can click the one you want build. And then it will say, okay, do you want this extra functionality? And I'm like, we are literally. This would literally cost thousands and thousands of dollars. And all it is now is a chip box.

It's the code version of. Do you want fries with that?

I always want the fries.

First of all, fries are delicious. Second of all, you always want them.

I know. Maybe one other note. There is, I think a lot of org you guys know this better. Led huge teams. The excuse is always like, okay, I don't have time to do that because I have to do my job. So you never have people actually take and set time away to go and learn the tools, document their processes. Not only does the technology make it faster, but I do think an organization or a boss needs to be like, okay, here's your week. Or here's your, like, every Wednesday. And maybe even some reward mechanism where it's like, you're actually rewarded for Automating parts of your job or.

Well, so that brings on another use case I was was going to bring, which kind of plays off. That is what is stopping anybody from using their calendar and whatever meeting zoom teams, whatever meeting recordings and transcripts, putting all of those into AI with their goals for the week, the month, the quarter, and looking at time allocation. Am I actually like, how am I actually spending my time and doing my work versus what I'm trying to accomplish? Because it gets to that point you're making, which is like, well, does the team really have capacity to do more or do they not? And that's actually a very hard thing to understand. But it's not really anymore because you can just look at that and say, oh, we're actually focused on all the right things and people are focusing all their time there. So yeah, we probably can't do more.

What's that you're actually talking about, which is something I've been thinking through, is in the future, what percentage of your time is doing your job and what percentage of your time is teaching an AI to do your job? Which does bring in a whole wrath of like awkward questions. But that is fundamentally what we are talking about here is that the more you can document what a human does and really be specific about that, the better the AI assistant will be. The one quick switch I want to make is, you are again, like an incredible creator. You've created a bunch of great courses, you're a great content creator. I'm curious about how you feel about AI in your day to day work as a creator. Because for a long time I was really kind of blase about AI could never help me do anything. I have the ideas. I'm a great content creator. AI is not any. It's just inferior to me at doing this. And so I cannot create it. I can't use it in any way. Claude 3.5 definitely changed the way I thought about AI completely in terms of how I use it for content creation. But how have you used it, if at all for content?

I mean, so I'll say just point blank, I'm not worried, but I'm partially not worried. Okay, so here, here's an example.

Drew and I had a really depressing AI conversation last night. So I'm glad you're not worried.

Depressing. When I say, oh, like you like AI, it's not compressing, it's just, okay, we can't.

Well, here, here's a story. I. I did a gifted program growing up and I was definitely.

That's the Least surprising thing that I've.

Heard on the show today, definitely the bottom tier of that class. Okay. What I learned in that class or being a part of that environment is like, I look around, there's absolutely like, ignore AI absolute people who are way smarter than you, like right in front of you. Right. Like, you guys are both very senior within HubSpot. I can guarantee there's smarter people below you in some fashion.

Yeah.

Somehow you've figured it out. Same thing with content creators. Like, there are definitely smarter, more creative and different variety people out there. And I've still been able to create interesting things. And so that actually brings me to a second example of something that I think people should be using AI for, which is a lot of people. Yes. Use it to create, you know, kind of stock, like images. They're moving into video. But people should be paying attention to the very interesting ways that people have created new art forms with AI. So let me give you a few examples of things that probably people have seen but not registered or clocked so that they're actually kind of new ways of creating.

Okay, real quick. On this show, we help you save time by pinpointing the AI strategies that actually deliver so you can ditch the ones that don't to help you stay focused. We just dropped the new AI Trends for Marketers report. This resource is packed with original research and findings on the latest trends in the AI space. And we've got some really surprising insights that you can only find right here. Now you're going to learn the top five AI tools across all of marketers. Who wouldn't want to know that information? We're going to tell you exactly how teams are investing in AI and where they are seeing a high return on investment. Is AI going to help or hurt your marketing career so you can steal hours of research for free? Just grab your copy from the link in the description below. And now we're going to get back to the show.

First one is the make it more trend. You guys remember that?

No, I'm sure you. I don't think we're cool enough to remember it.

No, it was so make it more anything. You have a bowl of ramen, make it spicier and then spicier and then spicier and spicier.

Oh yeah, I have seen same thing.

With like a surf picture. Like make Kelly Slater more badass. More bad. And then he's like jumping over sharks and like. Anyway, that seems again, really mundane, but this is an idea that you could not have done without AI. Let me give You a couple better examples. There was an AI generator that was an anagram, which mean a visual anagram, which meant one way it was Marilyn Monroe upside down. It was Edison. You can't do that. I mean, maybe some artists, like, you know, taking 300 hours, could have painted that. You could not do that with AI. Here's another example of something someone did before AI but could absolutely be done better and easier with AI, which is all of those. Have you seen the, like, musical Huberman clips? Or like, it's like basically a podcast clip that they have made into a song.

Yeah.

Yes, Right. And they're like tunes. They're really, really good. And tons of those have gone viral. And people. This is just like actual artists who are going and creating these. Why has no one done that with AI and made it even better? There are so many examples where, again, it's like you're actually extending the lane of creativity in some way with these tools and creating net new things that didn't exist before. And so for those reasons, I'm not really worried. I think humans will keep doing that.

I think it's a really interesting point. Like, we're here at HubSpot's inbound event, and on the main stage here, our team generated images in midjourney, animated them with Runway, and that's like the holding pattern for the slides. Like, that's the filler art. And it's not like historically you would have had, like these boring slides or you would have had a ton of money in custom animation. And you can now do that very quickly. And there are, like, lots of practical use cases for art. I also think they're just some really cool AI artists, like Claire Silver. Like, her art is incredible. Like, some of those folks out there are doing things that, you're right, would never been thought possible in other art forms. Right. They're literally impossible in other forms.

We talked about this, that it's. I think people get stuck that, hey, this is a copy and paste tool. But we think it's an accelerated learning tool and it's the ability to. I think AI allows you to unlock creativity. One of the things I'm fascinated by is the accelerated learning. Like, if you can actually distill down why someone is great at what they do or why this piece of work is great at what it is. And that's the thing I've been playing around with Cloud a lot, is if I give it enough of something, can it actually distill out why that thing is good and for everyone? I was doing that with OpenAI or the early cloud models, it was not good or like, it just. It was a thing you would expect a best practice because again, it's trained on the Internet and then it's a. It averages it all out. So it's exactly what you would expect the average to look like. The latest quad model is starting to do a really great job of trying to extract why something is good. A good example is I had a. You know, Sean, who's co host on my first million. Very good writer, has like a really great mix of wit and the ability to articulate things that people believe, but in a very differentiated way. And so I given a bunch of his content and I was like, well, why is this so good? Like, this content does really well on this platform. Why? And it could distill I really. Some things I hadn't thought about before of why that content works so well for that platform. And to me, that is how humans learn. Right. How I would have learned in the past is I would just read Sean and I would try to apply some of his stuff into my writing. But now I can just basically get it from Claude, the core things. And then I can say, well, what are the parts that I want to get better at myself and integrate into my own writing? And that to me is like the, like the creative unlock and the accelerated learning are the two parts that I am most excited by. By AI Not, I think, where people are really stuck and where they don't have much success is when they're like, oh, well, I can use it to copy and paste this thing and use it for my work.

Yeah.

And that is just again, the la. That's the lazy approach. And the people in the lazy bucket will always be the ones who get commoditized that much faster for sure.

How do you use AI Learn to learn.

I mean, I. I constantly ask it for, like, basically, as you're saying, a learning tool. So if I write something up, I'm like, grill this. Like, like, why is this terrible? And the same thing goes for. It doesn't have to be writing, it can be processes. And. And I feel like the same way that every time in something like midjourney, you're upvoting or you're deciding when I'm downloading an image, midjourney is circling that back into the algorithm to say, okay, this was what the person wanted. I think you should always have your own eyes on. In the same way where it's like, what did I like about this image? Why did I like this response from ChatGPT or Claude better than this one. Why do I have to keep prompting? Oh, most of the time it's because I'm not very clear. And so there's kind of a meta awareness that I think comes with using these tools that I don't think as many people pay attention to. They're just mostly like, it's a more active awareness of the fly.

I think what's really interesting that both of you are talking about, and we, I think maybe talked about this on pastures together is throughout human history there's been this thing of like, that person has taste, right? Oh, that person has a taste for making music, that person has a taste for art, they have a taste for fashion. And, and one of the use cases you're saying is like, hey, if you just focus on something, let's say fashion, right? And you just save pictures and annotate why they're good, after a while you can then feed all that to an AI model and actually document and create taste from that and in a new way that like before was like this very amorphous thing that it was like, oh, well, that person just has it or they don't have taste. And now it's like you can just document, anybody can have taste.

You can basically put things into style guides and then tweak. What I've been doing is creating style guides for anything and then within the style guide, starting to tweak the variables and seeing by tweaking the variables what is the output. And then like compare and contrast. And again, what's really happening is I'm just learning at an accelerated pace because I'm able to do that.

By the way, what people should do is if you just scroll midjourney, compile some of the best images every single day from their platform. And if you go to the homepage and scroll it, you'll see a bunch of images that will wow you in totally different ways, right? And if you click into them, you'll obviously see the prompt and then you can click into certain things will jump out at you as like, oh, what is this art form? Or like one of the ones. So for Internet pipes, the product I run, I have one prompt. It's about 30 words. I think there's two parts of it that generate maybe like 80% of the output. One of them is something called blind box design. I found it through this method of going to Midjourney's homepage, seeing some kind of like Asian looking, internety 3D sculpture. And I was like, what is that? And I clicked into blind box design. And I use this prompt for everything I can put in Berlin Skyline and then the 30 word prompt, because we're doing an event in Berlin, and it just spits out this perfected visual style that I had no idea existed.

Exactly. Yeah. It turns everything into building blocks.

Yes.

Like, it turns everything into building blocks and then you can kind of just start to stack, rank them on top of each other. I think that is like the best use of AI.

Well, AI takes all the unstructured aspects of our lives and makes them structured. Right. Turns them into the building blocks. And I think the hardest thing for all of us as humans is to understand that that is now possible. I find myself constantly going back and being, oh, I can't do that, or I shouldn't do that. And it's like, no, no, that is not possible. I need to go and, like, try to actually do that. And it's like breaking our way of thinking is just as important as actually using the technology, because I think we limit ourselves a ton. Like, even that midjourney example, Steph, it's like you could build a cool agent that just shows you what the cultural and taste aesthetic is like over time, Right? Yeah, you just snapshot that homepage every day and, like, index it. You can be like, oh, all right. Well, it's clear this type of imagery is going out. This type is coming in and this is why. And it's like, that's incredible. Like, you'd never be able to do that before.

Really simple. I mean, so something I used to do, it was literally a routine before AI was. There's all these tools like subreddit stats. There's a new one like Gummy Search that take platforms like Reddit, which just so many people engage in. And the fascinating part of the Internet, which is going to sound really mundane, but it's really important is you can see what millions of people, like, buy, care about. And obviously you guys know this from the SEO perspective, but that is, like, so underrated. And so you can go to something like a Gummy search and just see in the last week, month, year, all time, what was the most upvoted, AKA the most liked, resonated with thing on the platform. And you can do that for Reddit's particularly interesting because it's drilled down in specific communities. So you can do that and you're like, oh, this chart got 7,000 upvotes on R slash data is beautiful. Like, I wonder why? And you can stack it in the same way.

Exactly. And then you can actually build from that.

I'm just like. Sometimes I'm just like, technology is awesome. Before we wrap, the one thing I will say like back on your point is what that tells you is that we're all going to be able to learn so much more quickly that taste is going to change faster. Like that chart that is awesome. Will be awesome for that format. And waiting will. Won't last very long.

It'll.

Yeah, like everybody will just do that and somebody will then figure out a better way to do that.

As a whole show digital, we are never going to see the saturation curve get more exponential.

It'll be fast. But it's also like, taste is confusing.

It is.

Can go to a concert and hear objectively worse music.

That's true.

Than on Spotify, but they still love the concert. So I think like, taste is confusing. Something in the experience worth studying because there's just counterculture. Anytime something trends like AI, like there will be counterculture associated.

All right, to wrap, I would like both of you to close out with the number one use case that folks should leave today's show with and go use AI for.

I think you should document your process.

I believe Kieran's like, I'm going to steal steps with a good one.

You will see this on the LinkedIn post very soon. I actually do think that as the use case to do, I am a big proponent of documenting processes, creating onboarding docs.

Yeah, I'll just double down and say the same thing. But the frame people should take is, in my opinion, Async over sync. And there's this thing. I got a credit. I think the first place I saw it was through Jason Calacanis, but he said instead of room, you've got Zoom and then you've got Loom. And it progresses to what maybe is objectively more homey and engaging and emotional in the room, but way, way, way more productive and I guess something you can work with.

Yeah.

In Loom. So room, Zoom. Loom. I prefer Async Loom.

We all prefer Async. We're all antisocial. We talked a lot about use cases today, but the number one is how can you just. What did you call it at the start of the show? You said, like, take an Async week and kind of go and document Loom everything and. And get feedback from AI on what can be automated and how you can make it better.

Async week.

I love that. Async week. This has been another episode of Mark against the Grain. Big thank you to Steph Smith for joining today's show, and we'll see you next time.

Thank you.

Thanks, Steph.

This data is wrong. Every freaking time.

Have you heard of HubSpot? HubSpot is a CRM platform where everything is fully integrated.

Whoa.

I can see the client's whole history. Calls, support tickets, emails. And here's a task from three days ago I totally missed HubSpot Grow better.