Transcription
So, Anthropic, the inventor of Claude, just released internal videos showing how their own employees actually use Claude. Their marketing department, their legal department, they've all streamlined their entire process using Clawed Code. But the thing that surprised me is these people aren't coding wizards. These are just normal people with normal jobs that build systems that replaced entire departments in one of the fastest growing companies on the planet.
Now, I went through everything Anthropic has published, blog post, internal videos, research studies, and I found a pattern. And this is the same pattern I'm now seeing with the business owners I work with. So I'm going to break down exactly how people inside Claude are using Claude, what they build, the pattern behind it, and how you can apply it whether you're technical or not.
So getting to the first case study, this is all about marketing. For nearly a year during Anthropic's fastest growth period, their entire marketing operation was one guy. Austin Laauo, a non-technical growth lead, was running paid search, paid social, email, and SEO completely solo. And at most companies, that's probably a team of 10 to 50 people.
Now, here's what's wild about Austin, which besides having an absolute fire name before Anthropic, he worked at Dropbox, Web Flow, and Notion. Serious companies, but he had never written a line of code in his life. So, when Claude Code launched, he actually had to Google on his computer how to open a terminal. And his quote, which I loved, "My first reaction when we launched Claude Code was, I have zero idea what this product is for. It just didn't [music] really click in my mind."
Then somebody else on his team sent a guide in Slack about how you can set up claude code for non-technical employees and Austin followed it. And his first project he made a calculator app just to see how it worked. And then a week later he built two workflows that fundamentally changed how he operates day-to-day. And before I walk through these two workflows as well as the rest of this video, just have the mental frame that these are things that you can apply to your day-to-day because everything that they're doing you can do as well.
First workflow is a Figma plugin for ad variations. In about an hour, he was able to build a Figma plugin that generated ad creative variations with a single click. You can see it on screen, the plugin in action. And here's his direct quote: "where all I would have to do is copy and paste just once all the different variations and copy that I wanted to update." And with a click of a button, the Figma plugin will create all the different permutations for that single image. And what used to take 30 minutes now takes 30 seconds.
The second workflow is a Google Ads copy generation. He created a custom Claude skill/ RSA for responsive search ads. And I have a full video on my channel breaking down Claude skills and how you can use them. But anyway, when he types / RSA into Claude, it kicks off a whole workflow to go through this ad variation. Claude asks for campaign data, existing copy, any keywords, and then cross references against skills he's already created, which would be Anthropics, brand tone and voice, product accuracy guidelines, and Google Ads best practice. It then generates 15 headlines and four descriptions per ad, all within the character limits that Google requires. The skill that he created exports a CSV that he can upload directly to Google Ads. And the statistics for the improved output that he was able to do from all of this is pretty crazy. You can see on screen the change of time for specific things. And one thing to hammer home is that this isn't AI helping a giant marketing team. This is one person doing what used to take maybe 50 people to do.
And the reason this is so effective is Austin didn't just ask Claude to write ads. He built a system. He said something in an interview that stuck with me: "So the gap between I wish this existed and wow, I can actually build this myself is actually much smaller than people realize. [music] You don't need to know how to code. All you need to know is to how to explain your challenge and what you're trying to solve for in a very clear and concise manner. If you're able to provide a clear explanation of the problem, Claude can then figure out the solution. But the reality is it's not magic. It's iteration. It's troubleshooting. It's using your own domain knowledge to know what is a problem and what is actually the correct solution. That's where you provide your value and claude becomes a tool that multiplies your expertise."
Now I've personally have seen this exact pattern in one of the businesses that I work with. So I work with this engineering firm and we need to generate over 80 reports for something called local law 97 and this took weeks previously. But instead of doing that, we spent 2 hours going back and forth creating a single skill that probably saved us about 40 hours of work. And the reason we were able to do this is because I was working with an engineer who knew what a local law 97 report should look like. But without that expertise, we wouldn't have actually been able to create this skill. So we used his industry expertise as well as Claude skills to create a skill that can compound over time. So now as his business grows, this is like year two of him growing his business and it's scaling exponentially. He doesn't need to hire a new employee to complete these reports. Instead, he now has a skill that can do things at 100 times the speed. So human capital is no longer a bottleneck for this specific activity.
Going back to Austin at Claude, not to be confused Austin me, he puts it this way, "Growth marketing is going the way of almost like a product manager. We're able to actually build products in order to help us achieve our targets." So, he's not just using tools, he's actually building them for himself. That's a massive shift.
Now, before we get to the next case study, I know you might be thinking, "This is just Claude's team hyping up their product." And yes, there's probably a sense of truth there. However, I have personally noticed this firsthand, so they're not just like blowing smoke.
But let's go to case study 2, which is all about legal. Generally speaking, at most companies, before anything goes live publicly, the legal team has to review it. So that's landing page, ad copy, blog post, push notifications, emails, whatever. Anything that could get the company in trouble if the wording is wrong. And at anthropic, the night before a product launch, marketing would often say, "Hey, legal, can you please review this today? We need to go live tomorrow." Legal would then have to manually do whatever work was needed in a time crunch and get them back ASAP. This is obviously hard for legal and also hard for the marketing team because the reality is legal often takes longer than you would like.
So, insert Mark Pike, Anthropic's associate general counsel, who decided to fix this. Quick background about Pike. Before Anthropic, he was at Facebook and Slack. The guy knows legal at scale, but he doesn't know how to code. And his exact words, "I wouldn't say what I'm doing is learning to code. I partner with Claw to tackle certain projects that involve [clears throat] coding, but I am not the one coding. I'm just very good at troubleshooting." A lawyer, zero coding experience, about to automate the entire pre-launch legal process.
One thing that Pike said that's really important for you to understand is how he frames the whole thing. He says, "Don't start with what can AI do." He says, "Start with what do we wish we didn't have to do?" And he also says, "Be specific about the pain." Before Claude, I had a ton of tactical busy work, things I would put off until the end of the day because I just knew it would take a lot of time, but not using the best parts of my brain. So, this was a problem that he was having. So, he was like, "How can I tackle this?" It wasn't how do I use AI? Instead, it was what work drains me that doesn't require my actual expertise.
So, he decided to build four different tools. Tool one is a marketing review tool. He created a self-service tool pinned directly in Slack. Step one, marketers paste their content into the tool. Step two, the AI reads the entire thing and checks it against Anthropic's actual legal guidelines, which is stored as a skill. From Pike, he says it flags things like, "Hey, do you have the rights to this logo? Is this statistic accurate? This trademark needs a symbol." And if a landing page says Claude is the most secure AI on the market, the tool will flag it as an overstated claim because that stuff will get you in legal trouble and you have to avoid that. Based on the assessment, it'll assign a low, medium, or high risk. And then it'll tell you how to fix it. The marketer reads the flagged issues, makes the fixes, and then cleans it up before the lawyer even looks at it. That's tool one, and the other three tools are just as effective. Tool two is about contract redlinining. Tool three is outside business activity review and tool four is privacy impact assessments. And he also mentions he uses six clawed skills to power all of these. The first is marketing review skill, second privacy, third employee law skill, four commercial skill, five corporate skill, and six product legal briefing skill. All these use similar problem solving techniques to help streamline both his workflow as well as his team members workflow. He didn't just make his life easier, he made his team members' lives easier. help them help you save you time so you can focus on what matters.
And one thing to call out with all of this is if you've been watching this channel, you know that I've mentioned this concept of risk of error. This is really important to consider when you're determining how to use AI. And for something like legal, this has a very high risk of error. Simply put, if they mess something up, that's not a good thing. So, it's a high risk of error. So, to mitigate this, Pike still personally reviews everything before it goes live. This is human in the loop. and a quote from him: "I I'm always trying to ensure that a human is remains in the loop from the legal team. We know that AI systems can still hallucinate. I'm still making sure I'm reviewing the work. Uh but this is really helping us move with more speed and sort of [music] preemptively flagging things." So 80% of the work is caught through his tool and that last 20% is what he handles. His broader vision is the goal is to turn the legal team from the department of no into crossf functional thought partners that other parts of the company can collaborate with throughout the entire process. From his words, "We're not replacing lawyers, we're pushing out the frontier of what's possible."
I've actually seen this strategy work firsthand in a couple of ways from legal perspective. At my last startup, IK, we worked in the music industry. And so we would joke that we had in-house counsel, but really it was just AI with all the context about what we wanted to avoid as a business. So instead [snorts] of sending everything to our legal counsel and racking up our legal bill, we had our in-house legal counsel, aka AI, review these documents for us and flag anything that would be a concern. This way, we're not having them spend time looking at just boilerplate NDAs, but instead focusing on the points that matter. That's a whole other story. But clearly this can be functional and Claude is using it internally. You can use it externally between the marketer and the legal department within Claude.
There's a clear pattern that I see. You're not replacing the expert. You're building internal capacity that handles the repeatable stuff so that the expert can focus on what actually requires their judgment. And the one concept that Pike used that you should also consider is this micro tool concept. He built a tin legal review tool that outputs a Slack message. This is a micro tool that solves a very specific problem that he encounters throughout his specific day. Think about your day-to-day and I bet you there's probably 10 different micro tools that are small, focused, and solves one workflow issue.
So, zooming out, what's the broader pattern? Austin Laauo didn't just ask Claude to write ads. He built brand voice skills first. Wrote down what good ad copy looks like, what anthropics tone should be, then he built the automation. Pike did the same thing. Before building his legal review tool, he codified six different skills. Marketing law, privacy, employment, etc. That's his judgment. And as a lawyer, that is so important. From a marketing and a legal perspective, they codified and they documented their expertise before building anything. The reality is AI won't replace your judgment. It can scale it, but you do have to write it down first. Codifying your expertise first, then build systems on top of it.
So, think about your own work. What's the thing you do every week that drains you, but doesn't actually require your best thinking? It's a great starting point. There's a concept called human in the loop where a human steps in to help with AI. But I think it's actually reverse. It's not human in the loop, it's AI in the loop. You use AI where you think it's needed. And the better you're able to do that, the more you're able to multiply your output.
So here's your action plan. Pick one workflow you do every week, something that drains you. Write down the steps. What's the 80% that's routine? What's the 20% that actually requires your judgment? And just like they do inside of Claude, build a system for the 80% and stay in the loop for the last 20%.
And now, if you want a deeper breakdown on how to actually build these skills and how I work with Claude, I have a video walking through my exact setup along with a template you can copy and get started with today. So, go check out that video by clicking right here. I'll see you over there. Pace.