The Real AI Disruption: Why Your Platform is More at Risk Than Your Developers - Episode 136 of Commerce Today
===

Joshua: [00:00:00] Hi everybody. Welcome to this week's episode of Commerce Today. There's actually been a really fascinating conversation happening on LinkedIn right now around the e-commerce space that I wanted to dive into. Brent Peterson, as usual, has stirred up some interesting conversation and debate asking if Shopify is losing its edge and.

The usual suspects came up. People of course, were complaining about the percentage that Shopify likes to take of your revenues and just all the different extensions you have to add, just all the usual stuff. But then Nicholas Ner, one of the co-founders of Medusa through an interesting idea in there.

He said actually that there's a much bigger disruption happening right now that most are missing, and that is AI coding agents and how that's impacting platform selection. So that's my thesis for this episode is that it's not actually developers jobs that are being dis disrupted, although they are.

[00:01:00] But the bigger disruption that's coming is actually in the platforms. And I've actually seen this directly at Creatuity We recently needed a piece of software, not an e-commerce platform, but a piece of software or a SaaS to help with some sales automation work we were doing and looked around and.

I'd probably have to buy a couple of different things and spend 500,000 bucks a month. Once I put all these different things together, then I realized, wait a second, I bet Claude Code could actually just build me exactly what I needed and could pay for, 10 bucks a month worth a worth of hosting.

And next thing you know, thanks to these open source projects, I have a complete working solution that's meeting our needs and is not nearly as expensive. Really for the past year, years, we've heard. AI is coming for your job, especially for your developer. AI is definitely coming for your job.

That's the story. At least we've been told that GitHub co-pilot and tools like that, and then winder and all these things, we're gonna just completely disrupt the development jobs that these new a genic [00:02:00] coding models, things like cloud code are gonna automate coding completely, make developers obsolete.

It still might happen. There's some interesting stats I've been hearing about how in the next two years these coding agents will be about 16 times better than they are right now. And if you think about where they are right now, think about 16 times better. I guess if you haven't tried one, let just say 16 times better means.

They're better than any human developer out there, and that's gonna be pretty intense and pretty interesting. But I also think that they do require a developer's help and a developer's context and really. Providing the business context and providing the technical context even, and knowing how to operate those agents is so important.

Kind of a tangent, so I'm not gonna, that'd be a whole separate episode, and I've talked about that some. But I think the new thesis I'm developing that Nicholas inspired is that the real vulnerability here isn't people, isn't specific jobs, but it's actually the platforms [00:03:00] that they're working on.

So new tools like Cloud code. All these tools were trained on publicly available code. Do you know where a lot of publicly available code comes from? Open source projects. So cloud code thrives in open source and it's a command line based system. So anything that you can do on your computer on the command line cloud code can do.

That means any sort of platform work that can be done on the command line, Claude Code can do. And if you're not technical yet, you may not be connecting these dots. But I will say that basically that means if you're working on an open source platform that uses a lot of command line utilities, think Medusa, think Magento, things like that.

These. Are so much easier for agents like Cloud Code to do the implementation for you. These tools, however, struggle when they're working on a closed source platform when they're working, especially on things that are primarily behind like a web-based admin [00:04:00] panel. I'm sure the agents and the LLMs will get there, but they're not there yet, so they don't do as well.

So again, think SaaS platforms like Shopify, that's actually a much harder environment for these tools to work in. So the platform that you're choosing today could actually become a huge competitive disadvantage in 18 to 24 months, especially think. 18 to 24 months. These age agentic coding tools will be 16 times smarter and better than they are right now.

And that's actually, if you want to check my math on that, reach out to me. I will tell you it's based on Moore's Law. And it definitely fits with just the growth in benchmark scores we've seen in coding LLMs over the past few years. So yeah, think about it. If your competitors are able to build and iterate custom experiences at 10 to 16 times the speed of what you're able to do in a fraction of the cost, how's that gonna feel?

What's that gonna do for your business? All right. I dove right into the meat of the episode there. But let's start on the AI's [00:05:00] natural habitat. So the command line interface versus the graphical interface or the web. So what is Claude Code? Very simply, Claude Code is an AI assistant that lives on your command line.

You can talk to it, it reads, understands, and writes code directly to your project files. It allows developers to basically have an on-call senior developer as their pair programmer 24 7. And it's interesting Claude Codes pricing, and this is another little tangent, but I do have to, I feel like I have to share this.

In the race to get more users. They are anthropic, the authors and creators of cloud code. They're subsidizing things. Cloud code, you can get access for 20 bucks a month. And then there's larger plans at a hundred bucks a month and 200 bucks a month. And there are tools out there that'll help you see that.

If you don't subscribe, you can actually just pay per token, pay per use API pricing. And there's tools out there that'll help you see, okay, if I was paying for this via API pricing, what would it cost me? And [00:06:00] I've had days where I've used over a thousand dollars worth of Claude Code usage on my $200 a month plan.

So you can tell that. There's some major venture capital subsidization of the prices happening here. I don't think that's gonna last forever. I think something's gonna have to give, but I think until there's one clear winner in the agentic coating space, we're gonna see that. Level of subsidy bringing the prices down, making all of this not only possible, but within the reach of really any developer out there.

You can start for 20 bucks a month, but so that's what Claude Coat is. It's this system that literally you can sit down and actually for those that want a more technical deep dive, I'm gonna record a little screen share that I'll post after this episode goes live. Where I'm actually gonna show you how you could use Claude Code to spin up, say a Medusa website.

But you just sit down in your command line interface and you tell Claude code what you need it to do. It reviews your current code base, it edits [00:07:00] it, it adds new code, it improves existing code. It actually can, as I mentioned on an episode last month, it can deploy the site for you. I had Claude code.

Use the Amazon AWS command line utilities to spin up an entire hosting infrastructure for me, deploy a project, everything. So that is what cloud code is. Next up is the composable and open source workflow when using a tool like Cloud Code. So it's very much a local first experience. So developer clones a project onto their computer works.

Files. Again, it's very text-based, text in text out. It's able to view the entire code, base, understand dependencies and execute changes. Most importantly, it has access to search the web. So if it is struggling with how exactly do I set up this particular feature, it can go out there, search the web, find the documentation for that open source platform.

Figure out exactly how to write the code that [00:08:00] it needs to write. So you very much can go out there and say, Claude Code, create a e-commerce website for me based on Medusa. Now there's a lot more to it than that. Those of you that have been using these tools probably are even shaking your head saying it's not that simple.

It's definitely not quite that simple. There's a lot of little steps that you really have to get practice in using these tools to use. But my point overall is. Cloud code absolutely can do that. Same thing with Magento open source. If we go out there and say, build me a Magento open source site, cloud code can do it.

It can test it, it can deploy it. It's pretty insane. But let's think about then something like Shopify traditional SaaS workflow. So it's a hybrid. Some of the work, like the theming work could be done with local code, but a huge amount of the configuration and setup happens inside that web-based admin panel behind the login and an AI agent like CLO code while.

It's getting better browser use tools. It can't really see your Shopify admin panel. It can't click buttons, navigate menus, [00:09:00] fill out the forms to configure a shipping rule or add a new plugin. So you can't yet do true ag agentic automation for a project like that. So what does that mean for you? The strategic imp implications for e-commerce businesses?

I think we're gonna see a collapse of implementation costs. We're already seeing that five, 10 years ago there were these e-commerce replatforming projects would have a $500,000 budget, a million dollar budget, a $2 million budget. And the projects were. Costing that much and taking that long.

But through technical advances, platform advances and really just all of us getting a lot better at deploying new e-commerce platforms already, customization costs are come, have come down substantially. Like it is so rare that I see even a, a. $500,000 budget being necessary for an e-commerce replatforming project anymore.

I think though that you're gonna see that plummet even further. I think, way back when I started in the [00:10:00] Magento space about, oh goodness, what would that be? 15, 17 years ago? Long time ago. Where. Magento was really disruptive in the early days, is a single developer could go out and they could deploy a Magento open source site, launch an e-commerce business just through their own individual development efforts.

Now you can go out and deploy if you know how to use a tool like Cloud Code. You could build and deploy an e-commerce site running on Magento, open source, running on Medusa, running on one of these open source platforms pretty quickly. And you don't have to be as technical of a developer as you had to be in the early days of Magento in order to do that.

Most of y'all listening and watching though probably aren't gonna go try to spin up your own e-commerce site. You're probably still a kind of business that's gonna wanna hire someone to do it. But again. What used to be a six month, $300,000 project now might become a two month a hundred thousand dollars project.

And I think that AI already [00:11:00] could handle 70% of the code on a lot of the replatforming projects that I see. And it completely and totally changes the total cost of ownership calculation, especially when those advancements. And those reductions in costs, those aren't gonna apply to platforms like Shopify.

They're gonna apply to Magento open source, they're gonna apply to Medusa. I'm even seeing it, interestingly enough, I don't know if Adobe knew where things were going, saw where things were going or got lucky, but the app builder approach that Adobe has been touting and recommending for years now is also a simpler approach.

It's like a happy medium between. A truly like open source, text-based command line based system, and a SaaS with its web-based interface. So the way app builder projects and work is conducted is actually a way that is super easy for Claude Code to work on. So even those projects, I think you're gonna see the costs coming down.

And again, this is, if you're. Trying to evaluate platforms [00:12:00] right now. Good luck. I'm sorry. This is a very challenging time to do it because I think oddly enough, a lot of the calculations and recommendations you're getting are wrong because I think people are now basically underestimating how Shopify at times could be more complex or more expensive than something like Adobe Commerce or Medusa.

It just sounds weird saying it, but that is where we are in the technical space in the summer of 2025. But beyond cost, speed to market can be such a weapon and such a differentiator for so many of the brands I work with. And once you embrace these AI coding systems and these AI agents, you can spin up new features in hours.

Not days, not weeks, but hours. And so the companies that really embrace this and shift to platforms that are made for and really have embraced these AI agents they're gonna get features to market so much [00:13:00] faster than their competitors. 

They're gonna be able to respond to new market opportunities faster.

And it's gonna be just a start difference between the two. I also think that the entire moat around platforms is changing. SaaS platforms for so long had this moat around their super user friendly experience, the admin ui. I think the most valuable moat could actually become a platforms friendliness towards AI agents.

How easy is it for an AI to build on your platform? It's gonna be a critical question for the platforms and their product teams and for merchants as you're making platform decisions. So this, if you're not working with these tools, this probably sounds crazy, and this sounds like this isn't commerce today.

This is maybe commerce someday. But I am telling you these tools, I haven't seen a disruption like this possibly ever in my. 

Longer than I'd like to admit career in e-commerce development. These AI coding agents are [00:14:00] smart. They're getting smarter literally every day as we train them through using them.

And they are just, it's incredible what they can do. So this is what's happening now, and there are three ways that you need to prepare for that. So first of all, audit your customization friction. Get an honest accounting of the last custom feature you built. How many hours, meetings, and dollars did it really take from idea to launch figure out, what is the pain of your current process?

Maybe you actually already have a setup to where you're getting new features out, live quickly without a lot of dev hours needed and a lot of manual work needed. Maybe you're not gonna be as disrupted by this, but I know a lot of y'all out there, new features are a weeks long, months long, sometimes quarters long process.

The next thing I would recommend is run a AI sprint. So try it out. These AI tools, this is this is a new frontier. This is beyond cutting edge, like this truly is frontier work that is [00:15:00] changing every single day. There's not a there's a lot of people that'll sell you a lot of AI coding training courses.

There's not actually a proven playbook of here's the three steps to be successful using an AI agent to code on your e-commerce project. It takes practice. It takes. Sitting down with Claude Code, using it on your specific use case and really seeing how it does and working through that. So I would recommend.

Have one of your developers or work with an agency that is embracing AI to actually just experiment and see what does it look like to build a new feature using an AI agent to do the coding, and where are there possible friction points or problem points with either your current business setup or your current platform.

And then also add AI readiness to your vendor selection. So the next time you're looking at platforms, technologies or agencies, don't be careful how you ask this because if you say, what [00:16:00] sort of AI do you have, everybody, they have been slapping AI sauce on everything in the sales process for years now.

Instead, I would ask specifically, how is your architecture designed to work with ai, native agentic development workflows? Some big words in there and you might stump a few people with that, but really what you're trying to get to is if I wanna have an AI agent. Working in my command line, working with your tool, working with your software, is it possible?

I know MCP servers, which not gonna dive into that's a whole other episode, but it's a technology called model context protocol. It's basically an API technology for AI agents. That is the first step towards making sure that your code or your system, your service is AI friendly and ready for these AI agents.

Perhaps they might answer a question like what we just asked with, oh, we have an MCP server. If they say that, that's a good first sign. Yeah, it is. Every [00:17:00] time I stop and do an episode on ai especially this summer it's insane how much things have changed and how fast they're changing and how they're only changing even faster.

And I believe that the platforms that are embracing this AI native development world are going to empower merchants to innovate at an incredible pace faster than we have literally ever seen before. So again, I'm gonna make a short screen share. To show you this and approve this to you because outside of the AI world and the people that like to sit around on LinkedIn and talk about technology world, I don't see a lot of e-commerce business leaders and decision makers and e-commerce directors.

Really understanding and embracing the change that is coming through this AI agent revolution. So I'm gonna record that. I'll share it out on my LinkedIn. You can find me on LinkedIn as Joshua Warren with a creativity gold background behind my headshot. And if you have any questions, if your head is spinning after this [00:18:00] episode, you can always book a free 30 minute e-commerce problem solving session with me.

You'll find a link to do that on my LinkedIn profile. Would love to chat with you about this and would also don't forget to grab your copy of the E-Commerce Growth Playbook, my new book that is available on Amazon and at all fine bookstores near you. Definitely let me know what you think about the book.

And stay tuned for next week's episode of Commerce Today when we're actually gonna dive into another one of the playbooks from the book. I hope y'all are having a great summer and are embracing the really cool new things that AI agents are doing for us. Till next time.