Last week, the AFR ran a piece about an Australian ecommerce brand, Craft Club, that set out to save $18,500 a month by building its own software with AI. They specifically named the software they were replacing. Australian startups included.

Uncomfortable reading if you're a SaaS founder. Hopeful reading if you've recently stared at your SaaS invoices wondering what you're actually paying for.

The story writes itself: AI kills SaaS. Vibe coding saves retailers. Everyone cancels their subscriptions and net profit goes through the roof.

I don't buy it. Not fully, anyway.

The thing about building your own

AI has made it cheaper and faster to spin up bespoke tools. A motivated founder or a halfway-decent dev can now build something that covers 70% of what a $500/month subscription was doing. That part is real, and honestly, you should be experimenting. Not to rip out your core stack, not yet, but to solve the specific, annoying problems your existing software either can't or won't prioritise.

But 70% leaves a lot of room for error. Custom builds don't come with roadmaps. They don't get automatically updated when APIs change. They don't have customer success managers you can yell at call. You don't have other retailers pushing the limits alongside you.

Brands saving $18,500 a month today might be spending $150,000 in opportunity cost over 12 months when their homegrown stack needs maintenance, integration work, and a full rebuild because the AI that wrote it in 2025 has already been superseded.

Or AI custom-builds genuinely take over, meaning no outages, no salespeople playing games, and the world is truly your oyster. It's too early to tell. Both outcomes are plausible.

This isn't an argument against experimentation. It's an argument for understanding what you're actually buying when you buy SaaS. And right now, most SaaS companies aren't doing a great job of explaining that.

The real pivot: Service as Software

Here's the flip. SaaS isn't dead. It's just being asked to justify itself in a way it never had to before. For two decades, the model was simple: build the platform, charge the subscription, occasionally go out for a client lunch. The software was the product. The service was the wrapper around it.

That model is cracking. Not because the software isn't good, but because AI is commoditising what the software does. If your core function can be replicated by a founder and an AI agent on a Tuesday afternoon, the invoice is hard to justify.

The most defensible SaaS position right now isn't faster updates, more integrations, or bolted-on chatbots. It's becoming something an LLM can't replicate overnight. It's the ability to provide a service that solves problems from a position of genuine expertise, deep partnerships, and unique data that no internal team can build with a terminal window and a $99 LLM subscription.

None of that is software. It's a service, packaged inside software. The flip is Service as Software.

What this actually looks like in practice

A few patterns worth watching.

If you're an analytics platform, the moat isn't the dashboard. It's the aggregated, anonymised data from hundreds of ecommerce clients that tells a brand not just how they're performing, but how they sit relative to their category, their margin bracket, their time of year. A custom build gives you your own data. Nothing else. You can't vibe code first-party benchmarks.

Shopify is pivoting better than anyone. Their shift from "here is an ecommerce platform" to "here is the infrastructure layer for commerce" is the clearest example in the industry. UCP, App Store, Shopify Payments, Shopify Capital, Shopify Markets. They are no longer the software. They are the ecosystem. That's almost impossible to replicate with a weekend AI build - for now.

The partnership advantage is just as real. Over the last few months, we’ve seen the love fest between Google, Shopify and Klaviyo go to the next level. That's a network effect being built in plain sight. Clients inside that ecosystem get first access to new channels, new data connections, and new automations before they're available anywhere else. The brands building their own tools are opting out of that network, often without realising it.

And then there's community. Notice how many events are being hosted, not just sponsored, by SaaS platforms right now? That's deliberate. Bring people together, create relationships, build belonging. That's a moat that has nothing to do with the software and everything to do with who else is in the room.

The new role of SaaS

The brands most likely to churn to custom AI builds are the ones that never really understood the value they were buying in the first place. That's partly a product failure. Mostly, it's a positioning failure.

If your renewal conversation is about features versus cost, you've already lost it. If it's about intelligence, risk absorption, network access, and expertise that no internal build can replicate, that's a different negotiation entirely.

SaaS isn't dead. It's got an identity problem.

The era of charging $499 a month to be a prettier spreadsheet is over. The era of charging for collective expertise, aggregated intelligence, and first-mover advantage is very much open for business.

Flip the acronym. Service as Software, not Software as a Service. If there's no longer any service underneath it, Master Claude awaits you.

Cheers,
Bushy

Add To Cart is brought to you by

Need help making a big call or sense-checking your strategy? You can book a consultation session to tackle challenges, test assumptions or connect with the right people in ecommerce.

I also partner directly with retailers and tech providers on team coaching, transformation strategy and advisory roles. If that sounds like what you need, get in touch, and we’ll take it from there.