Ditching SaaS Platforms With AI
Microsoft CEO Satya Nadella has been widely misquoted from his interview on the BG2 Podcast as saying that, because of AI, SaaS (Software-as-a-Service) is dead.
What Nadella was really arguing is that most SaaS products are just a database wrapped in business logic and a UI, and as AI gets good at handling that business logic, the value drains out of the app itself. This shifts the build-versus-buy line that’s held for over a decade. Based on how he worded it, he seems to have been talking about all off-the-shelf software, not just SaaS. His point is a valid one, though. AI shifts the balance when it comes to the classic buy-vs-build decision.
In the blog post that follows, I’ll outline two ways we’ve seen this ourselves as we’ve shifted how we run our Trailhead away from two SaaS platforms in the just last several months. I hope it inspires you to rethink your own software situation.
The Argument
All other things being equal, nearly everyone would prefer custom software that conforms to their business processes rather than having to bend their business process around a piece of off-the-shelf software’s one-size-fits-all solution. Nobody actually likes the compromise, but the economics have historically forced their hand. By sharing the cost of building and maintaining software across many buyers, SaaS has made very capable tools more affordable to everyone. However, that shared cost comes with a built-in trade-off: the software has to be one-size-fits-all. You get something good enough, but usually not something that fits your needs perfectly.
AI-assisted development has been changing the calculation lately. For teams that know how to use it well, AI drops the cost and effort of building custom software by mulitple times. When the build cost falls that far, the line between “build it” and “buy it” moves, and that means a lot of decisions that were settled years ago might be worth re-opening.
This year we re-ran that math on two of our own systems, and both times the answer flipped from using SaaS to building our own solution. I’ve detailed those situations below for your own inspiration and edification.
Case Study 1: Moving Away From WordPress
For several years, we ran our website on WordPress. Between hosting, maintenance, and support, it cost us several thousand dollars a year.
Over the course of a couple of weeks, I was able to use Claude Code to download the entire thing, slice it up into components, and transition it to a static site that looks and functions exactly the same. I used Astro for static site generation, Cloudflare for hosting, a GitHub repo for the source, and GitHub Actions for builds.
Our new hosting bill is pennies a month. However, cost wasn’t even our primary motivation. Control was. WordPress and its plugins update constantly, and those updates would often break our website without us knowing. We got tired of discovering the site was broken when nobody had made any changes.
Of course, a static site doesn’t do that. It’s made up of static files that, once generated, don’t change. The HTML, CSS, and JS we deploy now stays exactly the same until we decide to change it. There are no silent nighttime WordPress or plugin updates. When we do want to change something, though, it’s fast and easy. No theme hierarchy or plugin conflicts to worry about. We just make changes in simple markdown files, and it quickly rebuilds the website. What’s more, we can now make changes even more quickly with the help of AI. That same AI used to struggle to make updates to a WordPress site.
For a site with almost 500 pages, the whole migration still took less than 2 weeks, with AI doing the heavy lifting. With the lessons we learned, we could do it even faster next time. Doing that work without AI would easily have taken 2-3 months, and at the going rate, it simply wouldn’t have been worth it.
Case Study 2: Breaking Up a CRM Monopoly
Our SaaS-based CRM cost us many thousands of dollars a year, but it bundled in social media automation, web form collection, and email marketing along with the core CRM functionality. Those other tools were sufficient, but not best-in-class. Because of integration, our CRM held a kind of monopoly over all our sales and marketing software.
We switched to the free tier and replaced the custom reports, forms, and automations with in-house build jobs with about two to three days of effort. We then added two inexpensive, best-of-breed tools for social media and email automations, and moved the new sync jobs out onto Azure Functions running on a consumption model at essentially zero cost. The record of who our clients are and what we’ve talked about doing for them now runs on free tooling.
Here’s the best part, though. This didn’t just let us rebuild what we had for less money. It let us build something that actually fit how we work, because we were the ones defining how it worked.
Take our spam problem. The old system dumped every form submission straight into the CRM, including all the spam and solicitations. In practice that meant the CRM was maybe 90-95% noise. The real signal was in there somewhere, buried under all the stuff we didn’t care about. In the new system we built, submissions get triaged before they ever touch the CRM, and only real leads make it in. Our CRM is a record of clients and prospects again, instead of a landfill.
Owning the system also let us wire it directly into the tools we already use every day. We can now automatically pull a client contact directly out of our M365 email and calendars and attach it to the right CRM record without anyone having to manually key in anything. Our old SaaS CRM technically had a feature that was supposed to do that, too, but it was so fragile and unreliable that we could never depend on it. We piled on to the support ticket reports and upvoted by hundreds of customers, but it never went anywhere. When you own the thing, though, edge cases like that can become first-class features instead of requests that die in a SaaS vendor’s backlog.
Where Buying Still Wins
To be clear, I don’t think companies are about to start building their own ERP systems, payroll systems, accounting packages, or video conferencing platforms. Those large-scale systems benefit enormously from shared development costs, and the economies of scale behind them are still substantial. For most organizations, buying remains the obvious choice.
What is changing is the category of software that sits closer to a company’s unique processes and competitive advantages.
Many organizations have spent years bending their workflows around software because building something tailored to them was simply too expensive. As AI-assisted development drives those costs down, some of those assumptions are worth revisiting. The build-versus-buy line may no longer be where it used to be in the past.
Is SaaS Dead?
While SaaS may not be dying off completely any time soon, its value proposition is going through a fundamental shift. That shift means questioning the old assumption that building it yourself is always the more expensive option. There are a lot of calculations worth re-running now. For some, the result will still be the same: rent your software. But for Trailhead twice already this year, and increasingly for many companies, the new answer comes back “build it”.
We help clients with this complex decision every day. If you’d like help balancing cost and control, Trailhead is ready to help.
