SaaS Is Dead. Long Live Software: Why Owning Software Wins in the AI Era

p>Let me start this post with an anecdotal story about a break-up.

For the past 12 years, my company has been a HubSpot customer. I’ve long been a big fan of HubSpot – I was an early proponent of inbound marketing as they defined it. I’ve been to Inbound multiple times. HubSpot really defined an entire segment of the digital industry for a long time – and they supported agencies along the way.

Like a lot of businesses, we grew into it over time. At first, it solved a real problem. Then it solved more. Over time, it became deeply embedded in how we handled sales and marketing. And along the way, the license fees kept climbing—quietly at first, then noticeably. What was the most frustrating experience for me? When their email plugin for sales started saving every recipient I was emailing, thus driving up my contact list, which we paid for by every thousand contacts. That, in my view, was a bit too much.

Eventually, we were spending the equivalent of a nice new car every year just to keep access to a system we were only using about 10% of.

That felt normal, but gave me a slight ick. Especially as a software developer. But that’s what established companies did. You subscribed to serious software and budgeted accordingly.

This week, we dropped our HubSpot subscription to its lowest possible tier, just to keep the account alive…

• Not because CRM became less important.
• Not because we downsized.
• But because we built what we actually needed: a sales management tool with custom automation to make our lives easier – the way we do business.

That decision wasn’t emotional or ideological. Though I have to admit it was hard to pull off the bandaid…!

It was practical, and it’s one we’re seeing more organizations arrive at as AI changes what’s possible.

For a long time, subscribing to software was the safest choice. Building custom systems meant long timelines, high upfront costs, and real technical risk. SaaS won because it removed friction. You could move fast, offload maintenance, and rely on best practices out of the box.

That logic held for years, but AI has quietly changed the equation.

Just look at the stock market this week to see what the street thinks about the future of SaaS. Salesforce is down 42% in the past year. HubSpot is down 68%! These are staggering numbers. In one day this week alone, HubSpot dropped 11%, all due to the SaaS space being driving lower by yet another release of an Anthropic product.

The hardest parts of building software, such as integrations, data handling, automation, and reporting, have become dramatically faster and more accessible.

Not trivial, but meaningfully easier. The bottleneck has shifted away from execution and toward decision-making: what to build, how it fits together, and how it evolves over time.

Software is no longer something you implement once and live with for years. It’s something that can adapt alongside your business.

That’s where the SaaS model starts to strain.

As I’ve said forever in this blog, SaaS platforms are designed to scale across thousands of customers. AI-driven systems perform best when they’re deeply contextual, when they understand your workflows, your data, and your edge cases. Those two goals don’t always align.

The result is familiar to a lot of teams: platforms filled with features you don’t use, workflows that don’t quite match how you operate, and rising costs justified by capabilities that remain largely generic. If a product doesn’t do what you need, then integrate it via API to another piece of software until you get the desired results. But eventually, you do that enough and wonder why you keep the software anyway. I recently heard an executive of a Fortune 100 company say they question the value of their CRM because, at this point, it’s simply become a “Centralized Database” that they could build themselves.

This doesn’t mean SaaS is going away. Commodity tools will always have a place. But when software defines how you sell, operate, or deliver value, renting it indefinitely starts to feel like a constraint.

That’s the shift we ran into.

We didn’t set out to replace HubSpot. We set out to remove friction. AI made it feasible to build a system around how we actually work, one that integrates where it needs to, ignores what it doesn’t, and evolves without waiting on someone else’s roadmap. And most importantly, we were able to integrate automation (AKA: AI) into key points that are very unique to our business.

At this point, it’s worth pausing to say something clearly: not everyone should build.

Building software, especially now, isn’t a badge of honor. It’s a strategic decision. But there are some strong signals that suggest it’s at least worth considering.

Cost

One of the most obvious is cost. If you’re spending a substantial amount on SaaS every year—especially across multiple tools that overlap or barely get used—that’s no longer just an operating expense. At a certain point, you’re paying not for value, but for convenience and inertia.

Cost is complicated, though. If you’re spending millions of dollars on licensed software, it’s hard not to recommend custom development, especially given the advances in software development we’ve been seeing. In some cases, it may come out to be slightly MORE expensive to build your own solution. But in that case, are the benefits of something bespoke worth a slightly larger line item? Is there value there for your organization?

Cost is indeed complicated – but you have to focus on value delivered in addition to bottom-line dollars.

Workarounds

Another signal is workarounds. If your team spends a lot of time bending tools to fit how you work, such as creating manual steps, exports, custom fields, side spreadsheets, and the like, that’s usually a sign the software was never designed for your reality. SaaS works best when your business looks like everyone else’s. The more unique your workflows become, the more friction you feel.

Use Cases

There’s also the issue of use cases. If your tools almost do what you need, but fall short when it comes to automation, integration, or decision-making. That last 20% often matters the most. And it’s usually where generic platforms stop. Growth and efficiencies more often than not live in that last 20%.

Strategic Importance

Finally, there’s strategic importance. Some systems quietly shape how you compete: sales processes, operational workflows, customer experiences. When those systems are central, outsourcing their evolution entirely to a vendor roadmap can become a real liability.

None of this means ripping out your entire stack.

It means being intentional.

AI hasn’t made good software design automatic. In many ways, it’s raised the stakes. Here’s something to consider: poor decisions move faster now. Architecture, data strategy, and long-term thinking matter more than ever. Building software responsibly is still a strategic exercise—not a weekend experiment.

What AI has done is make ownership a viable option again.

So, the better question for decision-makers today isn’t “build or buy?” It’s “which capabilities are important enough to our business that we should own them?”

Some systems are fine to rent forever. Others quietly define how you operate. Those are the ones worth examining closely. SaaS isn’t dead. But the assumption that every important system should be rented because it’s easier most definitely is.

AI didn’t eliminate software. It made it malleable. It turned software from a product you rent into a capability you own. And, it’s made it possible for every company to become a tech company – at a reasonable cost!

Once you experience that shift, it changes how you look at every subscription line item on the balance sheet.

To be clear, this isn’t about convincing every company to start building software.

A big part of what we do is help organizations decide what actually makes sense to build, and what doesn’t. In many cases, the right answer is still to license a tool and move on. In others, it’s to own a specific capability that’s become too important to rent.

AI has made building faster, but it hasn’t made decision-making easier. The hardest part isn’t writing code. That’s never been the problem. It’s knowing where ownership creates leverage and where it just adds complexity.

That’s the work we spend most of our time on now: evaluating systems, understanding workflows, and helping teams design software that fits how they actually operate, whether that means building something new, extending what already exists, or deliberately not building at all.

AI expands the set of options. Experience is what helps you choose the right ones!