So much of what I do these days is research – more or less focused on all aspects of AI and the implications and possibilities of this emerging technology. Some days, I worry about the consequences of AI in terms of the economy and job displacement. On other days, I see AI as an incredible technological breakthrough that will exponentially advance the human race. For me, it's a bit of a struggle balancing these two completely opposite viewpoints – when I grew up, our TVs were big, and the phone was on the wall, after all! With all of that said, regardless of your views on AI, things are going to change across many industries, and the way businesses operate in those industries will fundamentally be altered.
One area that a lot of futurists point to is the idea of SaaS applications. SaaS has been a massive part of the online industry for the past 20 years, brought about by browsers being able to handle complicated tasks and broadband plus reliable high-speed cellular being widely available. Software providers quickly realized the power of the subscription and began to develop their platforms to be web-based and operated online exclusively. Despite their strength in the market today, there is an opinion among many in the AI space that the future for SaaS applications is not bright.
This is for a variety of reasons. Firstly, it is because of our inherent dislike of SaaS. Let’s face it – the leverage is not in our favor. SaaS companies focus on embedding you into their ecosystem, making it extremely difficult to leave. We all know how invested we are in platforms like QuickBooks, HubSpot, or Salesforce. There are entire consultancies that focus on doing nothing other than transitioning their clients from one platform to another. Therefore, they have all control over the relationship, especially around costs. And worse yet, while they're helping us by providing a service, our data doesn't live in our custody, or at least in a format we can manipulate or relocate at will.
Come on, does anyone really like using all these SaaS apps, anyway?
Enter AI and its potential benefits. The first advantage is that as AI systems get more innovative and it's easier for people to manage their LLM trained and fine-tuned with everything they know, they can use those systems for much of what modern-day SaaS applications do. Rather than start with a small example, let me go right to the behemoth and talk about Salesforce. Salesforce, as a piece of software, could be entirely transformed by a fine-tuned LLM that knows all aspects of your business. There is no reason why a company's internal "brain" couldn't store all information about its users and communicate with company personnel about those users in a conversational and all-knowing way. And even do it in their custody, on-premises.
“Hey Pete, your customer John Doe was looking at his old invoices earlier. Do you want me to email him to see if he is interested in having a discussion about billing?”
"It's been three months since you last year from Jane Doe. Should we send a short note to check in and see how she is doing?"
"I noticed that your team is spending a lot of time on AI Prompt engineering. Should we write a blog post and publish to LinkedIn about it?"
Yikes!
Think about the everyday tasks that Salesforce completes. These things can be replaced with an LLM and perhaps some easy visualizations that can be built on top of it. The craziest part about it is that the LLM will eventually be able to create those components on top of itself. If you've done any work with coding on any of these systems, you'll know just how powerful they are. And keep in mind that we're still in the first inning of AI disruption. And the dawn of AI Agents to manage all the workflows for us!
Now, that's a more long-term picture because I don't think we see this type of disruption in the SaaS space for some time. In AI terms, some time means probably three or more years. And I suspect it won’t be because of the lack of technology but rather our unwillingness to give up our old UI/UX. As King George said after the revolution: “I pray that the United States does not suffer unduly from its want of a monarchy.” We’re just trained to work in a way dictated for us – for now.
But I do think in the short term, we can see interruption in other areas. There are many examples of licensed software that hold high levels of user dissatisfaction. Still, due to either being too embedded in the use of the software or because of a lack of other options, companies grin and bear it. Does anyone really love Salesforce, anyway?
Okay, enough of the Salesforce bashing...
There's always been the build vs. buy debate, and this is something that has gone on for as long as there have been software developers and people who had customized use cases. I've made my career by designing and developing websites, but a major part of our business and the most substantial use of our development resources is building custom software.
One of the #1 reasons why people choose to buy is because budgets don't allow for the build they desire.
Until now.
AI tools have the opportunity to streamline the software development process in a way that means custom-built software is now accessible to literally everyone.
The old process was slow and awkward.
There's a reason why custom software development is expensive. It's time-consuming, and it requires a lot of components. Typically, our first phase of an engagement with a client is to conduct an in-depth discovery process, where we work with the client to analyze use cases and develop a specification. This process can cost thousands to tens of thousands of dollars. Then, we’d start design, development, and ultimately work towards deployment.
The process is now becoming faster and more iterative.
AI tools are already capable of producing rudimentary prototypes quickly. And AI can drive discovery (with some tweaks and special sauce). I've played around with the tools available for these tasks now, and they still have a little way to go, but they do a decent enough job. And while they may not do the best job, you have other options as well. You could utilize Claude or ChatGPT to generate HTML-based prototypes based on your description of what you need on those pages – and it will even work with you to advance them.
Obviously, these tools require inputs, and inputs require expertise. So, we're not quite at the point yet where somebody can say that they have an idea for the next great application and have the entire thing figured out for them. However, what we see are tools that enable professionals to accomplish tasks much quicker than before.
So now we have a situation where determining what you want to build is faster and easier than before.
But what about the actual development?
I've built some simple applications utilizing existing LLM technology that, to be honest, have blown my mind. Granted, my background was a developer, but even with a little bit of rust, I was able to produce prototypes of ideas that I've had for years.
There are now a slew of developer tools that make it faster and easier to build full-stack software applications. These systems connect to LLMs (Large Language Models) to help the developer build and define functionality quickly. We're not at the point where someone can develop a spec, copy and paste the entire thing into some system, and have the whole app built for you. Web development is so inherently complex: you have back-end considerations, front-end considerations, and infrastructure to think about. Not to mention security, scaling, accessibility… We're just not there yet.
But we are definitely in a situation where development for complicated projects can be streamlined and built much faster. As I saw someone comment on X recently: “Who would’ve known the best programming language to learn would be English?”
What a thought.
So, back to the point of this post, is it possible that AI can usher in a new golden era of custom web development?
I think it's definitely possible. At this point, if a company is considering licensing software, the mathematics of deciding to build versus buy is getting more and more compelling. It used to be, and this is purely anecdotal from my experience, that it would take upwards of eight years or more to recoup your investment from custom software. Now granted, one of the major selling points of custom software is the ability to differentiate and drive additional revenue. But if you consider the economics of where clients are when they make the decision to build, and you contemplate the license fee of the software they might need to acquire, that would be very typical math.
Now, I think it's more likely that due to acceleration and AI tools, that cost-benefit analysis has been cut in half in favor of building.
And I'm not so concerned about AI doing all of this just yet because there will need to be a significant paradigm shift for that to happen. People will trust and be dependent on user interfaces for a little bit longer.
So, with all that said, if you have ever considered custom software before, now is a time to reconsider. We all know the hassle of working around the limitations of licensed software. Imagine being able to do precisely what you want to do with the economics now being in your favor. This is truly a breakthrough moment for differentiation from your competitors and the advancement of your business.
But the time to act is soon. With AI and what can be produced, having a first-mover advantage makes a big difference.
