Have you ever loved a product and then a new version is released and, even though it does a lot more, it stops doing what you loved about it?
There’s a reason why this happens.
It’s the same reason why AI is making SaaS and business intelligence applications obsolete. And it may also be a big reason why companies are laying off waves of people in the name of efficiency and productivity.
The thing is, I predicted this, back in 2023, and I got a lot of heat for it. But if you’ll allow me another predictive stab at the future, I can also tell you how it’s going to play out.
Why Do SaaS and BI Tools Age So Poorly?
Many pieces of software are initially designed to fix a specific problem. As they scale, developers add new features to solve more problems until, eventually, the original use case is lost.
Why Products Go From Life Saver to Pain Point
We’ve all experienced the enshitification lifecycle of a product, whether digital or physical, business or personal, but let’s stick to applications and tools we use in business, to align with the focus of this article — SaaS and BI.
These apps, the ones we love, often come into our lives as a joy to use and become “must haves” in our day-to-day work. Then future releases include some odd feature choices that we don’t need, but it’s still cool. Eventually, the app becomes this bloated, awful chore we can’t get away from.
There’s a simple reason for this. As the expansion of the digital platform evolved from the personal computer to the internet to mobile, software companies could no longer survive by just solving your problems. With lower barriers to entry, increased competition, and a user reach that included every pocket on the planet, software companies had to start solving everyone’s problems or risk being swallowed by the companies that did.
SaaS brought everyone access to the tools they needed to accomplish the tasks, thus “maximizing” efficiency. BI chained together all the data to make those tasks fruitful, thus “maximizing” productivity.
The quotation marks there aren’t snarky. They’re a nod to the lesson from the product devolution lifecycle: When you’re solving everyone’s problems, you eventually wind up not solving anyone’s problem at all. As SaaS and BI adoption went mainstream, apps that were built and meant “just for you” (or so it seemed) couldn’t and didn’t stay that way for long. And the more those apps were built for a global and universal population, the more poorly they solved individual problems.
Inefficient and Unproductive, But Also Expensive
Let’s take a step back into the pre-AI world and talk for a minute about how SaaS and BI platforms do what they do.
Both of these platforms essentially allow the user to aggregate data in some way and communicate with either people or other applications to act on that data. Simply put, your banking app’s data tells you how much money you have, and its bill payment feature interacts with other entities to pay your bills.
To accomplish this magic, you have to build the databases, code up the management of custom data input and storage, write SQL to wrangle the data the way the user wants it, write the code to display it, and create entire API libraries to interact with other API libraries to get things done.
And the more you did this in a way that anyone and everyone could do anything and everything related to your market and industry niche, the more you watered down the experience — efficiency and productivity — for each individual user.
Can Apps Solve Everyone’s Individual Problems?
AI is changing the product enshitification lifecycle. Or at least that’s the intent of enterprise AI.
In 2023, when I wrote the original SaaS-is-dead post, generative AI could replace a customer support rep or be a much better version of a chatbot. That’s how most people saw it.
But here’s the thing. I helped invent some of the first commercially used NLG and generative AI, and I knew from this experience that the output of generative AI was secondary to the input, i.e., the collection and aggregation of the data. Generative AI and LLMs purported to scrape and train the world’s “general knowledge base” into a proprietary prompt-and-response platform.
OK, so now you’ve got the world’s knowledge at your fingertips, as long as you’re using the right platform and it doesn’t hallucinate.
Retrieval augmented generation (RAG) frameworks have now come online to incorporate the individual data (company data, for now) into the generative AI experience. This allows the “world’s knowledge” to be contextualized through the lens of specific use cases regarding the company’s products, services, support, or whatever databases or documents they choose to integrate, including those same databases and documents that are used to feed business intelligence systems.
Co-pilots, quickly evolving into Agents, are there to, as I said above, “get things done.” This could range from making decisions from a set of options, interacting with other applications or even writing code.
Hey AI, Make Me More Money
The dream state here is not too far from what I laid out in 2023, and I’ll use the same example here.
Right now, we’re logging in to Google Analytics and customizing the way it works and creating individual data fields and setting preferences and building reports and — let’s not forget — first learning how to do all of that. Then we’re doing it repeatedly, setting benchmarks and goals and KPIs. Then we’re noting when realities change that require tweaking. Then we’re stepping out of the digital world back into the real world to formulate the right steps to take to make progress towards those goals and KPIs…
What if, instead of all that, you could just tell your AI platform to improve your conversion rate? And it just … did it.
OK. Yeah, that’s silly. That’s the dream state. I imagine this gets said in hushed tones at the end of the sales pitch to the company CEO who instantly thinks about all the resources, time and money that can be cut from the growth and profitability equation.
That’s how AI takes out SaaS and BI.
It’s happening low-key and behind the scenes in the sense that they’re not saying they’re doing this, because in a lot of cases these are SaaS and BI companies cannibalizing their own SaaS and BI offerings.
But when we talk about the enterprise, this is where it’s heading, with maybe 100-or-so dominant SaaS and BI companies becoming maybe a dozen-or-so dominant AI companies.
I bet you can name at least five. OpenAI, of course. Microsoft and Google, of course. Anthropic is one I think everyone knows. Then the fifth one can be your choice. X? Mistral? Meta? Apple? Amazon? DeepSeek?
There are a lot to choose from. For now. You can see why everyone else – the group of all the other SaaS and BI companies who aren’t on that list – are panicking.
