I remember when getting a document typed meant finding someone who could type. Then came the Lanier word processor, and suddenly we felt like we were living in the future. Then the Compaq luggable showed up — a "portable" computer roughly the size and weight of a small engine block — and we knew everything had changed.
What we didn't know was that decades later, we'd still be buying our way through every problem.
Let me explain what I mean by that.
For most of my career, technology decision-making in corporations has followed a reliable and deeply human pattern: something hurts, so you buy something to make it stop hurting. A CTO or CIO's job, if we're being blunt about it, has largely been sophisticated vendor management. Find the right product, negotiate the right deal, implement, complain, renew. Repeat.
Then SaaS arrived and democratized the whole circus. You no longer needed IT's blessing to buy a tool. A department head with a credit card and a 30-day free trial could onboard a platform by Thursday. This felt like freedom. And in many ways, it was.
But freedom without a plan is just chaos with better branding.
The result is that most mid-to-large companies now sit atop what I can only describe as a tech Jenga tower — dozens of SaaS products, each solving one slice of one problem, each with its own data silo, its own API quirks, its own customer success manager calling every quarter. These stacks weren't designed. They accumulated. Like debt. Like clutter. Like the apps on your phone you haven't opened since 2021.
And now, into this magnificent mess, walks AI.
The SaaS industry's response has been entirely predictable: bolt AI onto everything. Every platform now has an AI feature. An AI assistant. An AI-powered dashboard. AI that summarizes your AI. The problem is that each of these AI layers is built to play nicely within its own ecosystem — its own data, its own integrations, its own walled garden. You end up with fifteen different AI tools that can each see fifteen percent of your business, each convinced it has the full picture.
This is not a technology problem. Well, it is. But more importantly, it's a psychological one.
For the first time in most executives' careers, buying something won't fix this. There is no product you can purchase that will untangle what years of product purchasing created. The companies that figure this out first will have an almost unfair advantage over those still waiting for the right vendor to save them.
Here's what I think actually has to happen, and I'll use a metaphor because I find they clarify things that spreadsheets obscure.
Think of a moon shot. Stage one is a massive rocket — expensive, powerful, burns fast, and falls into the ocean when it's done its job. That's your current tech stack. Some of it served you well. Appreciate it. Then let it go. You will not get to the moon clinging to stage one.
Stage two has to be yours. Built for where you're going, not where you've been. Powered not by a vendor's roadmap, but by your own understanding of your business, your customers, your data, and the AI tools — many of them raw, many of them evolving hourly — that can be configured and composed into something purpose-built for you.
This requires something the product-buying era never really demanded: a plan. Not a vendor selection. Not an RFP. A genuine, strategic, slightly uncomfortable reckoning with what your business actually does and what it actually needs.
I spent a career watching smart people in large and small companies alike reach reflexively for products the way the rest of us reach for our phones. It's a deeply conditioned behavior. Products feel like decisions. They feel like progress. They generate invoices, which generate line items, which generate the sense that something is being done.
But the AI era is, among other things, a forced intervention on that addiction.
The executives who thrive in the next decade won't be the ones who found the best AI product. They'll be the ones who finally put down the catalog, looked at their business with clear eyes, and built something that was actually theirs.
The moon isn't going to wait.
.png)
.png)
.png)
