Stephanie Cole

We Asked Every Vendor If AI Would Kill Their Product. The Answers Were Wild.

February 13, 2026

There is a question making the rounds in every B2B buying conversation right now, and vendors hate it. Not because they don't have an answer - they have too many answers, rehearsed and polished and completely unconvincing. The question is simple: Is AI going to make your product irrelevant?

I started asking it directly. Every call, every demo, every renewal conversation. I figured if I was going to keep subscribing to a stack of tools - some of which I genuinely do not know the cost of because, honestly, someone from the office handles all of that - I should at least know whether the vendors themselves believed in what they were selling.

What I got back was a masterclass in corporate deflection. And one or two moments of startling honesty that I think about more than I expected to.

What's Actually Happening, First

The backdrop matters here because it is genuinely dramatic. The catalyst for the most recent software selloff was a single earnings call - when Palantir CEO Alex Karp announced in early February 2026 that AI had become so powerful at writing and managing enterprise software that many SaaS companies were in danger of becoming irrelevant, it triggered a selloff that wiped out $300 billion in market capitalization for Microsoft, Salesforce, ServiceNow, and others. That is not a small tremor. That is the ground shifting.

A basket of SaaS stocks compiled by Morgan Stanley saw a 15 percent decline by mid-January 2026 alone, following an 11 percent drop in 2025. Salesforce lost 26 percent of its market capitalization. Adobe fell by 19 percent, Atlassian by 30 percent. ServiceNow, HubSpot, and other industry giants experienced declines of 30, 40, or even 50 percent from their peaks.

The term "SaaSpocalypse," coined by a Jefferies trader, proved apt - hedge funds earned $24 billion by shorting software stocks.

So that's the context when I'm on these vendor calls asking the AI question. These are not idle philosophical inquiries. The market has already placed its bets. I'm just trying to figure out if the people selling me software have placed theirs.

The Five Types of Answers You Get

The first type is what I call the Integration Pivot. This is the most common response, and it goes roughly like this: "AI won't replace us - AI will work through us." As one managed services CEO put it: "I see things integrating into HubSpot or Salesforce, not replacing it." This is a real argument, by the way, not just spin. When you look at how enterprises are actually deploying AI agents in 2025 and 2026, they're not replacing their systems of record - they're building orchestration layers on top of them. The problem is that vendors say this whether it's true for their specific product or not. A payroll platform with deep compliance infrastructure saying it is different from an AI that might be replaced - that argument makes sense. A basic workflow tool using the same line? That's a prayer dressed up as a strategy.

The second type is the Feature Announcement Response. This is where the vendor tells you, with great energy, about everything they're adding to the product. AI summaries. AI-assisted writing. An AI copilot. In many cases, vendors are overpromising and underdelivering by adding AI features to tools where they are not necessary. I sat through a demo in January where someone showed me an AI button that auto-generated a subject line for an email campaign. That's a feature. It is not a survival strategy.

The third type is Honest Dread. A couple of vendors actually paused when I asked. One said something like: "That's the question we ask ourselves every week." I respected that more than I can articulate. As one industry observer noted, "there are a ton of underwhelming SaaS applications that are features not products" - and those will be the ones to go. The vendors who know which category they're in, and say so, are at least telling you the truth.

The fourth type is Confident Dismissal. This is Jensen Huang's territory. Nvidia's CEO said: "There's this notion that the software industry is in decline and will be replaced by AI" - and called it "the most illogical thing in the world," arguing instead that AI will use and enhance existing software tools rather than completely reinventing them. Some vendors parrot this. A few actually believe it, and for some of them, they're probably right.

The fifth type - and the rarest, and the most interesting - is the Full Concession. One vendor I spoke with, after some prodding, said their entire team was actively building the thing that would eventually compete with their existing product. They were essentially running a controlled demolition of their own revenue model before someone else did it for them. I found this refreshing in the way that I find a really good piece of news refreshing - it made me feel informed instead of managed.

The Honest Answer Is: It Depends on What Kind of Product You Are

Here is where I am going to land, and I am not going to soften it: most vendors are not giving you the real answer because most vendors don't know the real answer yet. But the data is pointing somewhere specific.

There are a ton of underwhelming SaaS applications that are features not products - and those will be the ones to go. The market already knows this. The SaaS boom happened because software disrupted spreadsheets - every janky process that lived in Excel got replaced by a polished product with a login page and a per-seat price tag. AI is now doing exactly the same thing to those polished products.

There are already cases where a SaaS vendor sends through their usual annual double-digit price increase, and teams are starting to ask "do we actually need to pay this, or could we just build what we need ourselves?" A year ago that would be a hypothetical question at best. Now it's a real option people are putting real effort into thinking through.

I used to just forward renewal invoices to whoever handles that. (I don't actually read them. Gerald - Linda's husband - would probably have thoughts about this approach, Linda mentions him often enough that I feel like I know the man. Anyway.) The point is: even people who have never interrogated a software cost in their lives are suddenly being handed a reason to interrogate it.

But here's the distinction that most of these vendor calls missed entirely: the determining factor for survival isn't the brand, or even the data that SaaS companies have - it's whether their core system is deterministic or probabilistic. Meaning: does your product do something precise, with consequences if it gets it wrong, embedded in regulatory and compliance infrastructure? Or does it do something fuzzy - generate content, summarize, suggest, organize - that AI can now approximate for almost free?

A Forrester vice president doubts that large-scale replacement will happen for this reason: "there are about 20,000 legal jurisdictions worldwide and complying with applicable regulation is a major reason why people trust vendors like SAP." He is right about that category. He is probably not right about every category.

What I Think Is Actually Going to Happen

I think we are about to see the sharpest bifurcation the software industry has experienced in a decade. The products that survive are not necessarily the biggest or the most famous. They are the ones that own something AI can't easily reach: if AI can replace the "business logic" layer that SaaS applications provide, then the SaaS layer survives only if it creates unique and indispensable value for customers - specifically what AI agents alone cannot offer, like industry-specific expertise, deeper integrations, or bespoke capabilities tailored to complex workflows.

The products that don't survive are the ones that built beautiful dashboards on top of workflows that AI can now handle directly. The "seat compression" phenomenon means a single AI agent can perform the tasks that previously required dozens of junior employees - for enterprise customers, the incentive to maintain thousands of individual licenses has evaporated. Why pay for 100 seats for a marketing department when three autonomous agents can execute the same volume of work using a single administrative API?

Tory - our resident life coach, who is currently navigating what I'd call a period of significant personal restructuring - said something at the last team lunch that actually applies here. He said: "The universe doesn't punish you for becoming obsolete. It just moves on." He seemed to be talking about his lease situation. But I think about it in the context of software vendors every time I sit in another one of these demos.

The vendors who stood out in my conversations were the ones who could explain, specifically, what their product does that an AI agent going straight to the API cannot do. Most could not. Your 2019-vintage SaaS app with its static dashboards and manual workflows now looks ancient next to AI-native interfaces. Users are experiencing Claude, ChatGPT, and AI copilots everywhere. Their expectations have shifted. They want natural language interfaces. They want AI that anticipates what they need. They want automation, not forms.

We had someone from the office spend most of a day running through our tool stack, asking exactly these questions. Chris helped coordinate it. He came into my office afterward looking genuinely unsettled, which is unusual for Chris - nothing seems to visibly bother him, which is part of his whole thing. He said a handful of the tools we were paying for were essentially solving problems that something like a configured AI agent could handle without the license fee. I said that was very interesting and then asked if he wanted tea. I am not sure I processed what he said in real time. I am processing it now.

What Should Actually Change in How You Buy Software

Stop asking vendors what features they have. Start asking them what their product does that AI cannot do, specifically, and make them get uncomfortable if they can't answer it.

Adding AI to a conventional SaaS solution in the form of a chatbot or other form of AI-generated add-on does not make a meaningful difference. Real modernization requires rethinking the SaaS workflow from the ground up. When a vendor shows you an AI button in a demo, that is not the answer to your question. It's a distraction from it.

The pricing model shift is also worth watching carefully. The percentage of companies still relying on seat-based pricing models fell from 21 to 15 percent within a year, while hybrid usage-based models increased from 27 to 41 percent. Gartner predicted that by 2025, more than 30 percent of SaaS solutions would integrate outcome-based pricing components. If a vendor is still selling you seats without any conversation about outcomes, that should raise a flag. It means they haven't figured out how to survive in the new model yet.

There is a very small category of tools - the ones with real compliance infrastructure, proprietary data, or deeply embedded workflow logic - that I think will come through this without much drama. For everything else, the vendor's answer to the AI question is not just interesting. It is the most important thing they will say to you. And most of them are still saying the wrong thing.

You can read more about the software stock dynamics playing out here in our piece on the SaaS crash nobody wants to call contagion. And if you're trying to figure out which categories of tool are most at risk in your own stack, the AI replatforming conversation is worth reading before your next renewal cycle.

Derek, for what it's worth, thinks the whole thing is overblown. He said the same thing would have happened to Star Wars if people had listened to the panickers. I did not follow the analogy but I nodded. He seems invested in the outcome.