Cal RiveraSoftware Vendors Are Terrified of Agents and They Should Be
April 1, 2026
I want to be honest about where I was when I fully processed this story. It was a Tuesday night, probably around 11pm. I was parked - well, sitting - in the parking garage of my old building, on the third level, because the wifi from the office still reaches up there and my phone data situation is a little complicated right now. Marcus had forwarded me the Futurum piece about the SaaS selloff, which was kind of him even though I know he was really just sending it to the group chat. But I read the whole thing and then I sat with it for about twenty minutes in the dark, which is more time than I usually give industry news, and I thought: yeah. Yeah, that's exactly right. The fear is correct.
Here is what is happening. The SaaS sector is in freefall. Not a correction. A structural repricing. Agentic AI is not just threatening how software vendors make money - it is threatening whether they make money at all. And what makes this moment genuinely interesting, and genuinely frightening if you run a software company, is that the vendors themselves know it. You can hear it in the earnings calls. You can see it in the pricing pivots. You can feel it in the kind of panicked bundling that looks like strategy until you read the fine print.
The Numbers Are Not Subtle
Let's start with what actually happened in the market, because this isn't vibes - there is data here that should make every SaaS founder lose sleep.
The SaaS index fell 6.5% in 2025 while the S&P 500 was up 17.6%. That gap is not noise. Then February 2026 arrived. The tipping point came during what's now widely called the "SaaSpocalypse" - a frantic 48-hour window in which the iShares Expanded Tech-Software ETF plummeted as institutional investors simultaneously realized that the productivity gains from agentic AI were not accruing to the software vendors, but to the end users and the AI model providers.
As of late March 2026, a brutal "Great Repricing" has wiped more than $1 trillion in market capitalization from the sector, as investors grapple with a world where autonomous AI agents are rapidly making the traditional per-seat licensing model obsolete.
A trillion dollars. Let that number sit there for a second. The median EV/Revenue multiple for public SaaS companies stands at 5.1x, down from the pandemic peak of 18-19x. The market isn't being irrational here. It's doing what markets occasionally do right: pricing in a structural shift before most people are willing to say the words out loud.
Mentions of "agentic AI" and "AI agent risk" on Q4 2025 and Q1 2026 earnings calls doubled compared to the prior quarter. CEOs who spent 2023 and 2024 pretending agents were a complement to their products are now publicly acknowledging them as competitive headwinds. That's not a small thing. That's a confession.
Why This Is Different From Every Other "Disruption" Story
I coach people for a living - or I used to, in a more officially structured way - and I've noticed that people use the word "disruption" so often it's lost all meaning. Cloud disrupted on-premise. Mobile disrupted desktop. Every three years someone declares that the old model is dead and then the old model... mostly survives, updated but recognizable.
This is different. And the reason it's different is specific and structural.
SaaS grew for two decades on a simple assumption: companies grow, they hire more people, and each new person needs a login. Per user, per month. A tidy, compounding machine. Agentic AI doesn't just help users do their jobs faster. It reduces the number of jobs that need a human seat in the first place.
That's the mechanism. It's not that agents are better software. It's that agents break the fundamental growth engine of the entire SaaS revenue model. The model assumed human headcount would keep climbing. Agents interrupt that assumption at the source.
One of the dirty secrets of the SaaS industry is that it's not that different from running a gym: your best customers are often those who pay for memberships - or in this case, seat licenses - they don't use. An agent, by definition, actually does the work. Every time it does the work, it removes a reason to pay for a seat. The more useful the agent, the more the vendor's revenue logic collapses.
In December 2024, Microsoft CEO Satya Nadella declared on the BG2 podcast that "SaaS is dead." The comment set off a shockwave across the technology industry. After all, software-as-a-service has defined enterprise computing for nearly two decades, representing a massive share of IT spending in 2024 and forming the backbone of digital transformation strategies worldwide. But Nadella wasn't being contrarian for sport. He was describing a structural problem his own company is also trying to solve. That's the tell.
The Panic Is Showing in the Pricing
Nothing reveals a company's fear like how it prices things. And if you want to watch a company thrash in real time, watch Salesforce try to figure out what to charge for Agentforce.
Salesforce has now shipped three different pricing models for Agentforce in roughly 18 months. First, $2 per conversation when they launched. Then Flex Credits at $0.10 per action in May 2025. And now per-user licenses starting at $125/user/month with what Benioff frames as "digital labor." Three pricing models. All running at the same time. On the same product.
I don't say this to mock them. I actually think there's something almost admirable about the chaos - it suggests they're paying attention to customer feedback rather than just defending a bad model. But it also reveals a company that doesn't yet know what its product is worth, or more precisely, doesn't yet know how to capture the value without destroying the customer relationship in the process.
What began as a simple-sounding $2 per conversation fee quickly met with customer confusion and backlash, forcing Salesforce to overhaul its approach. By 2025, many companies were in belt-tightening mode, with boards scrutinizing ROI on AI projects rather than green-lighting them on hype. So you had Salesforce charging for something nobody fully understood, in a currency nobody could forecast, during a period when CFOs were specifically told to stop approving things they didn't understand. No wonder it landed badly.
Nate pointed out last week - we were arguing about something unrelated, he was doing the thing where he defends a bad take by bringing up other bad takes - that Microsoft is doing the same bundling move. Microsoft is bundling AI capabilities into higher-priced seat licenses to get around the per-seat problem - announcing in December 2025 that core suite pricing goes up in July 2026, with Copilot Chat and security features folded in. It's the same instinct: wrap the new scary thing in the old comfortable structure and hope nobody does the math on whether they're actually getting more value.
What the Smart Money Is Actually Saying
The vendors who are not panicking are the ones who figured out that the unit of value has changed. Sierra AI uses pure outcome-based pricing - they only get paid when their agent successfully resolves an issue without human intervention. Co-founder Bret Taylor calls it "paying for a job well done." Sierra hit $100M ARR in just 21 months, then crossed $150M+ ARR by early 2026. That is not a company that is scared of agents. That is a company that built its entire commercial model around the thing everyone else is terrified of.
IDC predicts that by 2028, pure seat-based pricing will be obsolete, with 70% of software vendors refactoring their pricing strategies around new value metrics, such as consumption, outcomes, or organizational capability. Seventy percent. That's not a niche trend. That's the entire industry having to rebuild its commercial engine while the plane is already in the air.
And this is where it gets uncomfortable for the mid-market SaaS companies - the ones running point solutions, the ones that do one specific thing with a clean interface and a per-seat price that always felt just slightly too high. Gartner says: "By 2030, 35% of point-product SaaS tools will be replaced by AI agents or absorbed within larger agent ecosystems of major SaaS providers." That number isn't evenly distributed. It's going to absolutely destroy some categories and barely touch others.
Marketing automation, basic CRM, project management tools - categories that looked bulletproof five years ago now face serious questions about their long-term viability. I use several of these tools. I used to use more of them. The honest question I ask myself now when I'm renewing anything - and I ask it from a position of someone who has recently been forced to scrutinize every line item for reasons I don't need to go into - is: could an agent do this for me without this interface? And the answer is increasingly yes.
The Part Nobody Wants to Say Out Loud
Helen mentioned this at lunch on Friday - Harold had apparently asked her about it, because Harold apparently follows tech news more closely than any of us realized - and she framed it perfectly without meaning to. She said Harold wanted to know if the "software companies were going to be okay." And the honest answer is: some of them, yes. But not because they deserve to be. Because they moved fast enough.
That phrase - "workflow ownership" - is the key. For twenty years, SaaS vendors sold you workflows. You paid to use their interface for a task, and the task lived in their system. Agents don't use interfaces. They don't need them. They interact directly with APIs and databases without requiring graphical interfaces, orchestrating actions across multiple systems simultaneously. Where a human might log into three different applications to complete a workflow, an agent accesses all three systems concurrently, extracting data, applying logic, and executing actions in seconds rather than minutes.
The interface was the product. The interface is now optional. If you built a company around people needing to use your interface every day, you are in a genuinely difficult position and I think you know it.
Publicis Sapient reports actively reducing traditional SaaS licenses by approximately 50% - including major platforms like Adobe - by substituting them with generative AI tools and chatbots. An executive at the firm explains that AI agents are "10x faster, 100x smarter" than junior staff, creating a redundancy that directly cannibalizes the seat-based revenue underpinning commercial SaaS models.
Fifty percent. That's not a pilot. That's a budget decision that has already been made.
A Specific Thing I Think Businesses Should Pay Attention To
I ran an email sequence last month from that same parking garage - different night, different crisis, same wifi - using one of the tools I've had on autopilot for two years. While I was waiting for it to send, I found myself thinking: I'm paying for this every month and I use it maybe twice. An agent could do what I'm doing right now, and it would do it faster. I caught a segment error before it went out. But barely. Growth is sometimes a near miss.
The point is: I'm not a CIO. I'm one person with a modest stack and a complicated personal situation. And even I am starting to look at my software subscriptions the way you look at a gym membership you've been meaning to cancel. The CRM tools, the email platforms, the workflow software - all of it is now subject to a question that didn't exist two years ago: is this interface earning its seat cost, or is it just where my data happens to live?
For businesses running lean, this agentic shift is genuinely an opportunity. The biggest immediate effect for many businesses will be drops in pricing as CIOs negotiate better deals. Vendors are scared. Scared vendors negotiate. If you're renewing anything in 2026, you are in a better position than you were in 2024 and you should act like it.
But the bigger thing - the thing I keep coming back to - is about knowing which parts of your stack are actually safe. 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 tools with deep data, complex workflows, regulatory dependencies - those survive. The tools that exist primarily to give someone a place to click a button? Those are the ones with a problem.
"The way we're going to know who's going to be alive in 2030 is to see who launches really great AI agents that are doing full and complete jobs this year in 2026," said Tatyana Mamut, a former Salesforce VP. That's the right test. Not "are you adding AI features." Are you shipping agents that actually do the job.
The Optimistic Take, Which I'm Required to Have
I'm constitutionally obligated to end on something constructive, not because the site requires it but because it's literally how I function. Marcus would confirm this. I sent my 6am motivational text this morning from level three of the parking garage, and I sent it because I actually believe it. That's not performance. That's coping mechanism fully integrated into professional identity.
So here's what I actually think: the vendors who survive this are going to build better software than we've ever had. Not because they want to. Because they have to. The ones who figure it out won't do it by adding an AI tab to their settings page - they'll do it by making a real bet on what their product is actually for. Not "we help people do X" but "we make X happen." That's the difference. The companies that cross that line get to keep charging. The ones that don't become infrastructure someone else builds on top of.
Some of them will make that bet correctly. And when they do, the software that comes out the other side of this - priced on outcomes, built around agents, earning its revenue by proving value rather than by making cancellation annoying - will be genuinely better for the people using it.
The fear is correct. The outcome isn't settled. The vendors who are scared and doing something about it are the ones worth watching.
I'll be in the parking garage if you need me.