I am sick of seeing "traditional software is dead" posts.
They are everywhere. LinkedIn. X. Every VC with a Substack. And I do not think most of the people writing them know what that actually means.
So I am going to unpack what it really means here today.
Let me start with a challenge.
Name one.
Name one AI product that has fully replaced Salesforce. Or Jira. Or ServiceNow. Or SAP. Or NetSuite. Or QuickBooks. Or HubSpot. Or Workday. Or Zendesk. Or Shopify. Or Stripe. Or Slack. Or Figma. Or GitHub. Or DocuSign.
I will wait.
The answer is zero. Not one. In the entire history of generative AI, not a single vibe-coded, AI-native product has shown up on the market and completely replaced one of the traditional incumbents.
The closest anyone can point to is Klarna. Their CEO made a big splash in 2024 announcing they were ripping out Salesforce and Workday. The tech press loved it. AI replaces enterprise software! The future is here!
Except that is not what happened. Klarna replaced Salesforce with other SaaS tools. They moved to Deel for HR. They layered AI on top and called it a tech stack consolidation. It was a PR story, not a displacement story.
So what IS dead?
The business model around software is dead. The grow-at-all-costs model. The mentality of burning cash for 5, 7, sometimes 10 years without a dollar of profit. The 500-person company running a product that 50 people could run.
Software stocks got crushed in early 2026. Salesforce dropped 26%. ServiceNow dropped 23%. The entire IGV software ETF fell 30% from its peak. But here is the thing: Salesforce posted its best quarter ever during that sell-off. ServiceNow grew revenue 21% year over year. The products are fine. The valuations were not.
The market is not punishing the software. It is punishing the model we built around the software. The insane multiples. The assumption that growth never stops. The CEOs who are one of two things: either too much ego to see that their company has plateaued, or aware of the plateau but avoiding the hard work of becoming a profitable company. Because it is way more fun to be a growth company than a profitable company.
The money did not disappear. It moved.
If you are living under a rock, here is what happened with investors.
They got scared. But not in the way you think.
When interest rates were at zero, an investor had two choices: park money in a Treasury bond earning nothing, or bet on a SaaS company growing 30% a year. Easy call. But when rates went to 5%, that same investor can earn 5% doing literally nothing. So a SaaS company burning $50M a year to grow $20M in ARR suddenly has to compete with a risk-free government bond. That math breaks instantly.
Then growth rates fell. The median public SaaS company was growing 17% in 2023. By late 2025, that number was 12%. You cannot justify an 18x revenue multiple on a company growing 12%. The multiples that existed in 2021 were priced for 30-40% growth. Cut the growth rate in half and the multiple falls by more than half because it compounds down.
Then AI created a fear tax. Even though no AI product has replaced a single incumbent (see above), the fear alone was enough to compress valuations. Markets price in perceived risk. If investors believe there is even a 30% chance that AI agents replace the need for a CRM dashboard, they price that into the stock today. The software companies are still growing. Still profitable. But the stocks trade as if the disruption already happened.
The peak was 18.6x EV/Revenue in late 2021. By 2026, the median is 6-7x for public companies, 3-4x for private ones. That is a 60%+ compression. And the bottom quartile? Trading at 1-2x. Basically liquidation value.
But here is what nobody is saying out loud.
The 20x multiples did not disappear. They left your mid-market SaaS company and walked across the street.
Anthropic closed a $30 billion funding round in February 2026 at a $380 billion valuation. On secondary markets, the company is now trading at $1 trillion. Their annualized revenue went from $9 billion in late 2025 to $39 billion in March 2026. Five months. OpenAI closed a $110 billion round, the largest private tech raise in history, at an $852 billion valuation.
Investors are not scared of spending money. They are spending more than they ever have. They just concentrated it.
On one end of the barbell: Anthropic and OpenAI getting valued at a trillion dollars because they are growing faster than anything in the history of enterprise software. On the other end: deeply embedded, deeply profitable companies like CrowdStrike and ServiceNow still commanding 15-20x because ripping them out would require an act of God.
In the middle: your company. Growing 12%. Not profitable. No AI story. No moat. That company used to get 15x. Now it gets 3x. And the CEO is in denial about it because admitting you are a 3x company means admitting the growth story is over.
The investors did not get dumb. They got precise. They stopped spreading capital across the entire category and concentrated it on the two ends. If you are a CEO sitting in the middle, you have a choice: grow explosively (probably too late) or get profitable (still possible). Sitting in the middle hoping the old multiples come back is not a strategy. It is denial.
What actually changed: the eyeballs moved.
Here is the real shift, and it is not the one the headlines are covering.
Your users stopped looking at your software.
Think about your own behavior. Where do you sit all day now? You sit in Claude. Or ChatGPT. Or Gemini. You sit in your AI tool of choice, and you want it connected to everything else. You want to say "pull my pipeline from Salesforce, cross-reference it with the support tickets in Zendesk, and tell me which accounts are at risk." You just collapsed three departments into one question. The software is still there. Salesforce still holds the data. Zendesk still holds the tickets. But you are not in those tools anymore. You are in the AI, and the AI is using the software on your behalf.
That is why Salesforce just built MCP as a major feature. That is why CLI access matters now. That is why every serious software company is scrambling to make their product accessible to agents.
Think about what MCP actually is. Salesforce, one of the most valuable software companies on earth, built a feature whose entire purpose is to let other tools talk to it. That is an admission. The front door to their product is no longer their product. The front door is Claude.
Software becomes plumbing.
Here is the evolution happening right now.
Today: You sit in Claude. You ask it to pull data, analyze it, draft a response. You are the operator. AI is the assistant. The software is the database.
Emerging: Agents talk to the software directly. An agent monitors your pipeline, notices a deal stalling, pulls the last three support tickets, drafts a save strategy, and flags you only when a judgment call is needed. You are the orchestrator. Agents do the work. Software is still the database.
Where it is going: Agents talk to agents. Your sales agent negotiates with your customer's procurement agent. Your finance agent reconciles with your vendor's billing agent. You step in for strategy, exceptions, and decisions that require judgment. You are the architect.
At every layer, the software is still there. It still holds the data. It still processes the transactions. But nobody is logging into it to click around a dashboard anymore. The dashboard is dead. The data is not.
The prerequisite nobody is talking about.
Here is the part that should scare every software CEO who is not paying attention.
APIs are the foundation. Without clean APIs, you cannot build MCP or CLI access. Without MCP or CLI, your software cannot be reached by agents. Without agent access, your software is invisible to the new way work gets done.
You become the fax machine. You still exist. Nobody uses you.
Your product could be excellent. Best UI in the category. Beautiful dashboards. None of it matters if an agent cannot reach your data programmatically. The CEO whose eyeballs are in Claude is never going to leave Claude to go click around in your dashboard. That behavior is already gone.
Most software companies built their product for human eyes, not for machine consumption. Every dollar they spent on UI design, onboarding flows, tooltips, and dashboard layouts was an investment in a delivery mechanism that is becoming irrelevant. The data underneath is valuable. The wrapper is not. And most companies cannot separate the two because they never thought of them as separate things.
Software is not dead. The way we interact with software is dead.
The companies that survive understand their product is not the screens. Their product is the data and the logic layer underneath. Salesforce gets this. ServiceNow gets this. The ones who do not get it are still hiring designers to make their dashboards prettier while the world moves to a command line.
The constructed fallacy here: "We bought good software, so we are set." No. You bought software that was built to be looked at by humans. The humans stopped looking.
Now what?
Kathy
THE AI BOSS