Tired vs. Wired: Our Deep Dive on Why Software Spend Is Up Record Amounts … Yet Half of SaaS Is Still Dying

Two things are true in B2B right now, and they look like they can’t both be true.

Total software spend is growing 15% this year, the fastest in a decade, up from 12.8% last year. Gartner has it going from $1.2T to $1.4T.At the same time, public software is trading at a discount to the S&P 500 for the first time ever, and leaders like Monday, HubSpot, and Atlassian got cut 60% to 70% in a couple of months.- Spend is accelerating. Yet for many software leaders, valuations are collapsing. Both at once.

That’s the whole story of B2B in 2026, and it resolves the moment you stop looking at “software” as one thing. The market has split in two. One group is tapping AI budget and re-accelerating, in some cases to numbers we’ve never seen at scale. The other is running the same playbook from 18 months ago, waiting for a recovery that is not coming. There’s very little in the middle.

I walked through where that line actually falls, what moves a company from one side to the other, and what we learned running it ourselves with three humans and 21 agents. This is the full breakdown. I’ll be blunt where it helps, because the gap between the two camps gets wider every quarter, and the cost of being honest about which one you’re in keeps going up.

Here’s where it stands, in 10 points.

1. Vibe coding your own CRM is dead. The agent that books deals is what matters.

We vibe coded almost every app running SaaStr AI 2026. You can build incredible apps today, with or without a developer, in 1/20th the time. But the idea that you’re going to vibe code your own Salesforce is over. Go try it on Replit or Vercel right now. Say “build me my own Salesforce” and see what you get. It won’t work, and even if it did, it wouldn’t have the security, the collateral, or the integrations with the 100 apps in the building.

Who cares if you can build a mediocre clone of some segment of the functionality of Salesforce? Nobody wants a prettier leads tab with a purple gradient. They want deals on the calendar. Some vendors at this event are doing millions per week in new revenue, not because their opportunity tabs are nicer, but because they put real deals on your calendar. That’s worth $50K to $100K to even the smallest company.

2. Customers will prefer a great AI to a mediocre human.

Stop fighting the hallucination conversation from last year. Train an agent for 30 days, keep updating it weekly, and the guardrails are good now.

QB, our AI VP of Customer Success, ran the front line for 100+ sponsors at this event. QB reached out, found their issues, solved what it could autonomously over email and chat, and summarized the rest. 120 sponsors were mostly happy. This is not a chatbot. When we ran human CS, one year our two people didn’t send a single email until after the event because they were “still learning SaaStr” two months in. QB knows if you submitted your booth graphics on time. QB knows everything.

A great AI that solves the problem today beats a mediocre human who needs an engineer and an appointment next week. Build the great AI.

3. We went from 20+ humans to 3 humans and 21 agents, and we do more than before.

In 2024 we had 20-something humans. Today we run SaaStr with three humans and 21 agents, and we do more than we did with the bigger team.

Last year we had exactly one agent, a Delphi chatbot trained on everything including the WiFi password. It closed a $60K sponsorship deal on its own and we realized this was real. Out of last year’s event we went all in. Artisan for outbound first, then more, now 21 agents orchestrated and working together.

The math is worth sitting with. Our AI VP of Marketing and AI VP of CS cost $257 a month combined. They replaced roughly $500K of employees, and productivity went up, not down. Call it productivity or call it deflation. Either way it’s a lot to reflect on.

Nothing we did is clever. We just pushed the envelope a year before 99% of people. Anyone in the room can do their own version.

4. The SaaS-pocalypse is real, and the public markets gave up.

For the first time ever, public software is trading at a discount to the overall S&P 500. The whole point of software was that it was the best business model ever invented. Write once, sell everywhere. It always traded at a premium. Now the markets are saying it’s worse than almost any other business they can find.

They stopped believing 110% NRR is trustworthy. They decided the budget in the CIO’s office is going to AI spend, and that three-year contracts are hiding the fact that nobody renews in three years. They sent epic products like Monday, HubSpot, and Atlassian down 60% to 70% in a couple of months.

Founders have to let this go. It doesn’t change what you have to build.

5. But software spend is re-accelerating. You just have to grab it.

Even with the selloff, Gartner has total software spend going from $1.2T to $1.4T. That’s 12.8% growth last year to 15% growth this year. It’s accelerating faster than we’ve seen in a decade.

The catch is that about half of what CIOs are spending is net new for AI, and a lot of it goes to Anthropic and friends. To fund it, they’re cutting and consolidating older vendors. So it may not feel like spend is accelerating if you’re an older vendor. You have to tap AI budget or you feel none of it.

6. Re-acceleration is happening to old companies, not just AI natives.

This is the proof it’s not too late:

• Palantir went from 27% growth two and a half years ago to 85% today, with Karp projecting over 100% at their scale. It was a dog a couple years ago. People were leaving. Not now.
• Twilio was dead 18 months ago, single-digit growth, activists pushed Jeff Lawson out. Then the agents needed communication infrastructure. ElevenLabs and others run on Twilio. Growth went from ~4% to 20%.
• Datadog blew out the quarter because every AI hyperscaler runs on it. Same Datadog we always loved, new tailwind.
• Atlassian built a Rovo agent good enough that customers will actually pay, and growth moved to 32%.

The pattern: be the API or the platform that everything blowing up in AI also has to buy.

7. Buy, don’t build. Even though we build every day.

We build apps daily and I still tell you to buy. If a product here works for you, buy it. It’s not worth building.

We only build the 10% no vendor will ever make. Like the parking pass app. The event center hands us a 5,000-page PDF of custom parking passes. We used to pay a human to find each page, customize it, and send it, with mistakes and complaints and turnover. Amelia spun up an agent in a couple hours on Replit and automated the whole thing. No vendor is going to build that. No dev shop is patient enough. That’s when you build.

That parking pass automation wasn’t even possible until January or February of this year. The LLMs got that much better. If your own product isn’t dramatically better than three months ago, you get an F, because you should get a free boost just from the models improving.

8. The cheat code is consistency, not a brand.

People assume our agents only work because SaaStr has a brand. That’s not it. You have to earn a brand, but you really don’t need to go that far for your industry to have heard of you.

The reason our agents work is that we work with them every single day. We train them, iterate them, talk to them, make them better daily. Tools matter, but consistency beats everything. There’s no “set it and forget it” in agents yet. The daily reps are the whole game, and anyone can copy that.

9. Make your API agent-friendly. This might be the single highest-leverage thing you do.

When agents pick you, you show up in the LLM recommendations. Founders building agentic products will choose you because you’re the easiest to build on. We moved our email to Resend because it was the most agent-friendly platform. We’re leaving Marketo, our most dated software, because it runs out of API calls in an hour and makes you wait to get more.