E43: How to Survive The SaaS Sell Off with Dipti Desai

‍

In this episode of the SaaS Operators Podcast, we talked to Dipti Desai, the CEO of Crstl, and you get a live look at what's actually happening to software right now.

‍

Not "AI is coming for your job." Not "learn to vibe code."

‍

The real question is whether the thing you built can survive when the cost to rebuild it drops to near zero.

‍

The conversation starts with a simple observation. Monday.com is down 65%. Workflow tools without unique data underneath them are losing enterprise value fast. And the founders sitting on proprietary data assets are sleeping fine at night.

‍

And the implication is obvious.

‍

If your software is a workflow without proprietary data underneath it, the clock is already ticking.

‍

Traditional SaaS Is Three Problems

‍

Per-seat pricing is going away. Workflow software without unique data underneath it is going to be hard to retain. The old sales motion of hiring big teams to cold call doesn't hold enterprise value anymore.

‍

The only ways to survive are adding fintech, compute, or data to your model.

‍

Monday.com is the extreme example. No proprietary data. Down 65% over the last year. Purely a workflow tool.

‍

Now think about a B2B platform sitting on top of proprietary transaction data. Who does business with whom, how much, when, how. The more workflows you enable, the richer the data set gets. Nobody can vibe code that.

‍

Everyone is going to try to vibe code your product from a year ago. That is very possible. But if there's a proprietary data asset underneath, they can't.

‍

The Price Floor

‍

There's a price someone is going to be willing to pay forever to do something. If your option is to use Monday or build your own equivalent, there's a cost on both sides.

‍

At $60 a month, nobody is bothering to build a replacement. At $3,000 a month, someone on the team is building it internally.

‍

Software without AI can survive. At a very low commoditized price point.

‍

Phone contracts in India cost next to nothing because consumers switch constantly. That's what drives pricing down. We might be about to see software make switching just as easy. One button and your whole team is moved over.

‍

The cost to switch and the cost to vibe code are both moving at incredible rates. That's the problem if you're trying to discount future cash flows on a SaaS stock.

‍

The Feature Problem

‍

Everyone in DTC and ecom Twitter thinks vibe coding makes software cheaper. 

‍

Vibe coding gives you better features.

‍

There's a deck sealing company where the sales team goes door to door. The app they used to manage all of that was the dominant tool in the industry. Basically the HubSpot of home services. Everyone hated it.

‍

So they built their own. Vibe coded the whole thing, including a companion mobile app for the field team. And it was better. Because they actually live inside the problem every day. The company that built the original tool didn't.

‍

CRMs built for window cleaning companies are made by the Z class SaaS entrepreneurs. The products are horrible. Those industries are underserved. The unexpected categories might get disrupted the most.

‍

Companies are building something better than Fivetran for themselves because Fivetran can't do what they need. More traditional SaaS tools churned in the last four weeks than anyone thought possible. None of it because of price. All of it because of features.

‍

We're In The Belly Of It

‍

Agencies in ecom are heavy on OPEX. Most of the labor is people running ads, sending emails, doing routine tasks getting closer to full automation every quarter. So they're building their own tools.

‍

We're in that replacement category. Software for technically capable, OPEX heavy businesses whose teams can and will start building internally.

‍

It reminds me of my first brand. As soon as we hit reliable sales, someone posted it online. Piracy websites. Copycats reselling with slight tweaks. I had to release new stuff constantly or sales would drop off a cliff.

‍

Constant adaptation or death. That feeling is back.

‍

Consumers Aren't Vibe Coding

‍

Apps aren't about to be worthless. Mobile consumer apps were already commoditized. Many of them are performance marketing companies, not app companies.

‍

The average person is not downloading Replit on their phone.

‍

For a consumer who cares about cost, vibe coding your own app is actually the most expensive thing you could do. It's cheaper to buy the $2 a month thing. For a business spending $20,000, a completely different equation.

‍

The guy who owns a grocery store is never going to vibe code his POS system. If you sell software to grocery stores, you're probably safe. Just make it better.

‍

Building end to end solutions, taking care of customers, doing it consistently at a high bar is work. Most people don't want to do the work. AI or not, whoever shows up every day is still the one that wins.

‍

Taste Is The Only Thing That Matters

‍

Traditional product managers are middlemen between talented people. Pretty useless.

‍

But a product manager with taste, enough technical knowledge, and a few people to lean on can execute and iterate at incredible speed. What used to be an anchor on teams becomes the engine.

‍

Product management might be the most important job of the AI era. The other critical role is the enterprise seller. The end outcome with AI is that we all become pretty faces on the screen. We're here to make the other person feel good. Which is an important job.

‍

Read Sapiens. It's more relevant now than ever.

‍

We're social animals. We need to organize, communicate, build things together. That's baked into the wiring. AI changes what tasks we do. It doesn't change the fact that humans need other humans to feel like the work matters.

‍

So when people ask what jobs survive, start there. Anything that requires making another person feel understood, organized, or confident in a decision. That's not going anywhere.

‍

Ship Fast, But Ship Quality

‍

The biggest shift right now is that you have to ship product at unprecedented speed. Annual planning is dead. Quarterly goals barely hold.

‍

But the odds any single team ships faster than everyone else are actually small. If you have a family and you're not grinding 80 hours a week, it can't be about volume. It has to be about strategy.

‍

The better frame is the problem surface area. AI lets you solve adjacent use cases that would normally take two years to reach. Higher quality across more problems.

‍

The quality piece gets lost in the speed talk. There's an expedient path with AI that's lazy and probably the worst quality output. Optimizing for speed over solving real problems is the wrong goal.

‍

A few people with taste who solve real problems. That's the alpha.

‍

Anthropic Makes OpenAI Look Slow

‍

Anthropic is releasing something useful every two days. Nobody has ever seen anything like it from a single company.

‍

OpenAI is burdened by 800 million weekly actives pulling the entire company toward consumers. That makes shipping features horrible.

‍

And there's a data problem. OpenAI knows how you behave inside a chat session. Deeper than anyone in history in that one dimension. But they don't know anything outside their ecosystem. No pixel. No cross-device data.

‍

Meta knows you across Facebook, Instagram, and WhatsApp. Google has your browsing data. Amazon has your purchase history. OpenAI is uniquely a massive consumer company without a full view of the customer.

‍

Think about agentic shopping. OpenAI has no choice but to be consumer first. 800 million people are relying on them to be. Claude became the business tool. ChatGPT became the personal assistant. That divergence happened naturally, and it's widening.

‍

Second Place Might Be Fine

‍

Apple isn't in the AI race right now. Maybe they don't have to be. They already have the penetration and the user base.

‍

There's a mountain of success stories waiting. Rocket Internet watched what worked in America and copied it in Europe. Drata saw what Vanta did and out-executed them.

‍

But participation isn't enough when the stakes are this high. You actually need to win because of how much capital expenditure is required. You could end up under a mountain of infrastructure you never get out of. That's a real future too.

‍