The challenge with the "most SaaS is untouchable if it's a system if record" is that essentially we are saying we are fine with being the COBOL and mainframes of the technical transition to AI native software.
This is also often where the technical understanding of most investors shows it's limitation - the power of coding models like Opus 4.6 is not just coming up with new stuff, but utilizing existing (mostly open source) tools and frameworks. If the so called vibe coded CRM is built on top of a widely supported open source relational database, the ability to protect that data and move it to alternative architecture if required is much higher.
Now add in the concept of "skills" which essentially is premade template of things the model will do to set these systems up, and the "SaaS killer" becomes very much a reality for many use cases.
One consideration that’s completely missing from the bear case is the competitive dynamics *between* system of record companies. The problem isn’t that enterprises will vibe code their own CRM … it’s that all of the other system of record companies will start building and selling CRM on their platforms for cheaper.
Is not the main business case of SaaS is basically to enable workflows through the cloud and deliver it to clients CHEAPER than legacy on premise systems. The trick is that that cheap revenue is valued much much much higher than the old model. Hence, everyone and ttheir granmoms are in the SaaS business last 10-15 years punting like bandits.
Now, here we are, in a new junction, where the AI is offering you a completely natural language UI, doing all the complex tasks behind the scenes and then able to do it nonstop, factory like fashion. Now, anyone who defends SaaS or its legacy "system of record", "impossible to replicate" ya da ya da are either delusional or they dont remember how we come here from the days of mainframes to pre-internet era legacy set-ups.
Just think about what happened to print media with internet in early 2000s or, linear tv with VOD. I think everything we know about SW world will inevitable change and change fast, dramatically.
My guess is that SaaS will reorient to the way that Palantir approaches problems. Act as an OS for a problem space. The more that SaaS can become all parts of the end to end for particular role types, like Salesforce for sales people, including everything that currently doesn’t get captured (conversations, email fragments, intents, offers, negotiations) the better it can add value across the stack without the frontier model companies themselves becoming the destination and instead remaining the tool.
So the generic SaaS will die and the verticals will live or die by how closely they can resemble the workload of a real role, either as an amplifier of a real person or a substitute as a bot.
The challenge with the "most SaaS is untouchable if it's a system if record" is that essentially we are saying we are fine with being the COBOL and mainframes of the technical transition to AI native software.
This is also often where the technical understanding of most investors shows it's limitation - the power of coding models like Opus 4.6 is not just coming up with new stuff, but utilizing existing (mostly open source) tools and frameworks. If the so called vibe coded CRM is built on top of a widely supported open source relational database, the ability to protect that data and move it to alternative architecture if required is much higher.
Now add in the concept of "skills" which essentially is premade template of things the model will do to set these systems up, and the "SaaS killer" becomes very much a reality for many use cases.
One consideration that’s completely missing from the bear case is the competitive dynamics *between* system of record companies. The problem isn’t that enterprises will vibe code their own CRM … it’s that all of the other system of record companies will start building and selling CRM on their platforms for cheaper.
AI agents are still bound by traditional APIs, which just adds a reasoning layer on top of small data.
I still don't see how AI agent can distrupt enterprise software ?
Is not the main business case of SaaS is basically to enable workflows through the cloud and deliver it to clients CHEAPER than legacy on premise systems. The trick is that that cheap revenue is valued much much much higher than the old model. Hence, everyone and ttheir granmoms are in the SaaS business last 10-15 years punting like bandits.
Now, here we are, in a new junction, where the AI is offering you a completely natural language UI, doing all the complex tasks behind the scenes and then able to do it nonstop, factory like fashion. Now, anyone who defends SaaS or its legacy "system of record", "impossible to replicate" ya da ya da are either delusional or they dont remember how we come here from the days of mainframes to pre-internet era legacy set-ups.
Just think about what happened to print media with internet in early 2000s or, linear tv with VOD. I think everything we know about SW world will inevitable change and change fast, dramatically.
My guess is that SaaS will reorient to the way that Palantir approaches problems. Act as an OS for a problem space. The more that SaaS can become all parts of the end to end for particular role types, like Salesforce for sales people, including everything that currently doesn’t get captured (conversations, email fragments, intents, offers, negotiations) the better it can add value across the stack without the frontier model companies themselves becoming the destination and instead remaining the tool.
So the generic SaaS will die and the verticals will live or die by how closely they can resemble the workload of a real role, either as an amplifier of a real person or a substitute as a bot.
What about the "middle men", Dev Houses that developed software for companies? Pivot to AI product dev only hope?