Some IT pros don’t need SaaS vendors

It’s always cheaper to make your coffeebot at home.

Mike Toole, director of security and IT at Blumira, previously relied on a SaaS tool offered by an outside vendor to pair up employees for informal, get-to-know-you meetups. Then Toole discovered this matchmaking app’s functionality was something his team could brew on their own.

Toole and his team created their new “coffeebot” with an AI platform, using natural-language prompts like, “I want a platform to interface with the Slack API to pull all the users from a certain channel, match users once a week, and create a separate channel for matched employees to connect and sync their calendars.”

The original SaaS app option wasn’t “terribly expensive,” but the cloud-based cybersecurity company cut costs by getting rid of it.

“​We just basically built it ourselves and added some more features that they didn’t have,” Toole said. With their own app, Toole said the team could eliminate more personal questions that the vendor’s product would ask users, as well as offer a customized explanation of the intended purpose of the frequently caffeinated conversations—hence the company’s “coffeebot” nickname.

Toole’s effort reflects a worry felt in the tech industry (and the stock market) recently: a concern that IT pros don’t need as many SaaS vendors, given how they can use AI to quickly spin up their own, highly customized apps.

IT pros who spoke with us said AI has a chance to disrupt the application layer, but the emerging tech faces an uphill battle against established vendors offering software products with built-in quality assurance and granular controls.

Fully stocked. Many analysts and CEOs don’t see software companies losing their important role anytime soon, particularly in terms of the enterprise. “AI is more likely to be additive to software workflows in the near term rather than a substitute,” a JPMorgan analyst reportedly wrote to clients recently. At an industry conference in February, Nvidia CEO Jensen Huang said the idea of AI replacing software is “the most illogical thing in the world.”

Generating ideas. Juan Orlandini, CTO at solutions integrator Insight Enterprises, sees companies turning to AI to replace SaaS functions when they’re only using a small portion of their existing SaaS products’ functionality, or when the companies need light-lift, single-use solutions that don’t incorporate a lot of intellectual property (like a coffeebot). Homegrown efforts, he said, require maintenance and security.

He argues that SaaS vendors provide decades of domain-specific business logic and access to exclusive test harnesses that LLMs can’t utilize.

“What the SaaS vendors are providing for you is a framework for you to operate your business under, so that they can actually derive the value that they bring, so you can accelerate your business that has baked-in business logic. That’s not something that you can just generate with a prompt and a quick build of an app,” he said.

The stocks of software companies, including CRM giant Salesforce, have taken a hit as investors consider the possibility of AI replacing established SaaS models.

Nadia Hansen, global AI go-to-market leader at Salesforce, said AI can replace capabilities for solving one small problem, but established vendors offer capabilities like security that can’t be easily prompted in-house.

“How are you going to make sure the latest and greatest capabilities are going to be implemented, at a quick turnaround time, without you having to test every single functionality every time you change something?” Hansen told us.

For Toole, some business-critical platforms are best left to vendors—like accounting software, which can’t afford variability.

But for low-stakes, lighter-lift situations like a coffeebot, Toole can quickly customize without having to wait for a third-party vendor to add a new feature.

“There will for sure be some SaaS apps that if they don’t iterate and grow, they will be consumed into middleware that people are running themselves,” Toole said. “But I will also say that I don’t see anyone building their own internal EDR [endpoint detection and response] anytime soon, I don’t see people building firewalls.”