SaaSpocalypse? Maybe not, but SaaS applications are changing
03. 11. 26By Beth Pariseau, Senior News Writer | Ben Lutkevich on March 5, 2026
Rumors of a ‘SaaSpocalypse’ might be greatly exaggerated, according to IT leaders, but the way enterprises interact with SaaS applications is changing for good with the development of homegrown apps generated by AI agents.
With the rise of AI coding agents that can create applications from natural-language inputs alone, major SaaS vendors’ stocks have taken significant damage over the last three months. The Wall Street Journal estimated that investor fears about AI threats to these businesses wiped out $1.6 trillion in stock value in 2026 alone. The agent-based platform OpenClaw made headlines for disrupting traditional SaaS applications and raising software supply chain security concerns, while Citrini Research published nightmarish speculative fiction about an AI-fueled white-collar job-market collapse. In the past week, Twitter founder Jack Dorsey’s fintech startup Block Inc. eliminated 4,000 jobs, citing AI automation.
The democratization of application development brought about by AI agents is putting undeniable pressure on traditional SaaS vendors, eating away at some of their value proposition to users, and ultimately, long-term revenue growth, said Angie Jones, vice president of engineering, AI tools and enablement at Block, in an interview last month with Informa TechTarget.
Jones cited the examples of Spotify and Stripe, which use Block’s open-source framework Goose to coordinate agents and code custom apps, as an indication of the real risks faced by SaaS companies, even those that offer AI agents to customers.
“Instead of going and buying an agent from companies, they say, ‘Hmm, I can take an open source agent, I can then build on top of it for my needs. I can customize it in ways that we need. I can control the security, I can control the connections to other data, and I’m not relying on a SaaS company to provide that,” Jones said. “Those sorts of layers face a bit of risk.”
Reading the tea leaves for SaaS
For power users, AI agents can create apps or features that eat away at the edges of traditional SaaS vendors’ businesses. For example, one DevOps engineer at a Fortune 100 company singlehandedly developed an internal FinOps tool to monitor AWS accounts, resulting in significant savings without vendor intervention.
Suresh Gangula, who requested that his company not be named as he is prohibited by policy from speaking about it in the press, developed the tool during a hackathon at his company, using TypeScript, Amazon Bedrock and Claude 4.5. The tool monitors usage metrics such as CPU and memory utilization, feeds them to an AI agent that analyzes them according to rules defined by Gangula’s team, and produces a score based on those metrics and the expected service level, enabling the team to quickly delete, shut down, or resize services as necessary.
In the past, the company might have looked to a FinOps SaaS provider to achieve a similar end, but none of these tools provided the single-click functionality based on internal rules that Gangula’s in-house tool provides.
“If a particular resource is an orphan, if it’s not in use, with a single click we can just delete that resource,” he said. “We don’t even need to go into a particular account, log in and then search for that resource.”
Although the tool is in limited preview, Gangula estimated that the cost savings his tool could help the company realize would be huge, citing its success with one of the company’s smaller AWS accounts among hundreds across the company.
“That particular service bill is $30,000 a month, and we identified $10,000 in savings, so I think it could lower the cost by 30% at minimum,” Gangula said. Cloud resources for workloads such as log analysis, alert correlation, and vulnerability prioritization are other functions ripe for optimization, he said.
Still, while Gangula’s example might send a chill down the spines of companies whose bread and butter is exactly that type of FinOps feature, his company is far from unplugging any existing tools just yet. The cost management tool he vibe-coded is still limited to the AWS accounts Gangula works on, and it remains to be seen whether the 30% savings rate would remain consistent with scale.
“It’s very easy to build something that is shiny, but those things don’t run properly. They are vibe-coded. They are not on a proper IT infrastructure. They are not secure,” said Fabien Cros, chief data and AI officer at Ducker Carlisle.
Block’s Jones acknowledged that while the company’s finance team was reviewing tools it could eliminate with substitutes created using Goose, it hadn’t yet unplugged any. It was also unclear as of press time whether Jones was affected by the Block layoffs, which occurred after the interview with Informa TechTarget.
Ultimately, big SaaS vendors’ margins might be under pressure, but Microsoft, Salesforce and Oracle aren’t going anywhere, said Fabien Cros, chief data and AI officer at Ducker Carlisle, a global consulting and M&A firm that has begun to embrace AI agent coding tools for citizen developers in its organization.
“It’s purely a financial analyst’s reaction,” Cros said of the Wall Street SaaS stock sell-off. “It’s very easy to build something that is shiny, but those things don’t run properly. They are vibe-coded. They are not on a proper IT infrastructure. They are not secure.”
Block’s Jones countered that concerns about long-term maintenance could also shift with a vastly lowered barrier to writing – and rewriting – applications.
“My take is that, if I can spin up a new SaaS tool for myself in a day, then why do I even need to necessarily maintain it? If it starts acting wonky the next day, I’ll just build it all over again,” she said. “The cost of these things is much cheaper now.”
Meanwhile, the productivity gains for AI-driven application development have been pronounced, Jones said. For example, business users can now query data in natural language without writing SQL or waiting for a response from the analytics teams.
“We see our salespeople are able to do things like segment 100,000 leads in an hour, whereas it would’ve taken a whole week to get through that list,” she said.
SaaS application sea change underway
While no one can say for certain what the long-term future will hold, the role of SaaS applications at companies such as Ducker Carlisle has already profoundly changed with the introduction of AI, beginning with a citizen developer program in which business users develop their own software tools.
The program gives shy business users the support they need, bold business users the guardrails they need, and insulates the most technical cohort from small use cases that non-technical users can now handle themselves, freeing them up for bigger projects, Cros said.
SaaS is still present at Ducker Carlisle, but in new ways: the firm uses a drag-and-drop platform called StackAI to help business users create projects. StackAI also hosts evaluation and adoption monitoring, so that projects that naturally gain traction get adopted in production. When user-generated projects that address core functions at the company gain traction, they get referred to the expert cohort to build and own.
“It’s like a pyramid, where you have lots of use cases at the bottom, and then you bubble up the core use cases so your central team can take over,” Cros said.
The StackAI platform is good for simple apps, but core functions still require deep monitoring and full end-to-end control. Still, Cros estimated that 80% of Ducker Carlisle’s use cases now run through StackAI, while the central tech team manages the remaining 20% of high-value apps.
Ducker Carlisle has been able to eliminate several translation, market data, and AI meeting recorder apps the team was using, opting instead to stick with Microsoft Foundry to call Anthropic and OpenAI models using homegrown agents. Overall, the company achieved a 3% cost reduction, resulting in an annual savings of $1 million.
Another SaaS shift driven by AI at Ducker Carlisle is that the company has begun selling some of the tools created by its citizen developers to clients facing similar problems.
“We are becoming kind of a SaaS provider. Not by design, but the client says, ‘I want to do XYZ.’ And we’re like, ‘Yeah, we already did that for us because we had the same issue,'” Cros said. “‘Okay, can we just buy it from you because it’s cheaper than just hiring a bunch of people from Google?'”
The challenge Ducker Carlisle faces now is sustaining use cases beyond the excitement of the initial build stage in the citizen developer program, past the plateau of fatigue.
“We have an endless roadmap of use cases that we will extract from the program and build on our own,” he said.
What enterprises should do right now
One IT industry leader said he sees the creativity and personalization that are possible with agentic systems as a return to an earlier, pre-SaaS mode for applications, when engineers built custom systems for individual customers. At a certain point, it became too expensive to build custom systems, so people began to move to off-the-shelf products, said Bill Vass, CTO at consulting company Booz Allen Hamilton.
Now, agentic systems could read your business processes and build systems exactly the way you want, allowing business users to innovate again.
Instead of logging into and paying for 10 different SaaS services to complete a process such as employee onboarding, a business could train an agent on its specific process.
However, organizations must experiment with AI while ensuring security is built in by design.
As enterprises deploy AI agents, they must ensure proper authentication, auditing, reasoning rules, memory management, and cryptography.
With the right guardrails in place, enterprises can balance SaaS vendors and internal development by being model agile—able to switch between AI models and platforms to optimize cost and performance.