Sales & Automation
AI isn't satisfied with just our jobs—it's taking our SaaS too
July 16, 2026
6-minute read
The narrative that “AI will kill / replace / supplant (your choice) SaaS” sounds like pure bullshit marketing hype at first glance (okay, and at second glance too…), but exciting things have been happening over the past 1–2 years—and are still happening all the time—that really spark our imagination. To quote a classic, technological progress can even be seen from the moon.
On the SWE-bench Verified benchmark—which measures the ability to solve real GitHub issues—the performance of leading models increased from approximately 4–5% at the end of 2023 to 93.9% by 2026, a nearly 50-fold improvement in just under two and a half years. Meanwhile, the METR Research Institute’s “time horizon” metric (which measures how long it takes an AI agent to independently solve a task with a 50% probability of success)doubled every 7 months between 2019 and 2024, but since 2024, it has been doubling every 3–4 months—meaning that the pace of AI development is itself accelerating. (METR, March 2026)
Well, this is the kind of technological advancement that—if we extrapolate it to the coming years and multiply it by the computing resources currently being built amid the AI data center construction boom—will have legacy vendors and SaaS company owners rightfully shaking in their boots at the prospect of the software development capacity set to hit the market in the near future.
In early 2026, when an Anthropic product update (which really just drew attention to this area of AI technology development) wiped out approximately $300 billion from the software industry’s market value in two days. Salesforce, ServiceNow, Adobe, and Workday plummeted 7–11% in a single day, and the sector’s forward P/E ratio fell from 39x to around 21x in a matter of months. The press has since dubbed this episode the “SaaSocalypse,” and it continues to this day: the IGV software index is down 30% from its September peak, and the cumulative loss in market capitalization is now in the $2,000 billion range.
What’s really interesting is that the market isn’t pricing in the fact that companies will buy less software, and certainly no one thinks they’ll use less of it. However, they may be willing to pay for specific functionality delivered by AI-powered developers— greater efficiency, and lower vendor lock-in risk—and because of this, the SaaS “recurring revenue” dogma, upon which the high SaaS profits and predictability of the past decade were built, has suddenly evaporated.
In short, here’s a question I ask myself as a business leader: Why should I pay a lot of money to an external SaaS vendor when, with an ROI of 1–2 years (or sometimes even within a year), I can develop the technologies needed for my company’s processes on my own?
The cases uncovered by The Information and summarized by PYMNTS are particularly telling because they include the names of specific companies, specific dollar amounts, and specific dates:
Retool’s 2026 “Build vs. Buy” survey, which polled 817 enterprise developers and operations professionals, shows a similar pattern: ClickUp’s GTM team built six internal AI tools, saving $200,000 annually on automation software while integrating Salesforce, Zendesk, and Snowflake systems. A company called Harmonic rebuilt a $20,000-per-year tool in-house because it was faster than waiting for a response from customer support—today, they run 33 internal applications with integrations for Salesforce, Gong, and Slack.
The ERPClaw case is a pioneering example: a former Accenture consultant, who previously led SAP implementations for major energy industry clients (Allegheny Power, E.ON, American Water), single-handedly built a 45-module, open-source ERP system using AI tools. While a typical SAP license costs at least $50,000 per year, ERPClaw runs on a server that costs $20 per month. The project does not (yet) promise to replace SAP on a large-scale corporate level, but it demonstrates that the business model for software categories that previously carried six- or seven-digit price tags is also set to change rapidly. (HackerNoon, 2026)
Aspen Pumps (a British manufacturer of air conditioning parts and a user of SAP Business ByDesign / SAP Cloud ERP) has built 12 automation bots with its partner—including one that extracts data from CAD drawings and automatically generates a bill of materials (BOM). Result: a total annual savings of 10,000 hours, and the BOM bot alone saves 25,000 pounds per year by eliminating errors from manual processing. (SAP, case study)
Unified Women's Healthcare (a U.S. healthcare network), in partnership with a partner, redesigned and automated its NetSuite environment (script simplification, automation, optimization), achieving annual savings of 1,500+ work hours. (Rand Group)
Successful migrations have also been documented in the Power BI/Tableau market: Jean Mandarin, head of aMatillion’s data and insights team, publicly demonstrated how the team reduced the number of reporting error tickets by 80% by switching from Tableau to an AI-native solution with a different architecture. (Velosio, 2026)
The Swedish fintech company’s OpenAI-based customer service assistant, launched in February 2024, handled 2.3 million conversations in its first month, which the company says was equivalent to the workload of 700 full-time employees;By the third quarter of 2025, this figure had grown to the “work equivalent” of 853 employees and annual savings of approximately $60 million, response times had improved by 82%, and the Net Promoter Score (NPS) reached 73 points. (Twig, 2026)
It’s not just about American companies: the founder of DmarcDkim.com, a Berlin-based cybersecurity startup, switched from about ten SaaS products to self-hosted, open-source alternatives within a year (Rocket.chat instead of Slack, Twenty instead of HubSpot/Salesforce)—partly for cost reasons and partly due to European data sovereignty considerations. The same article profiles the Warp development tools team, where rebuilding an internal documentation product took only about two days of work and resulted in a better end product that aligns with the company’s brand. (LeadDev, 2026)
Explosively growing computing capacity multiplied by the exponentially increasing efficiency of AI coding, raised to the power of the frustration of business leaders at the mercy of SaaS and legacy tech companies =
BUT. Actually, DEDEDEDEDE. Let’s be careful, though. Just because this is a promising, exciting direction for our company—one where we have a lot to gain and that could solve many of our problems—doesn’t mean we should rush headlong into that particular forest.
There are plenty of risks. In the use of AI, as well as in the technology itself. In IT security and data protection. In companies’ AI readiness, the quality of their prior digitalization efforts, and the sophistication of their processes. In legal and ethical compliance. In the employees. And in a thousand other things.
In my next few posts, I’ll try to write about these risks—what they are, how to identify them early on, and how to either prevent them or simply mitigate them.
