In February 2026, the software industry experienced a massive shock as over $1 trillion in market capitalization was erased from software stocks. Fears that AI agents will replace traditional software triggered the “SaaSpocalypse,” sending Salesforce tumbling nearly 30% and Intuit down 34%. In a single trading session, $300 billion in market value evaporated.
I’m confident that this panic selling represents a major mispricing based on flawed logic. SaaS is not dead, it is evolving in response to the emergence of Agentic AI. While the learnings are still emerging, it is becoming clear that to thrive in a world of AI agents, software companies must rearchitect their business in three areas: Strategy, Product, and Pricing.
Implications on Strategy
Vibe coding is not an existential threat; CRUD is dead; the codebase is table stakes
The sell-off crystallized three learnings for SaaS investors and executives that have implications on software vendors’ strategy: The threat of “Vibe Coding” is overblown; CRUD software is obsolete, and competitive moats live outside the codebase.
The threat of “Vibe Coding” is overblown
Vibe coding empowers non-coding folk with the power to write software code, using AI. I now write scripts in Python and VBA, which is tremendous for my productivity. However, vibe-coded tools lack the reliability, security and maintainability required for company-wide deployment. Frameworks used for agent orchestration (e.g. OpenClaw, Goose) are not robust enough to meet enterprise security standards – though that may change in the future. Last but not least, companies who develop their own code base will have to maintain it, especially if this code replaces enterprise applications. Many will be reluctant to do so, as it’s not a core function or a core capability for most companies.
CRUD software is obsolete
Applications that simply execute basic CRUD (Create, Read, Update, Delete) operations to organize data behind a user interface are obsolete, as AI agents can replicate this functionality. CRM and project management software are an example of this category.
Competitive moats are found outside the codebase
Because code is easily generated by AI, sustainable competitive advantages are structural, not technical. Vendors must focus on building moats that cannot be prompt-engineered, including:
- Deterministic Outcomes Moat: Software executing exact, mission-critical functions (e.g., payroll or tax calculation) have a deterministic outcomes moat, because errors are unacceptable. Meanwhile, AI agents are prone to hallucinations,
- Data Moat: Platforms that serve as the “source of truth” for a company’s proprietary data (e.g., transaction histories, clinical outcomes) remain highly defensible via a Data moat. Think Salesforce for Sales, Hubspot for Marketing, Netsuite for Finance. AI agents require this data to function.
- Compliance Moat: Vertical software built for regulated industries, such as financial services (e.g., PCI), healthcare (e.g., HIPAA), or government (e.g., FedRAMP) has a compliance moat. AI agents cannot hold a FedRAMP license, or navigate strict compliance architectures.
- Liability Moat: While AI can write code, it cannot accept legal liability for a system failure or data breach, meaning companies still require a software vendor to own the operational risk and associated liability.
- Workflow Moat: Software integrated into cross-team operations retain a workflow moat due to the change management effort, staff retraining, and data migration required to replace it.
Implications on Product
10x is the new reality, the API is the product
While “vibe coding” is not an existential threat for SaaS vendors with moats, failing to adopt AI agents into the core business of shipping code, is. Vendors face a 10x productivity reality check. In addition, vendors need to prepare their product for a world where AI agents go mainstream. In a world of AI agents, the API becomes the product.
10x productivity is the new frontier
When it comes to agentic AI, software vendors are the proverbial canary in the coalmine. AI agents are uniquely suited to write and test code, and promise to unleash the productivity of software teams. I believe that software vendors that do not harness AI agents face an existential threat, as they will be left behind. Software development teams must learn to use AI agents to multiply their code output.
Software quality assurance teams must learn to use AI agents to test an order of magnitude more test cases, that previously went untested. Product owners must learn to use AI agents to deliver backlogs of new features and functionality faster than ever. DevOps teams must learn to use AI agents to accelerate code release into production.
The API is the Product
If AI agents become a key tool for work, AI agents will interact more effectively with an application through an API than through a GUI. AI agents can potentially trigger hundreds of API calls in the time it takes a human to move a mouse. Therefore vendors will need to harden their backend infrastructure. A robust, documented and scalable API will become more important than an easy to use, responsive GUI. The GUI will not disappear but will likely transform into a governance dashboard—a “Control Tower” for humans to audit the AI’s reasoning audit trail, and set approval guardrails.
Implications on Pricing
Seat-based licensing is at risk, explore a pivot to consumption-based and / or outcomes-based pricing.
The possible demise of seat-based pricing
If AI agents can autonomously perform the work of software end users, albeit much faster, organizations will be incentivized to shrink the number of seats they license. This will upend the traditional SaaS revenue model. We are in the first inning on this shift, so for now the risk that end customers are keeping their license, but reducing seats just needs to be monitored actively.
Prepare a pivot to outcomes and / or consumption pricing
If the seat-based pricing model fails software vendors because one AI agent is now doing the work of ten employees, they will need to price their product in a way that does not erode revenue. Vendors can consider at least two options to replace seat-based pricing: consumption-based and outcome-based pricing. With the former, vendors charge for API usage e.g. how many API calls are triggered.
With the latter, vendors charge customers for results achieved (e.g. resolve a customer query, for a CCaaS software). Both models allow vendors to capture a share of the productivity value they facilitate. Both also protect vendors’ gross margins from the variable compute costs introduced by AI. However, neither is a silver bullet.
Buyers like the seat-based model because it’s predictable for budgeting purposes. In contrast, neither consumption or outcomes-based pricing model is. Both are inherently variable, and thus unsuited to easy budgeting. Moreover, an outcomes-based pricing requires a precise definition of what is an outcome that a customer will pay for. I discuss pricing in greater detail HERE.
In summary, the “SaaSpocalypse” is a call to innovate, not an extinction event. What’s clear is that, as an industry, SaaS has become much more dynamic, and the bar for excellence has been raised. SaaS companies need to strengthen defensible competitive moats, harness AI agents to 10x their output, pivot to an API-first product architecture, and explore new ways to charge beyond seat-based pricing. SaaS executives will have their hands full of concerns beyond just hitting next year’s revenue target.
What are your thoughts on the threats and opportunities of Agentic AI for SaaS? How is your company adapting to the age of AI agents? Share your experience and thoughts in the comments below!
Mores articles on Revenue
Receive exclusive Benchmarks and Research directly in your inbox Subscribe to our ASTERI+
Monthly Summary of ASTERI Insights and more Subscribe on Linkedin
