Will Agentic AI Obliterate SaaS? A Niche Dive into the Bain Debate on Software's Survival
Key Takeaways
- A Paradigm Shift: Goldman Sachs predicts over 60% of software value will flow through AI agent systems by 2030, moving from "Software-as-a-Service" (SaaS) to "Results-as-a-Service" (RaaS).
- The Threat to SaaS: The risk is that SaaS products become commoditized, low-margin API backbones, while agent providers capture the majority of economic value by handling user intent directly.
- Evolution, Not Extinction: The likely outcome is a metamorphosis where SaaS companies survive by becoming the indispensable, trusted infrastructure for agents, focusing on proprietary data, deep domain knowledge, and robust APIs instead of just the user interface.
Another day, another "AI will kill [insert industry here]" headline. But this one feels different.
Goldman Sachs dropped a bomb of a prediction: by 2030, over 60% of all software economic value could flow through AI agent systems, not traditional licenses. We’re not talking about AI assisting software. We’re talking about AI becoming the transaction layer for the entire software economy.
This isn't some far-off sci-fi fantasy. It’s a debate raging right now, and consulting giant Bain & Company just threw a tanker of fuel on the fire with a detailed report on how agentic AI could systematically dismantle the SaaS model we’ve spent two decades building. So, is your favorite SaaS tool about to become a fossil?
The Catalyst: Deconstructing the Bain & Company Thesis
First, let's get our terms straight. "AI" is a hopelessly broad term.
What is 'Agentic AI'? (From Chatbot to Do-er)
We’ve all used chatbots. You ask a question, it gives an answer. An agentic AI is a different beast entirely. It’s a do-er, not just a talker.
You give it a goal—"Prepare the weekly sales report, identify the top three underperforming regions, draft an email to those regional managers, and schedule follow-up meetings." It doesn't just generate text; it interprets your intent, reasons a plan, and then executes it by interacting with other applications via APIs. It logs into Salesforce, pulls data from a BI tool, opens your calendar, and drafts the email.
It’s an autonomous worker. True agency is about autonomous execution, not just clever chat.
The Core Argument: Why Interfaces Could Vanish
The core of the Bain argument is an economic one. Foundation model costs have plummeted—OpenAI’s latest models dropped costs by 80% in just two months. This makes it economically viable to shift routine tasks from a "human + app interface" model to an "AI agent + app API" model.
Why click through five screens in your CRM to log a call when you can just tell an agent, "Log my 2 PM call with Acme Corp"? The Graphical User Interface (GUI), for decades the main way we interact with software, suddenly becomes a clunky, inefficient middleman.
'Results-as-a-Service' vs. 'Software-as-a-Service'
This leads to a fundamental paradigm shift. We're moving from Software-as-a-Service (SaaS), where you pay a subscription for access to a tool, to what can be called Results-as-a-Service (RaaS).
You’re not paying for the shovel anymore; you’re paying for the hole. The value isn’t in the software license; it’s in the successful completion of the task. And if a generic agent from a hyperscaler (think Google, OpenAI) can achieve that result by stringing together a few cheap APIs, what’s the point of the expensive, seat-based SaaS subscription?
Argument for Obliteration: The Bear Case for SaaS
This is where things get scary for incumbent SaaS players. The bear case argues that SaaS becomes a commoditized, low-margin utility, like electricity or water.
The End of the GUI: When the Agent is the Application
Imagine your desktop is just a single command prompt. You don't open apps. You state intents. This is the ultimate conclusion of the agentic shift: the agent is the application.
Economic Inversion: Paying for Outcomes, Not Seats
In Bain's "AI Cannibalizes SaaS" scenario, the entire business model inverts. Instead of Salesforce charging $150/user/month, a generic agent performs CRM tasks for pennies per action, paying Salesforce a micro-transaction for API access. The vast majority of the economic value is captured by the agent provider, not the underlying software.
The SaaS tool becomes a "silent workhorse," a database with an API, and its margins evaporate.
Early Tremors: Examples of Agents in the Wild
This isn't just theory. We're seeing powerful, real-world examples of agentic workflows taking over complex jobs that once required specialized software and human operators.
A major insurance firm is using agents for autonomous claims validation, a process that used to be entirely human-driven. A UK utility company built an agentic system to handle complex outage notifications for vulnerable customers.
These aren't simple chatbots. They are complex, multi-step autonomous systems eating away at the value proposition of traditional software.
Argument for Evolution: The Bull Case for SaaS's Survival
Okay, time for the counter-argument. SaaS is not going to roll over and die. It's more likely to undergo a profound transformation.
The 'Last Mile' Problem: Handling Complexity, Nuance, and Trust
Agents are great at the common 80% of tasks. But the final 20%—the "last mile"—is fraught with industry-specific nuance, regulatory hurdles, and complex edge cases. Can you trust a generic agent to handle clinical trial data or manage a billion-dollar construction project's finances?
This is where incumbents have a powerful moat: proprietary data, deep domain knowledge, and trust built over years. Furthermore, as these powerful agents are used by more non-technical staff, the need for governance becomes paramount. You can't have ungoverned super-agents running wild in your enterprise.
Infrastructure is King: Agents Need a SaaS Backbone to Operate On
Here’s the biggest point for survival: an agent can’t do anything in a vacuum. It needs systems of record to act upon. It needs a CRM to update, an ERP to query, and a cloud platform to run on.
SaaS companies can pivot from being the interface to being the indispensable infrastructure. They become the robust, secure, compliant backbones that agents rely on to get work done.
The Hybrid Future: SaaS with a Powerful 'Agentic Layer'
This leads to Bain's "AI Outshines SaaS" scenario, which is the most likely. The SaaS product doesn't disappear; its UI just becomes secondary. The real value is in the data, the business logic, and an agent-friendly API.
We're already seeing this happen as enterprise giants build their own agentic layers on top of their core platforms. eBay's Mercury Platform, for example, uses agentic workflows to build autonomous recommendation systems. eBay didn't rip out its old systems; it built a powerful new capability on top of them.
The Founder's Playbook: How to Navigate the Agentic Shift
So if you're building or running a SaaS company today, what's the move?
From Feature Roadmaps to Intent Roadmaps
Stop thinking about what button to add next. Start thinking about what user intent you can fully automate.
Your new roadmap shouldn't be a list of features. It should be a list of goals your product can autonomously achieve for the user.
Building an 'Agent-Ready' Architecture
Your API is no longer a side project; it's your primary product. It needs to be robust, well-documented, and designed for an AI agent to use, not just a human developer. Your new customer is another machine.
Surviving the Transition: New Moats, New Business Models
Your old moat (a slick UI) is crumbling. Your new moats are proprietary data, bulletproof compliance, and undeniable trust. Your business model may need to shift from per-seat licensing to consumption-based pricing based on outcomes or API calls.
Conclusion
So, will agentic AI obliterate SaaS? No. That’s too simple a narrative.
What we're about to witness is not an extinction event, but a great metamorphosis. SaaS is shedding its skin—the GUI—and evolving. The "body," the powerful engine of data, logic, and workflow that lives behind the API, is becoming more critical than ever.
The front door to the application is moving. For many tasks, it will no longer be a login screen but a simple, goal-oriented prompt. The companies that thrive will be the ones that stop selling front doors and start perfecting the engine inside.
Recommended Watch
💬 Thoughts? Share in the comments below!
Comments
Post a Comment