RISING WATERS - AI & The Drowning of Traditional Software
The Water Is Rising. Can You Swim?
There is a simple, unforgiving truth about drowning: once the water goes over your head, the rest is just physics. It doesn't matter how tall you are, how well you once swam, or how confidently you told everyone you'd be fine. The water doesn't negotiate. It just rises.
Artificial intelligence is that water. And for an enormous swath of the software industry, the tide is already lapping at the chin.
In the first weeks of 2026, over one trillion dollars in market capitalization evaporated from software companies. The so-called SaaSpocalypse wiped approximately $285 billion from software stock valuations in a matter of days. Salesforce's stock plummeted nearly 40% from its 2025 high. ServiceNow's stock cratered 11% in a single day—not because AI failed to deliver, but because it delivered too well. Customers were canceling seats, not adding them.
Mustafa Suleyman, CEO of Microsoft AI, recently offered a timeline that should make every SaaS executive pour something stronger than coffee: virtually all white-collar tasks that involve sitting at a computer will be automated within 12 to 18 months. Accounting. Legal. Marketing. Project management. All of it.
This isn't theoretical anymore. This is the weather forecast. And the storm surge is organized into three distinct waterlines—each one marking the depth at which a category of software stops treading water and starts sinking.
The Three Waterlines: A Framework for Software Survival
Not all software drowns at the same depth. The metaphor is deliberate: the water rises uniformly, but what determines survival is how deep your roots go. The more contextual depth a piece of software possesses—the more it knows about the specific, gnarly, exception-laden reality of a domain—the longer it keeps its head above water.
Think of it as three concentric rings of defensibility, each one harder for AI to replicate. The further out you sit, the sooner you drown.
Waterline
Category
Examples
Timeline
1: Legacy File Format
Low-context, broad utility
Word, Excel, Notepad, basic design tools
Now – 12 months
2: Legacy Interface Provider
Broad context, packaged workflows
HR SaaS, project management, marketing automation
12 – 36 months
3: Context Service Provider
Deep domain, nuanced workflows
Field service, medical practice, trade management
3 – 7+ years
4: ???
Unknown
TBD
Eventually
Waterline 1: The Legacy File Format
Status: Water at the nose. Gurgling sounds imminent.
What It Is
These are the broad-utility, low-context applications we've used for decades to perform basic knowledge work: writing, calculating, organizing, presenting. Microsoft Word. Excel. Google Docs. Notepad. Basic design tools. They are the digital equivalent of the hammer and screwdriver—universal, reliable, and increasingly beside the point.
Why It's Drowning
Here is an honest question: When was the last time you opened Microsoft Word to think?
If you're up to speed on AI—and if you're reading this, you probably are—you don't write in Word anymore. You write in ChatGPT, Claude, or Gemini. You draft, iterate, restructure, and polish inside a conversation with an AI. Then, and only then, do you copy-paste the output into Word. Not because Word adds value to the writing process, but because .docx is still the transit vehicle that gets your work from point A to another human's inbox.
Word hasn't been a writing tool for months. It's a file format with a formatting bar. And that's a terrible reason to exist.
The data supports the shift. Publicis Sapient has reported actively reducing traditional SaaS licenses by approximately 50%, including major platforms like Adobe, by substituting them with generative AI tools and chatbots. AI-native app spending jumped over 108% in the past year, according to Zylo's 2026 SaaS Management Index. The old utilities aren't being improved—they're being bypassed.
What Survival Looks Like
These tools won't vanish overnight. File formats are sticky. Enterprises have compliance requirements, template libraries, and deeply ingrained workflows that keep .docx, .xlsx, and .pptx alive as containers. But make no mistake: the application layer around these formats is becoming a thin shell. Within months, the primary interface for creating documents, spreadsheets, and presentations will be conversational AI, not a toolbar with 400 buttons nobody uses.
The irony is exquisite: this very whitepaper was drafted entirely in Claude, then exported to Word. The file format survived. The application didn't.
Waterline 2: The Legacy Interface Provider
Status: Water at the chest. Still standing, but the current is getting stronger.
What It Is
These are the broad-context SaaS applications that package domain knowledge into structured workflows. Think HR management systems like BambooHR or Workday. Project management tools like Asana or Monday.com. Marketing automation platforms like HubSpot or Marketo. They understand the rules, roles, and rhythms of a functional domain and wrap them in a pre-built interface.
Why It's Drowning (Slowly)
The value proposition of a Legacy Interface Provider was always twofold: (1) it knew the domain better than you did, and (2) it gave you a structured way to act on that knowledge. An HR platform knows the onboarding workflow. A CRM knows the sales pipeline. You didn't have to build the logic; you just had to fill in the fields.
AI is dismantling both halves of that proposition.
On the knowledge side, large language models are rapidly absorbing the contextual understanding that made these platforms valuable. Every HR workflow, every project management methodology, every marketing automation sequence—it's all training data now. AI doesn't need BambooHR to know what an I-9 form is or when to send a 90-day review reminder.
On the interface side, something more subtle and arguably more devastating is happening: disintermediation. Users are increasingly interacting with these tools through AI rather than with them directly. Consider the pattern: I don't open Canva to design a social graphic anymore. I tell ChatGPT to use Canva to design it for me. I don't navigate BambooHR's interface to look up a policy; I ask Claude to check it. The tool is still doing work, but the user never sees it. The UI becomes invisible.
This is the abstraction layer effect, and it's lethal to SaaS economics. When AI becomes the primary interface, the underlying tool gets commoditized. If users never see your UI, they don't care whose engine is underneath. And when they don't care, switching costs collapse. As Bain & Company noted in their 2025 technology report, when one user equipped with AI agents can accomplish the work of five traditional employees, the per-seat pricing model that has underpinned SaaS for two decades begins to implode.
What Survival Looks Like
Legacy Interface Providers won't all drown at once. The smart ones are already repositioning as context APIs—making their domain knowledge and workflows available to AI agents rather than insisting users interact through their proprietary UI. The ones that resist, clinging to their dashboards and seat licenses, will watch their moats evaporate like morning fog.
Gartner predicts that 35% of point-product SaaS tools will be replaced by AI agents by 2030. That's a conservative estimate. The 65% that survive will be the ones that figured out how to be useful to AI, not just alongside it.
Waterline 3: The Context Service Provider
Status: Water at the ankles. Plenty of time—but the tide is coming.
What It Is
This is the deep end. Context Service Providers (CSPs) are software systems built for specific, highly specialized domains: eye-care practice management, field service management for plumbing and electrical contractors, construction project management, veterinary clinic operations. They don't just know a category; they know the exceptions, edge cases, and regulatory quirks that accumulate over 20 to 30 years of serving a particular industry.
A field service management platform doesn't just schedule jobs. It knows that a residential HVAC callback in Arizona in July has different urgency rules than one in Maine in January. It knows the licensing requirements differ by county. It knows that the technician's truck inventory needs to sync with supplier lead times that vary by region and season. This isn't broad knowledge—it's deep, gnarly, scar-tissue knowledge built from decades of real-world operations.
Why It Survives the Longest
AI is brilliant at absorbing broad knowledge, but it struggles with the kind of contextual depth that CSPs have baked into their systems. As SiliconANGLE noted in their 2026 predictions, the biggest gap in enterprise AI is the distance between transactional data and the why behind it. A knowledge graph can traverse the data, but it typically misses the nuance. That nuance—the decades of domain-specific exceptions—is precisely what Context Service Providers have encoded into their workflows.
The data moats here are real and deep. A CSP's competitive advantage isn't its interface or its brand; it's the accumulated institutional knowledge of an entire industry baked into business rules, validation logic, and workflow automation. You can't just prompt your way to that understanding. You have to earn it, customer by customer, exception by exception, over years.
But They're Not Immune
Even CSPs will eventually face the same disintermediation pattern. The AI abstraction layer is coming for them too—it'll just take longer. When it arrives, the most modern CSPs will have already transitioned their value from providing the interface to providing the context. They'll become the knowledge backbone that AI agents call upon, not the screens that users stare at.
HSBC's contrarian analysis supports this trajectory. Their argument: software won't be devoured by AI; it will be the delivery mechanism for AI. Agents must operate within the parameters and permissions defined by software. The bounded agent—one constrained by a CSP's deep domain rules—is precisely what enterprises need for AI risk management. This reframe suggests that CSPs don't just survive; they become more valuable in an AI-first world.
Waterline 4: The Unknown Depth
Status: Below the surface. We can't see it yet.
Every honest framework has a blank space at the bottom, and this is ours. Waterline 4 is the depth at which even Context Service Providers run out of air. We don't know exactly what it looks like yet, but the shape is starting to form.
Perhaps it's the point at which AI doesn't just absorb domain context but generates it—creating new operational knowledge faster than human experience can accumulate. Perhaps it's when AI agents don't just orchestrate existing workflows but design better ones, rendering the accumulated wisdom of 30 years of software development quaint.
Or perhaps Waterline 4 is the moment when AI doesn't need software at all—when the concept of a discrete application dissolves entirely into a continuous, adaptive, context-aware intelligence layer that simply does whatever needs doing, without the abstraction of tools, apps, or interfaces.
We don't know. But it's coming. And the organizations building for Waterlines 1 through 3 had better keep one eye on the horizon.
So What? The Strategic Imperative
If you're a software company, this framework isn't academic. It's a survival map. The question isn't if AI will reach your waterline; it's when. And the answer to when is determined by the depth of your context.
For Legacy File Formats (Waterline 1):
The game is already over. Your only remaining value is the file format itself. Pivot to being the world's best rendering engine for AI-generated content, or get acquired by someone who will. Microsoft clearly understands this—their investment in Copilot isn't about saving Word; it's about making Word the output layer for AI.
For Legacy Interface Providers (Waterline 2):
You have 12 to 36 months to transform from a user-facing application into an AI-facing context service. Open your APIs. Make your domain knowledge accessible to agents. Stop measuring success in daily active users and start measuring it in API calls from AI systems. The companies that survive this waterline will be the ones that realized their interface was never the product—their knowledge was.
For Context Service Providers (Waterline 3):
You have the most time, but not infinite time. Your competitive advantage is real, but it's not permanent. Invest aggressively in deepening your domain knowledge, enriching your data, and building the semantic layers that will make your context accessible to the AI systems that are coming. Be the engine, not the dashboard. Be the knowledge, not the screen. The CSPs that thrive will be the ones that embrace being the context backbone of an AI-native world.
The Water Doesn't Care
Here's the thing about rising water: it doesn't care about your revenue, your brand equity, your installed base, or your roadmap. It doesn't care that you have 50,000 customers or that your NPS is 72. It rises at its own pace, driven by forces far larger than any single company's strategy.
The organizations that survive aren't the ones that deny the flood. They're the ones that understand where the water is and build accordingly.
Waterline 1 is breached. Waterline 2 is rising fast. Waterline 3 is the high ground—for now.
And somewhere below the surface, Waterline 4 is waiting.
The question isn't whether you can swim. It's whether you've built something tall enough to stand on.
