AI Is Not Killing SaaS. It Is Rebuilding It Around Intent.
#150: How AI coordination layers and intent orchestration are reshaping enterprise software economics.
“Video killed the radio star.”
The Buggles were wrong.
Video didn’t kill radio. It recontextualized it.
Radio became ambient. Background. Infrastructure.
Now we are watching the same drama replay itself.
This time the lyric reads:
AI isn’t killing SaaS.
It’s dissolving it into intent.
As generative AI becomes embedded in enterprise software, the future of SaaS is shifting from application interfaces to AI orchestration layers.
The headlines are louder than the analysis. “AI eats software.” “The end of SaaS.” “Vertical AI will replace incumbents.” “The death of the subscription model.”
But the real shift is deeper and more destabilizing than product replacement.
This is not about tools getting smarter.
It is about software losing its status as the unit of value.
The shift isn’t tools → systems. It’s systems that learn and coordinate—which is the core of an agentic operating model →
1. The End of SaaS Interfaces in the Age of Generative AI
For 20 years, SaaS trained the economy to think in dashboards.
You logged into a product.
You navigated tabs.
You clicked buttons.
You exported CSVs.
Value lived inside the interface.
Generative AI collapses the interface.
With models like GPT-4o (May 2024), multimodal copilots, and API-level automation, the user increasingly starts not with navigation but with intent:
“Close the books.”
“Find churn risk.”
“Prepare a QBR.”
“Optimize inventory.”
The interface becomes optional. The model becomes the mediator.
This is why companies like Klarna announced in 2024 that their AI assistant was handling the equivalent workload of hundreds of customer service agents. The announcement wasn’t just about automation—it was about bypassing workflow software entirely. Tasks moved from UI-driven systems to intent-driven systems.
The economic signal is subtle but profound:
If users no longer need to learn your interface, your moat shrinks to the underlying logic.
And logic is now portable.
2. What “AI Eating Software” Actually Means
The phrase suggests annihilation. That’s theatrics.
What’s happening is disintermediation.
Enterprise AI adoption is accelerating not because companies want smarter dashboards, but because they want outcome automation.
The economic driver is not interface enhancement — it is coordination-level margin capture and leverage.
AI systems sit above software stacks and orchestrate them. Instead of replacing CRM, ERP, HRIS, analytics tools outright, AI agents increasingly:
pull data from multiple systems,
interpret it in context,
execute across them,
and return outcomes rather than screens.
Software becomes a backend capability layer.
This is already visible in enterprise rollouts of Microsoft Copilot, Salesforce Einstein Copilot, and Google Workspace AI integrations. These systems are not separate applications; they are orchestration layers.
When Microsoft reported accelerating Copilot adoption in enterprise accounts in 2024 earnings calls, the strategic subtext was clear: value is migrating to the coordination layer.
SaaS used to own workflow.
AI now owns coordination.
And coordination captures more leverage than workflow ever did.
So is AI replacing SaaS? Not exactly. It is repositioning SaaS beneath an AI orchestration layer.
3. The SaaSpocalypse Narrative (and Why It’s Partly Wrong)
The bearish case goes like this:
AI startups can build vertical tools in weeks.
Foundation models commoditize features.
Distribution shifts to API-level aggregation.
SaaS incumbents lose pricing power.
Startups are squeezed between hyperscalers and open-source models.
There is truth here.
But three assumptions deserve scrutiny:
Assumption 1: Models are commoditized
Foundation models are becoming more accessible. But orchestration, data rights, evaluation, guardrails, and compliance are not.
Foundation models compress differentiation through feature commoditization, making surface-level advantages increasingly fragile.
Enterprise AI adoption remains gated by governance complexity, not model access.
Assumption 2: Software margins collapse uniformly
Some margins compress. But coordination layers can increase capture by bundling previously separate workflows.
In history, abstraction layers often capture disproportionate value (operating systems, cloud platforms).
Assumption 3: Startups can outpace incumbents indefinitely
Startups move faster. Incumbents own data gravity, distribution, and trust.
AI reduces switching costs at the interface layer—but increases dependence at the infrastructure layer.
The battleground shifts. It does not disappear.
4. Real Signals from the Market (2024–2025)
Several real events sharpen the picture:
Klarna (2024): Announced AI handling two-thirds of customer service chats, materially impacting labor demand.
Salesforce (2024): Workforce restructuring while doubling down on AI productization and monetization.
Microsoft (2024 earnings): Reporting strong enterprise AI attach rates tied to Copilot.
OpenAI enterprise adoption growth (2024): Rapid uptake of API-based integration across SaaS layers.
McKinsey (2024): Estimated generative AI could add $2.6–$4.4 trillion annually to the global economy.
IMF (2024): Estimated AI could affect nearly 40% of global employment.
These are not speculative think pieces. They are operational reallocations.
These developments reflect a broader acceleration in enterprise AI adoption across SaaS platforms.
Capital expenditure is shifting from application sprawl to model integration.
The budget line item is migrating.
If the interface collapses and coordination becomes the control surface, what happens next?
Here’s where value actually relocates.
5. From SaaS Applications to AI Intent Infrastructure
SaaS companies monetize functionality.
AI systems monetize resolution.
Functionality:
“Here is a tool to perform a task.”
Resolution:
“Tell me the outcome you want.”
In a post-software world, economic activity reorganizes around:
Intent capture
Constraint interpretation
Execution orchestration
Feedback learning
The firm that owns intent mediation may not own the underlying software at all.
This reframes value creation:
Not features.
Not seats.
Not dashboards.
But decision throughput.
6. What Happens to Vertical AI Startups and SaaS Companies?
As in prior platform shifts, coordination nodes accumulate leverage through platform-level coordination advantage.
Three possible futures:
Scenario A: Vertical AI Dominance
Small, domain-specific AI firms replace narrow SaaS tools by embedding directly into operational flows.
Pros:
Faster iteration.
Tighter domain adaptation.
Lower overhead.
Risks:
Dependency on foundation model providers.
Pricing pressure.
Integration fragility.
Scenario B: Incumbent Absorption
Large SaaS platforms integrate AI deeply and retain customers via:
Data gravity.
Compliance infrastructure.
Multi-product bundling.
Pros:
Stable enterprise revenue.
Regulatory advantage.
Lower switching.
Risks:
Organizational inertia.
Cannibalization dilemmas.
Scenario C: Coordination Layer Supremacy
A small set of orchestration platforms sit above SaaS and command intent routing.
Pros:
Cross-system leverage.
High lock-in.
Network effects.
Risks:
Regulatory scrutiny.
Commoditization from open-source orchestration.
The likely outcome? A hybrid stack.
7. The Economic Impact of AI on SaaS and White-Collar Work
The more radical question is not “what happens to SaaS?”
It is:
What happens to GDP when software stops being a discrete purchase?
In classical economics, software is a capital good.
In an AI-mediated world, software becomes ambient infrastructure.
Three macro shifts follow:
1. Labor Compression in Cognitive Middle Layers
Routine cognitive work—analysis, reporting, documentation—compresses.
The IMF (2024) suggests advanced economies may see larger exposure due to task composition.
This does not eliminate jobs wholesale.
It reorganizes value toward:
judgment under uncertainty,
coordination of AI systems,
trust-bearing roles.
2. Margin Redistribution
SaaS historically enjoyed 70–90% gross margins.
If AI commoditizes features, value redistributes toward:
compute providers,
model developers,
orchestration owners,
proprietary data holders.
Margin pools shift up and down the stack simultaneously.
3. Acceleration of Firm Velocity
If intent replaces workflow navigation, decision cycles shorten.
Shorter cycles increase competitive pressure.
Markets may become more volatile—not because AI is irrational, but because latency collapses.
But macro shifts don’t determine winners. Control does.
