Software Isn’t Dead - But Its Cozy Business Model May Be

Software Isn’t Dead -- But Its Cozy Business Model May Be

Software isn’t disappearing. But the model that made it one of the safest, smoothest, most predictable corners of the market is being quietly dismantled. For two decades, enterprise software enjoyed a golden era. Recurring subscriptions, high switching costs, and relentless digitization created an industry investor could rely on. Revenue compounded. Margins expanded. Valuations soared. Artificial intelligence doesn’t end that story — but it rewrites the financial logic underneath it. The shift underway isn’t about whether AI will “kill” software companies. It’s about how AI changes the unit of value. And when the unit of value changes, pricing changes. When pricing changes, predictability changes. And when predictability changes, so do valuations.

The Per-Seat Era: Why It Worked So Well

For decades, the dominant pricing model was simple: per seat. Each employee received a license granting unlimited use of a suite of tools. Whether it was productivity software, CRM, HR systems, or collaboration tools, companies paid a fixed monthly or annual fee per user. This model worked beautifully for three key reasons:

Budget Predictability: CFOs could forecast software spending easily. If headcount rose 5%, software spending rose 5%.

Vendor Stability: Recurring subscription revenue created steady cash flow. Once a seat was purchased and workflows integrated, it rarely disappeared.

Investor Appeal: Wall Street rewarded predictability. High gross margins plus recurring revenue streams justified premium valuations. Private equity firms loved SaaS because steady cash flows supported leverage.

Software became less like a product and more like an annuity.

AI Breaks the Seat

AI agents challenge the seat-based model because they shift the economic unit from people to tasks. In an AI-enabled enterprise a bot drafts marketing copy; an agent resolves support tickets; an automated system reconciles invoices; or a model analyzes contracts. The question becomes: Why pay per human user when machines are doing the work? Instead of seats, companies are beginning to charge for activities like conversations, actions taken, queries processed, data tokens consumed, or compute cycles used.

This is a profound shift. Under per-seat pricing, revenue scales with headcount. Under usage-based pricing, revenue scales with activity. That makes software spending more dynamic — and potentially more volatile.

The Rise of Consumption Models

Some companies have already embraced usage-based pricing.

Snowflake charges customers based on compute and storage consumption.

Databricks follows a similar model tied to processing workloads.

ServiceNow and others experiment with hybrid systems — base subscription plus usage add-ons.

The experimentation hasn’t been seamless. Salesforce initially priced its AI customer-service bot at $2 per conversation. Customers resisted. The company pivoted to a broader pricing menu: action-based billing, prepaid credits, postpaid invoicing, or flat-fee unlimited plans. This trial-and-error phase reflects an industry searching for equilibrium. Pure consumption models create flexibility for customers — but uncertainty for vendors. Hybrid models may become the bridge between stability and scalability.

What This Means for Investors

Predictability was software’s superpower. Annual recurring revenue (ARR) smoothed earnings. Investors rewarded SaaS companies with high multiples because revenue was recurring, visible and sticky. If revenue becomes consumption-driven, earnings may start to resemble utilities or even retail — subject to seasonal swings, cyclical slowdowns, or macroeconomic dips. That doesn’t mean growth disappears. It means volatility increases. Valuation multiples may compress slightly — not because software becomes less important, but because its financial profile becomes less serene.

Sticky — Just Different

Despite pricing changes, switching costs remain formidable. A company deeply integrated into enterprise platforms won’t casually rip them out. Data migration risk, trust, employee retraining and workflow disruption all create inertia.

In highly regulated industries like financial services, strict compliance requirements will entrench core software platforms rather than allow outright destabilization. Governance, auditability, data lineage, and integration with legacy systems are non-negotiable. Financial institutions will require AI solutions to address vertical specific requirements. The industry is unlikely to gamble on unproven or industry-generic vendors.

AI will more likely be developed and deployed into existing control architectures, creating structural support for workflow-owning, compliance-heavy software vendors. Software may lose its uniform revenue smoothness, but it does not lose its embedded nature.

The Bigger Shift: From IT Budget to Labor Budget

Here’s where the story becomes more interesting. AI agents are not merely tools. They increasingly resemble digital workers. And companies don’t think about workers in per-seat licensing terms — they think in terms of output, productivity, and wages. If AI agents are treated like employees, spending may shift from:

IT budgets vs. Labor budgets

Labor budgets are vastly larger. If a digital agent replaces or augments a $90,000-per-year role, paying $10,000 or even $20,000 annually for that capability feels rational. This reframes the conversation. The total addressable market for software expands dramatically if it taps into wage pools rather than IT line items. Goldman Sachs estimates US software spending could approach $2.8 trillion by 2037, driven by automation and productivity gains. Even if revenue becomes more usage-driven and uneven, the overall pie could grow substantially.

Volatility vs. Opportunity

The coming era won’t eliminate software profits. It will redistribute them. Winners will likely share several traits:

· Flexible pricing models

· Strong data network effects

· Deep enterprise integration

· Clear productivity ROI
Losers may cling too tightly to rigid seat-based models in a world that demands elasticity. The comfort blanket of predictable ARR may thin. But the opportunity expands as AI agents blur the line between software and labor.

This is where many analyses go wrong, by assuming software must converge on a single new economic model.

What is emerging is not a single pricing model, but a stratification of software economics. AI is forcing the industry to split into three distinct classes. Each carries a different risk profile, growth dynamic, and valuation logic.

Metered Utilities
These are elastic, usage-driven services that scale with activity. These businesses can grow substantially, but revenue becomes more volatile and increasingly utility-like.

Hybrid Platforms
This will likely form the majority. It combines a base layer of recurring subscription revenue with usage bands, action-based pricing, or outcome-linked caps. This model trades some predictability for flexibility while preserving an underwriting floor.

Labor-Indexed Systems
AI is embedded directly into roles and workflows, priced against labor capacity rather than seats or tokens. These systems behave less like software licenses and more like contracted digital workers.
They restore a different kind of predictability. One anchored to roles and labor economics rather than seats or raw usage.

The mistake is treating these as variations of the same model. They represent fundamentally different financial profiles.

A New Financial Identity for Tech
Software once behaved like a subscription utility — steady, compounding, almost boring. Parts of the software industry may look more like a metered service economy with dynamic, usage-based and sometimes volatile, while other parts begin to resemble digital labor infrastructure, priced against output and roles rather than licenses with both tied directly to measurable output. The question isn’t whether software survives. It’s whether investors, executives, and customers adjust to its new identity.

Software isn’t dead. But its cozy business model may be.