The Subscription Model Built an Industry. AI Is Now Breaking It.
I've watched enterprise software go through two complete economic reinventions in my career.
The first: license and maintenance to cloud and subscription. That shift took a decade, destroyed some incumbents, and created trillion-dollar companies.
The second is happening right now. And it's moving faster.
🎯 The per-seat subscription model the engine that powered SAP, Salesforce, ServiceNow, Workday and an entire generation of SaaS is structurally breaking down.
AI didn't just add a new feature category. It broke the core assumption that revenue grows when headcount grows.
📊 What the numbers are saying
For roughly 20 years, enterprise software companies optimised around seats. Add users, grow annual recurring revenue, expand multiples. That model powered the rise of an entire SaaS generation whose valuations rested on predictable per-user subscriptions.
AI agents are now doing work that used to require named employees. A customer support platform powered by AI may resolve a growing share of tickets autonomously. Charging per human support representative becomes less intuitive when much of the work is automated.
By 2025, 85% of SaaS leaders were using some form of usage-based or hybrid pricing model. High-growth SaaS companies using hybrid models showed 21% median growth.
➡️ The direction is clear. The debate is only about how fast.
❌ The model that's breaking
Seat-based pricing worked because it aligned three things simultaneously: customers could forecast spend based on hiring plans, vendors enjoyed recurring revenue visibility, and public markets rewarded net booking and retention built on seat growth.
AI agents complicate that structure. The link between headcount and software revenue is weakening.
📌 Think about what that means for a CFO approving a renewal. The old question was: how many users do we have? The new question is: how much value did the software actually produce?
Those are entirely different conversations. And most vendors aren't ready for the second one.
🎯 What's replacing it and it's not one thing
There is no clean successor to the subscription model. What's emerging is messier and more interesting.
Instead of charging per user, vendors are experimenting with pricing tied to tokens consumed, workflows executed, transactions processed, or measurable business outcomes delivered.
✅ The models gaining ground:
→ Consumption / usage-based , you pay for what runs. Tokens, API calls, compute minutes. GitHub Copilot moved from unlimited to usage-based in 2025, charging $0.04 per extra request beyond a base allowance.
→ Outcome-based , you pay for what gets resolved. Zendesk charges $1.50 per AI-resolved ticket. HubSpot links tiers to customer metrics.
→ Hybrid , a base subscription floor with consumption layers on top. Hybrid models combining base subscriptions with variable usage tiers are expected to be dominant in enterprise AI by 2026. Valueships
→ Credit systems , buy a pool of AI capacity, allocate across use cases. ServiceNow went this route. So did others.
🔑 None of these are simple to buy, forecast, or govern. That's the problem hiding inside the opportunity.
💡 What I find most interesting
Only 16% of SaaS incumbents have commercialised AI as a standalone product. But those that have report two to three times higher customer traction and revenue and per McKinsey & Company
That gap is the tell.
Most incumbents are still treating AI as a feature add-on a surcharge on top of an existing subscription. Microsoft added Copilot to Microsoft 365 and raised prices by $3 per user per month. Google embedded AI into Workspace at no added cost.
Two entirely different bets on where value will be captured.
➡️ One is protecting the subscription model. The other is betting the subscription model is already commoditised.
I think Google is reading the room more clearly. But the enterprise buying cycle will take years to prove it.
📊 The margin problem nobody is talking about loudly enough
Subscription SaaS at scale ran at 70–85% gross margins. That's what made the model so attractive.
AI-first businesses are targeting 60–70% gross margin at scale closer to a cloud services company than a pure software company, but with higher growth potential.
Vendors often lure customers with generous pilot credits, yet scaling to production routinely reveals 500–1,000% cost underestimation.
📌 That's not a rounding error. That's a business model risk disguised as a pricing conversation.
CFOs who signed AI deals in 2024 based on pilot economics are finding very different numbers in 2026. The FinOps function built to manage cloud infrastructure is now scrambling to cover SaaS consumption costs it was never designed to track.
🔑 The thing that doesn't get said
The shift from license to subscription wasn't just a pricing change. It was a complete reorganisation of how software companies thought about sales, customer success, renewal, and product investment.
The shift from subscription to consumption-outcome models is the same scale of change. Maybe bigger.
Traditional hand-offs between sales, customer success, and renewals are dissolving. Go-to-market, product, billing, finance, and investor relations are all being restructured around the new models. McKinsey & Company
✅ What this means for buyers:
→ Your renewal conversation is changing value proof, not seat count
→ Your budget volatility is increasing consumption is harder to predict than headcount
→ Your vendor relationships are changing outcome accountability is real now, not just a pitch
✅ What this means for vendors:
→ Pricing is now a product decision, not a finance one
→ ARR as a health metric is becoming unreliable without usage context
→ The companies that survive this will look more like services businesses than pure software
🤝 Where I'm watching most closely
The incumbents will adapt slowly. They have installed bases to protect and Wall Street expectations to manage.
The interesting signals are coming from the edges vendors who built for AI-native economics from day one, who never had a seat-based model to defend.
👉 The companies building clean consumption or outcome models with transparent unit economics will define what enterprise software pricing looks like in five years.
The ones layering AI surcharges onto 2018-era subscription structures are buying time, not building moats.
We spent 20 years convincing the enterprise world that subscriptions were the future.
What do we call it when the subscription becomes the legacy?
Sources
→ McKinsey & Company : Upgrading Software Business Models to Thrive in the AI Era (Sep 2025) · mckinsey.com
→ L.E.K. Consulting : How AI Is Changing SaaS Pricing (Mar 2026) · lek.com
→ Flexera : From Seats to Consumption: Why SaaS Pricing Has Entered Its Hybrid Era (Feb 2026) · flexera.com
→ PYMNTS : AI Moves SaaS from Subscriptions to Consumption (Feb 2026) · pymnts.com
→ Getmonetizely : The 2026 Guide to SaaS, AI, and Agentic Pricing Models · getmonetizely.com