The software industry has changed its business model three times in the last 30 years. First came perpetual licences: pay once, own forever. Then came SaaS: pay monthly per seat, access from anywhere. Now comes GaaS — and the transition is moving faster than the previous two combined.

This is not a theoretical framework. It is being implemented in live products, pricing pages, and enterprise contracts right now. If you run a SaaS business or buy SaaS software, understanding GaaS is not optional — it is the lens through which the next five years of enterprise software will be built and bought.

What Is GaaS?

GaaS stands for Agentic as a Service (also written as AaaS). The term gained mainstream traction when NVIDIA CEO Jensen Huang used it at GTC 2026 on March 16 to describe the next era of enterprise software. The core idea is elegantly simple: instead of paying for software your employees use, you pay for outcomes that AI agents deliver on your behalf.

To understand why GaaS matters, it helps to trace the lineage. The cloud era gave us IaaS (Infrastructure as a Service — pay for computing resources), PaaS (Platform as a Service — pay to build on a platform), and SaaS (Software as a Service — pay to use a software product). Each layer abstracted away a lower-level concern. GaaS follows the same naming logic: it abstracts away human execution. You don’t buy a tool for your employees to use. You buy the completed work itself.

The distinction sounds subtle. Its implications are not. As Gennaro Cuofano put it precisely: “SaaS charges for access to tools. GaaS charges for the execution of outcomes.” That sentence rewrites every metric, every pricing page, every sales motion, and every customer success playbook in enterprise software.

Jensen Huang’s GTC 2026 Declaration

NVIDIA’s GTC 2026 conference ran from March 16–19 at the SAP Center in San Jose. Jensen Huang delivered the keynote on March 16 to a sold-out arena with 30,000+ developers in attendance and 450 companies sponsoring. The address ran just over two hours and covered the Vera Rubin chip platform, DLSS 5, the OpenClaw agentic framework, and physical AI. But the statement that reverberated furthest through the tech industry was this:

“Every single IT company, every single company, every SaaS company will become a GaaS company, no question about it. Every SaaS company will become an agentic-as-a-service company, and every engineer will carry an annual token budget alongside their salary.” — Jensen Huang, NVIDIA GTC 2026 keynote, March 16, 2026

Huang didn’t hedge. He didn’t say “SaaS will evolve” or “AI will transform software.” He said every SaaS company will become a GaaS company. The word “every” is doing significant work there. The prediction is not about market leaders or AI-native startups — it is about the entire enterprise software industry.

The infrastructure context matters. Huang also announced that NVIDIA projects at least $1 trillion in AI compute demand through 2027 — up from a previous $500 billion projection. He positioned OpenClaw (an open-source agentic AI framework he compared to Linux) as the operating system for the agentic era, and announced NemoClaw as its enterprise-secure deployment layer. The hardware, the framework, and the pricing model are all pointing in the same direction: a world where tokens are the unit of production, agents are the workers, and outcomes are what enterprises pay for.

Why Per-Seat Pricing Is Breaking Down

Per-seat pricing worked because, for two decades, there was a reliable correlation between the number of software users and the value delivered. More employees using a CRM meant more deals being tracked. More agents on a support platform meant more tickets being handled. Seats were a reasonable proxy for value, and they were easy to measure, audit, and invoice.

AI agents have shattered that correlation. When one AI agent can resolve hundreds of support tickets without a single human login, charging per seat becomes a measurement of something that no longer reflects the value being created. According to Profitwell’s 2026 benchmark, seat-based pricing has already dropped from 21% to 15% of SaaS companies in just 12 months — the fastest structural shift in SaaS pricing since the move from perpetual to subscription licensing in the early 2010s.

The problem cuts in two directions simultaneously. From the buyer’s side: a company that deploys an AI agent to handle customer support might go from needing 50 software seats to needing 15 — but the value delivered could be identical or greater. Per-seat pricing punishes the customer for becoming efficient. From the vendor’s side: Salesforce built Agentforce, which is growing at 114% year-on-year — but it is actively cannibalising its own seat-based revenue, because customers using Agentforce need fewer human operators. Salesforce built a product that eats its own business model.

Outcome-Based Pricing: How It Works

Outcome-based pricing sounds straightforward: charge for results, not access. In practice, it requires four conditions to be true simultaneously. Chargebee’s 2026 AI pricing guide identifies them clearly: you must control the work being done, you must be able to prove the results, you must be able to predict the range of effort, and the customer must already measure that outcome as part of their own business operations.

When all four conditions exist, outcome-based pricing creates a fundamentally different economic relationship between vendor and customer. The vendor’s incentive becomes delivering more of the outcome faster and more reliably — not maximising seat counts. The customer’s risk is reduced to near-zero, because they only pay when value is created. And expansion becomes automatic: as the customer’s outcome volume grows, revenue grows without a sales call or a contract renegotiation.

Chargebee describes this as the “outcome-as-acquisition” model — where the pricing itself functions as a growth mechanic. Intercom’s $0.99 per resolution removes the buyer’s risk (“I only pay when it works”), which accelerates initial adoption. It creates a natural expansion path (more tickets resolved = more revenue). And it aligns incentives precisely: Intercom only makes money when its customer gets value. That is structurally different from a SaaS model, where the vendor is paid regardless of whether the software delivers value.

Real-World GaaS Examples Already Operating at Scale

The GaaS model is not a future state. Multiple companies are billing on outcome-based models in production right now, with verified revenue figures.

Intercom Fin has processed over 40 million resolved conversations at a 67% resolution rate. At $0.99 per resolution, that is a nine-figure ARR product growing at 393% annualised. The math is straightforward and the proof of concept is undeniable. Zendesk followed with $1.50–$2.00 per automated resolution, validating the model at a different price point. Salesforce’s $2.00 per Agentforce conversation represents the largest enterprise software company in the world publicly endorsing the transition.

The Four GaaS Pricing Models

GaaS pricing is not a single model. It is a family of approaches, each suited to different types of work, different degrees of outcome measurability, and different relationships between vendor risk and reward.

The Transition Playbook for SaaS Founders

Step 1 — Identify Your Measurable Outcome

The first question is: what result does your product deliver that your customer already measures? Not what your product does — what the customer’s business changes as a result. Intercom doesn’t ask “how many support interactions did our tool facilitate?” It asks “how many customer issues were resolved without a human?” That is a metric the customer already tracks in their support operations. Map your product to your customer’s existing success metrics — that is your outcome unit.

Step 2 — Build the Attribution Layer

You can only bill for outcomes you can prove. This requires instrumenting your product to track and report on the outcome unit in real time. If you charge per resolved ticket, your product must define “resolved” unambiguously, track it automatically, and surface it in a customer-facing dashboard. The technical and commercial infrastructure for outcome-based billing is non-trivial — billing platforms like Chargebee and Stripe are building metered billing primitives that support this, but the product instrumentation is your responsibility.

Step 3 — Start With a Hybrid Model

The fastest route to GaaS is not a full pricing overhaul on day one. It is a hybrid: a base platform fee that covers infrastructure, security, and integrations — plus an outcome-based variable component. This gives existing customers a familiar anchor (a predictable base fee) while introducing outcome alignment incrementally. It also protects your revenue floor while you calibrate your outcome unit and pricing. Most successful transitions to outcome-based pricing run through this hybrid phase for 12–18 months.

Step 4 — Redefine Your Growth Metrics

Once you move to GaaS pricing, your North Star metric is no longer net seat expansion. It is outcome volume per account — how many outcomes are being delivered, and at what rate that volume is growing. Monthly Outcomes Delivered replaces Monthly Active Users. Outcome volume per account replaces average seats per customer. Expansion becomes vertical (more outcomes within the same account) rather than horizontal (more seats across the organisation). For more on how AI is reshaping SaaS metrics broadly, see our guide on 10 ways AI is changing B2B SaaS forever.