Vijay Raina has spent his career at the sharp edge of enterprise SaaS, witnessing the industry’s transition from clunky on-premise installations to the streamlined, seat-based subscriptions that defined the last decade. As a specialist in software design and architecture, he has helped dozens of organizations navigate the complexities of digital transformation. Today, he finds himself at the center of a new, more disruptive shift: the erosion of the per-user revenue model by autonomous AI agents. His insights bridge the gap between the technical reality of high-compute infrastructure and the economic necessity of sustainable business scaling, offering a roadmap for companies whose very success now threatens their traditional bottom line.
The discussion centers on the growing “pricing paradox” where AI tools that deliver exceptional value—such as resolving 80% of customer support tickets without human intervention—simultaneously reduce the need for the user seats that generate revenue. This dialogue explores the shift toward consumption and outcome-based billing, the hidden risks of rewarding system failures in usage-based models, and the necessity of creating “verified outcomes” to maintain trust between vendors and buyers. It also touches on the infrastructure costs that make AI SaaS fundamentally different from its predecessors and the strategic shift toward hybrid models that combine flat platform fees with metered performance metrics.
Traditional per-seat pricing has long been the gold standard for SaaS, assuming that human headcount is a direct proxy for software value. How is the rise of autonomous AI agents fundamentally dismantling this logic?
The per-seat model was built on a very simple, comfortable assumption: if a company is growing its headcount, it is getting more value from its software, and therefore it should pay more. We saw this in every department; a larger support team meant more helpdesk licenses, and more salespeople meant a bigger CRM bill. But when an AI agent can step in and handle the workloads once shared across whole teams, that link between headcount and value snaps. We are seeing a shift where single operators now oversee an array of agents, effectively replacing three or four coordinator roles with a single license and an automated workflow. The customer is actually getting more value—faster resolutions and higher accuracy—but because they need fewer humans logged in, the vendor’s revenue takes a hit. It is a strange paradox where the better the product works, the less the customer feels compelled to pay under the old rules.
When a product works so well that it actively cuts customer headcount, the vendor faces a revenue problem they essentially created. How should companies address this “pricing paradox” without punishing their own innovation?
This is the most significant challenge facing SaaS founders today because if you stay tied to seat-based billing, you are effectively incentivized to make your product less efficient. Leading companies like Intercom and Salesforce are already moving toward conversation-based or credit-based pricing for their agents to circumvent this. McKinsey’s research highlights that about 40% of AI-native companies have already pivoted to activity or consumption metrics to ensure their revenue reflects the work the software is actually doing. The key is to stop charging for the “seat” and start charging for the “scarce value” the product creates, whether that is a resolved case or a processed transaction. If you don’t make this jump, you’ll find your revenue shrinking even as your impact on the client’s business reaches an all-time high.
Some experts argue that usage-based pricing has a “dark side” because it can inadvertently reward failure—charging more for retries or dead tool calls. How can vendors avoid creating an environment where the customer funds the system’s error rate?
This is a very real danger that doesn’t get enough attention in the excitement over consumption models. In a usage-based setup, every time an agent fails and has to retry a task, it burns compute and tokens, which then shows up on the customer’s invoice as ordinary consumption. This creates a “proof debt,” where the buyer is essentially paying to find out which parts of the AI’s work are actually finished and which are just hallucinations or errors. To fix this, we have to move toward a model where failed attempts belong on the vendor’s ledger, not the customer’s. The only unit worth selling in a mature AI economy is a verified outcome—a result that both sides can trust as complete and accurate. If you charge for raw throughput without verification, you are just rebranding usage and marking it up, which will eventually erode customer trust.
The shift toward “outcome-based” pricing seems like the most honest alignment with customer value, but it brings up difficult questions about attribution. How do you implement a “verifier” that both the vendor and the buyer can trust?
Outcome-based pricing is the “holy grail” because it ties the bill directly to measurable business results, like a recovered payment or a closed support ticket. However, the hard part is defining what constitutes a “win” in a way that can’t be contested by either side’s finance team. In sectors like construction or finance, this might be an approved change order or a cost report that reconciles perfectly across systems. We have to move toward a “working rule” where if an outcome cannot be independently verified by the system, it shouldn’t be billed as a completed task. Without a reliable verifier, outcome pricing becomes a series of claims rather than a transparent transaction, which is why many are looking at hybrid models to bridge the gap.
Unlike traditional SaaS with near-zero marginal costs, AI agents run on infrastructure that burns significant compute and model tokens. How does this “cost of work” change the way companies must think about their pricing floors?
In the old world of SaaS, adding a user cost the vendor almost nothing, but an AI agent running for four hours on a complex infrastructure task carries a very real, tangible cost. Every scan or workflow run consumes compute power and expensive tokens, meaning that usage-based billing isn’t just a strategy—it’s an honest reflection of the vendor’s expenses. If a vendor doesn’t set pricing floors, they might find themselves undercharging high-usage customers while failing to build the margins necessary to cover these orchestration expenses. This is especially true as inference prices continue to drop; while it makes the technology more accessible, it also introduces a deflationary pressure on revenue. Vendors must be careful to ensure that their efficiency gains don’t reduce their revenue faster than they can expand into new, more complex workflows.
Predictability is often cited as the biggest hurdle for consumption-based models, as finance teams loathe “lumpy” bills. Is the hybrid model of a flat platform fee plus metered tasks the only way to satisfy both the vendor and the buyer?
The hybrid setup is currently the most sustainable path forward because it offers the best of both worlds: predictability and scalability. A flat platform fee covers the essentials—security, integrations, management, and governance—which gives the customer’s finance team a baseline they can count on every month. On top of that, you layer the metered rates for successful agent tasks or specific outcomes, which allows the vendor to capture the upside of the value they provide. We are seeing this pattern emerge at major players like Salesforce, where they provide a layered model rather than a clean replacement for the subscription. It protects the vendor’s cash flow while ensuring the customer only pays extra when the AI is actually delivering measurable work.
What is your forecast for the survival of the per-seat model in the enterprise sector over the next two years?
The per-seat model won’t disappear entirely by 2028, but it will be relegated to “dumb” tools that don’t utilize autonomous agency. For any software that claims to perform work rather than just provide a place for humans to work, the seat-based approach will become a relic of the past. We will see a massive shift where the “seat” is replaced by the “agent,” and the primary metric for success will be the volume of verified work completed rather than the number of employees with a login. Companies that fail to adapt their billing to this reality will find themselves replaced by AI-native competitors who are more than happy to charge for outcomes. The winners will be those who can provide transparent measurement, showing exactly what the agent replaced and what human review remains necessary, turning the “pricing paradox” into a competitive advantage.
