For nearly two decades, software companies have ridden a familiar playbook: term licenses, then subscriptions, then SaaS and “born-in-the-cloud” models. These approaches delivered predictable revenue and eye-catching growth—20%, 30%, even 40% annually. That era is over.
Many mature SaaS businesses now resemble “utilities and pipelines”: durable and necessary, but no longer the primary engines of growth. Net retention rates that once routinely hit 125–130% have compressed toward 110%. CFOs are pushing back on sprawling vendor portfolios and seat-based pricing that no longer match how work is actually done.
At the same time, AI—especially large language models (LLMs) exposed directly to consumers—has reset expectations about how software should be used and paid for. Buyers now think less in terms of licenses and more in terms of tasks, outcomes, and measurable value.
The implication for software and AI-native companies is clear: pricing models must evolve. But there is no universal answer or “silver bullet.” Leaders need a portfolio mindset, grounded in data, experimentation, and operational discipline.
Most software companies still rely heavily on traditional models: annual subscriptions with maintenance, or per-seat pricing. These approaches are:
However, they share a critical weakness: they are poorly aligned to actual usage and value delivered. When software is triggered by systems, runs autonomously in the background, or varies dramatically by workflow, “per seat” quickly loses coherence.
In response, three families of alternative models are emerging.
1. Task-based pricing
Task-based models charge per discrete action or interaction—for example, a chatbot priced at $2 per customer interaction. These models can:
The trade-off is vendor complexity: cash flows become more volatile, billing and forecasting harder, and infrastructure costs more sensitive to usage spikes.
2. Credit-based or “infrastructure token” models
Here, customers pre-purchase a pool of credits—similar to game tokens—and burn them down over time, often across multiple features or workloads. This structure can:
Think of how many AI and infrastructure platforms now sell usage: customers buy capacity, then allocate it dynamically across use cases.
3. Hybrid seat + usage models
The most common pattern today is hybrid: a base subscription (often per seat) plus usage-based tiers or credits layered on top—Microsoft Copilot being a widely understood example. Customers pay a predictable per-user fee and then incur additional variable charges for advanced features or custom agents.
Hybrid models aim to blend:
As AI becomes embedded in core workflows, pricing innovation is accelerating further. Two patterns are gaining prominence: agentic pricing and outcome-based pricing.
Agentic “seat” models
In agentic models, what is priced is not a human user, but an AI agent that completes workflows autonomously. This can work well when:
However, these models introduce governance challenges. When many end users rely on many agents, costs can spiral quickly if usage is not carefully monitored, capped, and optimized.
Outcome-based pricing
Outcome-based models tie revenue directly to measurable success. For example, a customer support platform might be paid per successfully resolved issue, not per seat or interaction. This is already visible in vendors that charge when a chatbot fully resolves a customer’s problem, rather than just engaging with them.
Outcome-based pricing promises tight value alignment—but it is hard to execute. It requires leaders to grapple with:
For AI-native companies, building around data from day one makes this more feasible. For established SaaS players, retrofitting outcome-based pricing onto legacy systems and processes can be far more complex.
Several powerful forces are pushing software leaders to rethink pricing.
1. New variable cost structures
AI-infused software often carries highly variable infrastructure costs. A single misconfigured query, recursive join, or poorly governed model can produce an eight-figure cloud bill in weeks. Such dynamics simply did not exist in the world of fixed hosting and linear scaling.
This volatility affects:
2. Changing buyer behavior
CFOs no longer accept paying for unused seats or maintaining dozens of overlapping software vendors. They want:
This is prompting some companies to deliberately de-emphasize pure usage in favor of more stable subscription revenue—sometimes even changing sales compensation so that a dollar of subscription counts more than a dollar of usage toward quota.
3. The new reality of net retention
Best-in-class net retention has reset materially. A 110% net retention rate is now, in many segments, the new 130%. Behind that shift:
In this environment, pricing strategy is no longer a marginal lever. It is central to defending the base, enabling expansion, and sustaining investor confidence.
Evolving pricing is not just a commercial decision; it is an operational transformation that touches product, finance, sales, billing, and investor relations.
Design pricing into new product introduction
When launching new products or AI features, leaders should explicitly design for:
Buyers may be increasingly sophisticated, but they still want pricing that is easy to understand, compare, and explain internally.
Align sales incentives with the pricing strategy
Sales compensation can either reinforce or undermine your chosen pricing model. Companies must decide how they will reward:
As net retention moderates and CFOs push for predictability, many organizations are reweighting compensation toward more stable revenue streams, even if that means moderating short-term usage growth.
The common thread across all advanced pricing models is data. To price on tasks, credits, or outcomes, companies must reliably capture and connect product usage, customer contracts, billing, and financial reporting.
That is elevating a once-niche role: the data steward for revenue and pricing.
In leading firms, this role typically:
Crucially, the most effective stewards are not just data experts; they are business translators. They can explain what the numbers mean, how pricing changes will affect them, and how investors will interpret them.
For fast-growing AI-native companies, hiring this role early—sometimes as early as $10 million in ARR—can prevent inconsistent deal-making, ad hoc discounting, and opaque revenue patterns that become costly to untangle later. For large incumbents, the challenge is often carving out budget and clarity of ownership in organizations where data responsibility is diffuse.
Across stages, the imperative is the same: do not underinvest in the data foundation. The cost of fixing it under the scrutiny of late-stage investors or public markets is far higher than getting it right early.