.png)
Frameworks, core principles and top case studies for SaaS pricing, learnt and refined over 28+ years of SaaS-monetization experience.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
How SaaS Founders Should Choose Between Payg Fixed and Hybrid Pricing for AI Services
AI has turned pricing into a product decision. A SaaS founder can still charge per seat, but inference costs can rise with every complex prompt, agent run, or customer conversation. The old subscription model can therefore expose the vendor to heavy use without added revenue. A pure usage model solves that cost problem, yet can make buyers fear an invoice they cannot predict.
The harder issue is not whether usage should appear somewhere on the rate card. It is what the customer is truly buying. A developer using an AI coding assistant is buying faster personal output. A support leader buying an agent that resolves tickets is buying completed work. Those are different products, even when both rely on the same models.
Monetizely's position is clear: founders should use fixed pricing when a human remains the center of value, PAYG when an agent independently produces a discrete and verifiable outcome, and a fixed-led hybrid when the agent performs meaningful work but reliability, cost, or attribution has not yet earned an outcome price. Hybrid is not a compromise. It is a specific architecture with a named primary meter: predictable access or committed capacity first, variable usage second.
A named primary meter prevents hybrid pricing from becoming a muddle
A hybrid plan without a primary meter asks buyers to decode two pricing systems at once. Sales then describes one component as “the platform,” finance treats the other as “the real price,” and the customer waits for a surprise invoice. The company has not shared risk thoughtfully. It has hidden the decision.
Founders should instead decide which unit anchors the commercial relationship. Fixed pricing anchors the relationship to access, usually a seat or annual platform fee. PAYG anchors it to a measured result. A fixed-led hybrid anchors it to access or committed capacity, while variable charges protect margins and let revenue rise as adoption grows.
The table points to a simple rule: the variable component should never be the place where a founder tries to repair a weak primary meter.
Monetizely's 5-Step Pricing Framework puts the rate card near the end of the decision, not at the start. As developed in Monetizing Agentic AI, the framework begins with the business goal and the customer segment, then determines the package, the metric, the price point, and finally the operational system that makes the model work. That order matters because a price cannot rescue a package built for the wrong buyer.
For this question, Step 3 is decisive, but Steps 1 and 2 set its limits. A founder selling to individual developers, engineering teams, and regulated enterprises may use the same AI model across every plan. The packages should still differ because the buyers need different levels of administration, security, and purchasing support. Cursor’s public plans show that pattern: individual access begins at $20 per month, team access at $40 per user per month, and larger organizations receive pooled usage and enterprise controls. As of September 8, 2026, its pricing also includes model-usage allowances and on-demand charges after included usage is consumed.
The lesson is not that every AI product should copy Cursor. It is that packaging should separate customers by how they buy and govern AI, while the meter reflects how the product creates value.
Greater autonomy moves pricing from access toward completed work
The Agentic Monetization Spectrum (AMS) gives founders a disciplined way to make that call. It scores an agent on three dimensions: zero-human ability, or how little human work remains; operational domain, or whether the agent handles one task, one business function, or work across functions; and output/cost ratio, or whether value rises roughly with compute cost or far faster than it. As autonomy, scope, and value relative to cost increase, pricing should move away from the human seat and toward output or outcome.
The score below uses 1 for Small or Linear, 2 for Medium or Inflecting, and 3 for Large or Exponential. It is not a mathematical formula. It is a forcing device that makes founders state what their agent actually does today, rather than what a product demo suggests it may do someday.
The scoring makes the strategic dividing line visible. An AI copilot can be impressive without replacing the person who owns the result. A task agent begins to do work independently, but the customer may still need to review, retry, or repair that work. A resolution agent earns PAYG only when it produces a result that both parties can identify without argument.
The market already contains useful examples, provided founders study the design rather than imitate the headline price. Several leading vendors combine subscriptions and usage, but they do so for different reasons.
These examples support a stronger conclusion than “use a mix.” Cursor, GitHub Copilot, and Devin use variable charges to contain the cost of increasingly agentic work, yet each retains a fixed commercial anchor because a human user or team remains central. Intercom charges for defined outcomes because the customer can see whether a resolution, handoff, disqualification, or qualification occurred. Salesforce offers both paths because its product spans employee tools and customer-facing agents.
Many founders mistake model activity for customer value. Token counts, agent steps, API calls, and tool invocations are all easy to meter. None is automatically a good commercial unit. Buyers do not want to explain to a CFO why an AI service used 28 million tokens during a difficult month.
A PAYG unit must meet four tests before it becomes the primary meter:
- It is discrete. A customer can count it without reading model logs. “Resolved ticket” is clearer than “agent workflow.”
- It matters to the economic buyer. A support leader can connect a resolution to avoided workload, service quality, retention, or revenue.
- It has clear attribution. The company can distinguish an AI-completed result from a human-completed result or an abandoned attempt.
- It can be governed in advance. Buyers can set a budget, see usage, and understand what happens when they approach a limit.
Intercom’s published outcome rules demonstrate the standard. It charges no more than once per conversation and does not bill for failed attempts. That design limits disputes because the chargeable event is visible to both parties.
| Observable product condition | Recommended model | Primary meter | Variable element, if any |
|---|---|---|---|
| A user remains responsible for creating, approving, or sending the final work | Fixed | Seat or workspace | Published overage only for exceptional, high-cost usage |
| An agent executes tasks independently, but customers still inspect or rework a meaningful share | Fixed-led hybrid | Named user, team commitment, or platform capacity | Credits, compute-linked units, or task capacity |
| An agent completes a customer-recognized job with immediate, objective proof | PAYG | Resolution, qualified meeting, verified claim, or completed transaction | None required beyond volume discounts or commitments |
| An agent performs broad work but success cannot yet be attributed cleanly | Fixed-led hybrid | Platform access or committed capacity | A measurable intermediate unit, not a disputed “outcome” |
The final row deserves particular attention. Founders should not force outcome pricing because it sounds modern. When an AI sales agent sends outreach but a human closes the deal, “revenue influenced” is too contestable to be the invoice unit. A platform fee plus a measured, intermediate usage unit is more credible until the product can prove ownership of a result.
Billing clarity becomes part of the product promise
The choice among fixed, PAYG, and hybrid pricing is also a promise about control. A customer that cannot forecast spend will limit adoption. A founder that cannot see high-cost behavior will discover margin loss after renewal.
Every variable component therefore needs three operating controls:
- A buyer-visible definition of the chargeable event.
- Real-time or near-real-time visibility into usage and remaining budget.
- A documented path for caps, alerts, exceptions, and invoice disputes.
Operational discipline matters because agent pricing asks billing systems to handle more than annual seat counts. Monetizely’s framework places operationalization last because metering, rating, feature entitlements, and invoice explanation must support the strategy already chosen.
Founders should earn their way from access pricing to outcome revenue
Monetizely’s position remains firm: fixed pricing carries assistance, fixed-led hybrid pricing carries emerging delegation, and PAYG carries verified autonomous outcomes. That progression is a migration path. Founders who move toward variable revenue before their product can prove results will create sales friction; founders who remain fixed after their agents do real work will leave revenue and margin exposed.
Choose the business you intend to become over the next 24 months. A land-grab product for individual users should preserve simple entry pricing, while a margin-focused enterprise service can ask for commitments earlier.
Set one buyer-recognized primary meter for each package. If a sales representative needs several minutes to explain what customers pay for, the pricing model is not ready to scale.
Build packages around distinct buying environments, not model capability. Self-service users, managed teams, and regulated enterprises may use similar AI features, but they buy different levels of control, support, and commercial certainty.
Treat pure PAYG as a product milestone. Fund the instrumentation, workflow ownership, and success measurement required to prove an outcome before changing the primary meter.
Write the migration trigger before launch. Define the reliability, adoption, and attribution thresholds that will justify moving a task agent from fixed-led hybrid pricing to a true outcome model.
Footnotes
- https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
- Cursor, “Pricing” and “Models & Pricing,” accessed September 8, 2026. (cursor.com)
- Cognition, “Self-serve plans” and “Billing” in Devin documentation, accessed September 8, 2026. (docs.devin.ai)
- GitHub, “GitHub Copilot Plans & Pricing,” accessed September 8, 2026. (github.com)
- Salesforce, “Agentforce Pricing,” accessed September 8, 2026. (salesforce.com)
- Intercom, “Fin AI Agent outcomes,” July 30, 2026. (intercom.com)