Written by Nishant Patel, BGV Associate
As Generative AI becomes foundational to enterprise software, companies must rethink their pricing strategies. The traditional SaaS playbook—centered on per-seat licenses—no longer fits the economic realities or usage patterns of AI-driven solutions. This internal memo outlines key considerations and emerging best practices for pricing agentic AI, based on field insights and current industry experimentation.
Why Per-Seat Pricing Breaks for AI
SaaS businesses have historically priced per user. This worked because the cost to serve each user was roughly flat, and software usage tracked closely with the number of seats. With AI, that assumption no longer holds. The marginal cost of serving users varies dramatically due to:
- Token-based costs for LLM usage
- Unpredictable variance in usage per user
- Agentic automation reducing seat-based headcount
One team member may query an AI 1K times a day while another barely uses it. Under per-seat pricing, both cost the same. Worse, if your AI drives productivity so efficiently that fewer users are needed, your pricing drops as your value increases—an unsustainable inverse relationship.
Current Alternatives (and Their Limitations)
- Per Token / Cost-Based Pricing
- Transparent and accurate for vendors
- But burdensome for buyers who can’t predict costs
- Causes cognitive friction: “Can I afford to ask this question?”
- Task-Based Pricing
- Viable when outcomes are consistent (e.g. form filling)
- Breaks when tasks vary in complexity or success rate
- Creates customer dissatisfaction when quality is uneven
The Rise of Value-Based Pricing
The most promising direction is outcome-aligned pricing—charging only when customers receive measurable value. This aligns incentives, builds trust, and avoids punishing your best users. Examples include:
- Fin (Customer Support Chatbot): Charges only for resolved support tickets. Customers define success. If Fin fails to resolve the issue, there’s no charge.
- Evinced (BGV Portfolio Company): Uses per-day pricing for dev use. Customers pay only on days the tool is activated for accessibility testing, aligning spend with dev workflow.
- Chargeflow (Chargeback Recovery): Takes a percentage of recovered funds. If no money is recovered, the service is free.
Building a Value-Based Pricing Model: Key Questions
- Are You Clearly Saving Costs?
- Can you isolate incremental savings over existing solutions?
- If yes, consider pricing as a % of savings delivered.
- Are You Driving Measurable Revenue?
- If so, price like a commission model—e.g. % of qualified leads or closed deals.
- Can You Anchor to an Industry KPI?
- If your tool moves a metric customers already care about, price on the delta.
- Caveat: Make sure you control enough of the process to be accountable.
- Is Usage Predictable?
- If behavior and cost are stable, usage- or token-based pricing might work.
- Consider fixed tier bundles or hybrid plans to manage variability.
- None of the Above?
- Get creative. Talk to customers. Explore hybrid or novel models.
- Build in feedback loops to refine over time.
Hybrid Approaches and Customer Segmentation
- Enterprise Customers: May demand outcome-based or usage-linked contracts. Often need proof via pilots before committing.
- Startups: Prefer simplicity. Offer tiered pricing or free usage to accelerate adoption.
- SMBs: Crave predictability. Per-seat works if you limit AI usage or price as an add-on.
In hybrid SaaS/AI tools, keep the SaaS component on a per-seat model and price AI capabilities separately. This ensures revenue doesn’t drop as seats decline due to AI efficiencies.
Free Trials and Abuse Prevention
AI tools are expensive to run and easy to exploit. Avoid open-ended free trials. Instead:
- Offer usage-capped free trials (e.g. tokens, days, tasks)
- Provide enough runway to demonstrate value without incurring runaway costs
Outstanding Industry Questions
- How do users respond to value-based models over time? Do perceptions shift?
- What happens if foundational model pricing changes? Are vendors exposed?
- Will AI pricing converge around a dominant model, or remain fragmented by use case?
There is no one-size-fits-all solution. Pricing must evolve alongside your product, your customer’s maturity, and the underlying technology stack. What matters most is aligning cost with value—for both sides. Done right, pricing becomes not just a revenue lever, but a trust signal and product strategy advantage.
Questions? Feedback? We’re here to support you.
