Should we charge per user, per case or per completed AI task?

Understanding Pricing Units in SaaS: Per User, Per Case, or Per AI Task

Choosing the right pricing unit is a critical decision for SaaS founders and product owners. It affects customer perception, revenue predictability, and operational complexity. This article compares three common pricing units: per user, per case, and per completed AI task. We focus on aligning pricing with customer value, ensuring predictable billing, and accurately measuring usage, including handling failed tasks and staff review.

1. Pricing Per User: Simplicity and Predictability

Charging per user (also called per seat) is the most straightforward model. Customers pay a fixed fee for each user who can access the software. This model works well when:

• Value is tied to the number of people using the product.
• Usage is relatively consistent per user.
• Customers prefer predictable monthly or annual bills.

Advantages include easy billing, straightforward sales conversations, and predictable revenue. However, it may not capture value if some users use the product heavily and others lightly.

For South African businesses, integrating per user pricing with payment platforms like Payfast can leverage their subscription billing features for automated recurring payments Payfast subscriptions.

2. Pricing Per Case: Aligning with Customer Outcomes

Per case pricing charges customers based on discrete units of work or cases processed. This suits products where the customer values each completed case, such as legal matter management or insurance claims.

Benefits:

• Aligns price with customer value delivered.
• Customers pay only for what they use.

Challenges:

• Usage can be unpredictable, leading to variable monthly bills.
• Requires clear definitions of what constitutes a case.
• Handling failed or incomplete cases needs policy clarity.

Product owners should carefully define "case" and consider including staff review time or failed attempts in pricing to reflect true usage.

3. Pricing Per Completed AI Task: Precision and Complexity

When SaaS products incorporate AI tasks (e.g., document analysis, data extraction), pricing per completed AI task can directly link cost to usage.

Pros:

• Granular measurement of usage.
• Customers pay for the defined completed product outcome. Provider token consumption and retries are internal cost drivers, not automatically additional billable tasks.

Cons:

• Tracking failed tasks or retries complicates billing.
• Staff review time may be significant but hard to meter.
• Requires robust usage metering and integration with billing systems like Stripe's usage-based billing Stripe usage-based billing.

Founders must decide whether to bill only successful tasks or include failed and reviewed tasks, balancing fairness and revenue.

4. Measuring Usage Accurately: Including Failed Tasks and Staff Review

Ignoring failed AI tasks or staff review time can underprice the service and hurt margins. Consider:

• Logging all AI tasks, successful or failed.
• Defining policies on whether failed tasks are billable.
• Estimating average staff review time per task and including it as a cost factor.

For a completed-task offer, the proposed rule here counts a task only when its defined accepted outcome exists. Model failed attempts and review time internally when choosing a sustainable published price. A separately chargeable review service needs its own definition and explicit agreement.

5. Predictable Bills Versus Usage Variability

Customers value predictable bills for budgeting. Per user pricing offers this but may not reflect actual usage.

Per case and per AI task pricing can cause bill fluctuations. Mitigate this by:

• Offering usage caps or tiers.
• Providing monthly usage reports.
• Allowing prepayment or credits.

6. Aligning Pricing Units with Customer Value

Choose the pricing unit that best matches how customers perceive value:

• If value is tied to team access, choose per user.
• If value is per outcome or project, choose per case.
• If value is per automated AI action, choose per AI task.

Consider customer journey and usage patterns User journey to understand touchpoints and value drivers.

7. Integrating Pricing Models with Payment Platforms

Check merchant-entity and product eligibility before comparing implementations. Stripe lists South Africa through an extended network linked to Paystack; that does not establish access to Stripe Billing, Metronome or Connect through a Paystack account. Each provider has its own contract:

• Payfast supports subscription billing with flexible schedules Payfast subscriptions.
• Stripe offers advanced usage-based billing and real-time metering Stripe usage-based billing.

Ensure your pricing unit is compatible with your payment platform's capabilities.

8. Practical Pricing-Unit Decision Worksheet

Use the following worksheet to evaluate your pricing unit choice. Score each factor 1 (low) to 5 (high) for each pricing unit option.

How to use:

1. Score each factor honestly for your product and customers.
2. Sum scores per pricing unit.
3. Discuss totals only after mandatory customer, measurement and provider constraints pass. A high score cannot make an unavailable product or undefined unit workable.

Hypothetical Worked Example

All scores are fictional discussion values. Higher operational-simplicity scores mean less complexity; they are not researched demand or measured customer outcomes.

A SaaS startup offers AI-powered document processing with staff review. They score:

They select per user pricing for predictability and simplicity, but plan to add usage tiers.

9. Handling Failed AI Tasks and Staff Review in Pricing Models

A key challenge in pricing per AI task is deciding how to handle failed tasks and staff review time, which are often invisible to customers but critical to operational costs.

Defining Failed Tasks and Their Impact

Failed AI tasks occur when the automated process does not complete successfully, requiring retries or manual intervention. Ignoring these in billing can lead to under-recovery of compute costs and staff time.

Proposed Policy Example:

• Exclude failed provider attempts from the completed-unit count; include expected failure/retry costs in the internal price model.
• State whether review is included or is a separately agreed service; do not add an undisclosed fee after processing.

This is a proposed rule, not a provider requirement or measured fairness result. Validate customer understanding and unit economics before adopting it.

Measuring Staff Review Time and Costs

Estimate average staff review time per task or batch. For example, if staff spend 10 minutes reviewing 20 tasks, allocate a fixed review fee per 20 tasks processed.

Recorded Fields:

• Accepted completed tasks, separate from failed provider attempts.
• Number of failed AI tasks.
• Staff review time logged per batch.

Expected Results:

• Accurate cost reflection in pricing.
• Transparent billing statements explaining fees.

Recovery Steps:

• Monitor failure rates monthly; if failures increase, investigate AI model improvements.
• Adjust review fee quarterly based on actual staff time.

10. Integrating Usage-Based Pricing with Payment Platforms in South Africa

South African SaaS businesses must consider payment platform capabilities when choosing pricing units.

Payfast Integration

Payfast’s cited feature page supports recurring billing and management; it does not establish a native usage-metering contract. Verify additional products instead of claiming they are unavailable. For per user pricing, Payfast subscriptions work well with fixed recurring charges Payfast subscriptions.

For usage-based models:

• Implement usage metering in your SaaS backend.
• Generate monthly invoices externally.
• Verify the supported collection flow and merchant authority for each invoice; the cited subscription page does not establish a one-off usage-charge API contract.

Verify Provider and Product Eligibility

Stripe usage-based billing describes product capabilities for an eligible integration. Stripe’s country page lists South Africa in an extended network linked to Paystack. That is not proof that Paystack exposes Stripe Billing APIs or that a South African entity can use every Stripe product.

For a verified entity/product combination, review how it accepts your chosen unit and generates invoices. Stripe usage-event documentation supports mechanics for its applicable metering integration, not the decision about which customer outcome to charge. Verify identifiers, corrections and period handling against the product actually selected.

Where Stripe webhooks apply, its webhook guidance covers signatures, duplicate events and delivery order. Quick acknowledgement alone does not prevent duplicates. No cited source proves one provider is cheaper or supports the claim that Stripe through Paystack offers better metering.

11. Worked Hypothetical Case: Choosing the Pricing Unit for "DocuAI"

Background:

DocuAI is a South African SaaS startup offering AI-powered document processing with manual staff review. Customers range from small legal firms to insurance agencies.

Step 1: Define Customer Value Drivers

• Customers value processed documents (cases) completed accurately.
• Staff review improves quality but adds cost.
• Customers want predictable bills but accept some variability.

Step 2: Score Pricing Units Using the Worksheet

Step 3: Decision

Per case pricing scores highest, balancing value alignment and handling of failed tasks and staff review.

Step 4: Define Pricing Policy

• Charge a fixed fee per completed case.
• Exclude failed provider attempts from the completed-case count and model their costs internally.
• Add a fixed monthly fee for staff review proportional to expected volume.
• Offer usage caps with overage fees to control bill variability.

Step 5: Acceptance and Failure Tests

Step 6: Implementation Notes

• Integrate usage logging for cases and AI tasks.
• Automate billing calculations monthly.
• Use Payfast for base subscription and manual invoicing for usage fees.

This structured approach ensures pricing aligns with customer value, recovers costs fairly, and maintains predictable revenue streams.

---

Documentation Boundaries for This Decision

The pricing-unit worksheet is a proposed aid. Payfast’s feature page supports recurring schedules and subscription management, not a negative assertion that every usage-metering product is unavailable. Stripe documents usage-based products, but entity and product eligibility must be established independently before adopting an implementation.

Keep provider consumption, review cost and the customer's completed billable unit separate. Model failures internally without turning a completed-task promise into charges for every retry. No cited provider recommends charging 50% for failed tasks or proves such a policy fair.

Define whether one case contains several AI calls, corrections and staff review. Show a worked bill and ask the customer to explain the unit back. Test two calls contributing to one accepted case, a timeout retry, a rejected output and a later approved correction. These expose ambiguous definitions before billing automation makes them costly to change.

Frequently asked questions

What if my AI tasks often fail? Should I bill for failed tasks?

For the proposed completed-task offer, count the accepted outcome, not a failed provider attempt. Track failures internally and investigate excessive failures. A separately charged diagnostic or review service requires its own explicit agreement; provider consumption alone is not billing authority.

How do I handle staff review time in pricing?

Estimate average review time per unit and include it as a fixed fee or part of per case pricing to reflect true costs.

Can I combine pricing units?

Yes. Hybrid models (e.g., base per user fee plus per AI task charges) can balance predictability and value alignment.

How do I integrate usage-based pricing with Payfast or Stripe?

Verify entity eligibility, the exact product and collection contract. A Paystack extended-network listing does not establish Stripe Billing access, and a subscriptions feature page does not prove native usage-metering capability. See Stripe country and extended-network availability.

Conclusion

Choosing between per user, per case, or per completed AI task pricing requires balancing customer value alignment, billing predictability, and operational feasibility. Use the pricing-unit decision worksheet to evaluate your options objectively. If your business needs help designing or implementing pricing models, get in touch with Symaxx for expert SaaS development support SaaS development pricing models.

For wider implementation planning, use SaaS development alongside pricing models.

Related guides

For further background, read our website maintenance costs and CMS vs custom development. Understanding the user journey helps align pricing with customer value.

Sources

• Stripe usage-based billing
• Payfast subscriptions
• Stripe country and extended-network availability
• Stripe webhook guidance
• Stripe usage event reporting