Select a pricing structure for each phase of an AI service instead of forcing one model onto every engagement.
Decision framework with examples
The fastest decision rule
- Unknown scope: hourly or paid discovery.
- Clear deliverable: fixed project.
- Reserved recurring capacity: retainer.
- Variable measurable consumption: usage-based.
- Large, evidence-backed business impact: value-based component.
Six pricing models compared
Best when the task cannot be estimated reliably. Buyer sees time used, but total cost is uncertain. Protect with a cap and reporting cadence.
Best for defined inputs, outputs, revisions, and acceptance criteria. Seller carries overrun risk, so assumptions must be explicit.
Best when requirements, data, or integrations are unclear. The deliverable is a diagnosis, scope, plan, and estimate—not free presales work.
Best for recurring production, monitoring, or reserved capacity. Define monthly limits, response times, rollover, and out-of-scope work.
Best when cost or value tracks a measurable unit such as processed records or generated reports. Define the meter and minimum fee.
Best when impact is material and evidence is credible. Use only when attribution, scope, and responsibility are clear.
Pricing-model risk matrix
Hourly and paid discovery shift uncertainty into measured work.
Fixed projects provide the clearest upfront total when scope is stable.
Retainers become unprofitable when requests and response times are unlimited.
Usage and value pricing fail when the unit or attribution is disputed.
Somez AI Lab Example: one project, three models
Separate pricing-model choice from price level
The model answers how you charge; it does not determine the number by itself. Use the Somez AI Service Pricing Benchmarks to compare model-generated project quotes under published assumptions, then choose hourly, fixed, discovery, retainer, usage, or value pricing based on uncertainty and responsibility.
Build the quote in this order
- Estimate labor and tool costs.
- Add a contingency for scope and delivery risk.
- Choose the model that best matches uncertainty.
- Define client inputs, approvals, revisions, and delays.
- State how variable usage or extra work is billed.
- Compare the result with buyer value without inventing ROI.
Useful hybrid structures
- Discovery + fixed implementation: suitable for unclear projects that become defined after analysis.
- Fixed setup + retainer: suitable for ongoing monitoring and improvement.
- Base retainer + usage: suitable when minimum capacity is reserved but consumption varies.
- Fixed fee + success bonus: only when success can be measured fairly and external factors are addressed.
Pricing-model mistakes
- Using a fixed fee before understanding data and integrations.
- Calling unlimited support a retainer.
- Passing variable API costs through without a clear formula.
- Using value-based language with no credible business baseline.
- Charging hourly for a standardized package that already has reliable delivery data.
Frequently asked questions
Which model is best for beginners?
Fixed-scope pricing works well for a service you already understand. Use paid discovery or hourly billing when uncertainty is high.
Can models be combined?
Yes. Different phases often need different models.
When is a retainer appropriate?
When recurring work and reserved capacity are real, measurable, and limited. Use the Freelance Retainer Pricing Calculator to model reserved monthly hours, recurring costs, capacity buffer, and target margin.
Should tool costs be marked up?
You may include administration and risk in your price, but explain unusual variable charges clearly and avoid surprise fees.
All prices are illustrative. Pricing depends on scope, market, expertise, responsibility, taxes, and client requirements.
Written and reviewed using the standards in our Editorial Policy.