Tier Design

In traditional SaaS, tiers are often “features + seats.” Agent product tiers also have to package unstable runtime cost into customer-legible capacity and promises.

A good tier should answer more than “which features unlock?” It should answer:

  1. How much work can be completed?
  2. How hard can that work be?
  3. What happens on overage or high-risk tasks?

Four Tier Dimensions

DimensionMeaning
CapacityTask volume, token budget, runtime, concurrency, storage, or connector limits
CapabilityDefault model routing, context length, complex-task handling
PermissionsTools, connectors, execution environments, team management, audit
ServiceSLA, human support, compliance, dedicated deployment, or data region

Feature differences still matter, but if tiers do not control capacity and overage behavior, margin is exposed to a small number of heavy users.

Choosing Capacity Units

Do not expose tokens as the only customer-facing unit. Tokens are useful for internal controls, but not always for procurement.

Common combinations:

Capacity unitGood fit
Task countClear workflows with relatively stable task size
Successful task countClear success criteria and outcome-oriented buyers
Budget poolEnterprise customers sharing capacity across many task types
Runtime / sessionBrowser, sandbox, or hosted execution cost is meaningful
TokensEngineering/API customers who understand technical billing

Many products use a two-layer expression: customers see tasks or budget pools; the system accounts internally with tokens, model routing, and runtime.

Budget And Model Routing

Tiers can bind default models, or they can bind budget only.

DesignAdvantageRisk
Tier binds default modelCost control and simple explanationUsers feel artificially constrained
Tier binds budget; model selectableStrong user control, better for complex tasksMust show how different models consume budget
Automatic routingUsers do not need to understand modelsMust be explainable or customers distrust cost changes

A durable design is automatic routing by default, with advanced overrides and a clear explanation that stronger routes consume budget faster.

Capacity Ladder

Tier Capacity Ladder Each step should unlock more capacity, harder work, clearer permissions, or stronger service Free try core workflow more routine capacity Lite stable daily work more tools and teams Pro high-intensity team use custom capacity and service Ultra contracted scale Use real usage data to set limits; the visual shows separation, not recommended quotas.

Tier steps need to feel meaningfully different, or upgrading feels pointless. But if the gap is too large, customers get stuck between tiers. Instead of chasing a fixed multiplier, watch three signals:

  • Can customers easily see what more work the next tier enables?
  • Does the current tier often approach limits without converting to overage?
  • Does the next tier feel like a forced large upgrade for a small increment?

Capacity usually widens with customer maturity: entry tiers prove value, middle tiers cover stable workflows, and higher tiers support harder tasks, more teams, and stronger service commitments.

Overage Behavior

Overage rules need to be visible before they happen. Common strategies:

StrategyGood fitRisk
Hard stopFree, trial, and budget-sensitive customersInterrupted work hurts experience
Soft degradeRoutine tasks tolerate lower capabilityUsers may notice quality decline
Ask for confirmationHigh-value tasks and enterprise usersAdds friction
Automatic overage billingSigned hybrid or enterprise contractsRequires very transparent billing
Upgrade suggestionUsers repeatedly approach limitsCan feel pushy if poorly timed

The worst design is no warning, continued execution, and an end-of-month bill the customer cannot explain.

Example Structure

The following is synthetic, not recommended pricing:

TierMain promiseTypical limitOverage behavior
FreeTry the core workflowSmall task volume, low concurrency, limited toolsHard stop
TeamStable daily workMedium budget, standard routing, shared team capacityConfirmation or soft degrade
BusinessHigh-intensity team useHigher budget, more connectors, audit and adminExplicit overage billing
EnterpriseContracted capacity and serviceCustom budget, permissions, compliance, and SLAContract-defined

The point is not the names. Each tier explains capacity, capability, permissions, and overage together.

Naming And Presentation

Tier names should help customers identify themselves, not make them decode your cost structure.

Good presentation:

  • Explain capacity with typical monthly work.
  • Explain capability with task complexity.
  • Explain enterprise differences with team, audit, permissions, and support.
  • Explain overage with progress bars, alerts, and bill forecasts.

Avoid:

  • Listing model names directly as the pricing page’s main structure.
  • Making tokens the only external quota.
  • Hiding overage rules in terms of service.
  • Showing Enterprise as only “contact us” without saying who it is for.

Cross-section Connections

Was this page helpful?