MyDot Consultancy

Featured image: AI-generated conceptual illustration. Sources checked 26 October 2026.

Agent pricing discussions often jump straight to “charge for outcomes.” That can be a good model when the outcome is observable and the provider meaningfully controls it. It becomes contentious when success depends on the customer's staff, an external vendor, or a long chain of events the agent only partially influences.

A subscription, per-task fee, seat price, and hybrid model each allocate risk differently. The right choice should reflect the service being sold rather than the latest pricing fashion. This article provides a decision framework using proposed models, not a universal recommendation.

Define value before choosing a meter

Ask what the buyer wants to accomplish and what evidence shows it happened. A prepared document, a verified account update, and a won sale are different value events. The farther the event lies from the agent's direct action, the harder attribution usually becomes.

Also ask when value becomes visible. A support answer may be checked quickly, while the effect of a renewal-preparation workflow may take months. Delayed outcomes can make invoicing and disputes difficult, especially for a small service business with limited cash reserves.

Do not confuse customer value with activity. Tokens, messages, and tool calls are useful cost signals, but buyers may not consider them meaningful results. If you charge for them, explain why and give customers enough visibility to budget.

When a subscription fits

A subscription can work when the service provides ongoing availability, maintained integrations, and a predictable bundle of work. Customers receive a stable budget, while the provider takes some usage risk. Define allowances and exceptions so stability does not become an unlimited obligation.

Subscriptions can also suit services whose value is hard to attribute to individual events, such as knowledge maintenance or operational monitoring. The customer is paying for a maintained capability and accountable service rather than a count of isolated successes.

The weakness is perceived value during low-usage periods. Address this with clear deliverables, reporting, and an appropriate cancellation policy. Do not invent unnecessary activity merely to make the subscription look busy.

When per-task pricing fits

Per-task pricing works best when tasks have a clear boundary and reasonably consistent complexity. An approved document extraction or a defined evaluation run may fit. Specify how duplicates, retries, cancellations, and incomplete inputs are treated.

A task fee does not have to imply guaranteed success. The commercial definition can distinguish eligible processing from a verified outcome, but the buyer must understand the difference. Avoid calling every attempted task a completed result.

Watch for complexity variation. Two documents can differ enormously in length, quality, and review needs. Use transparent bands or exclusions rather than silently subsidizing the hardest work through an average that may not hold.

When outcome pricing fits

Outcome pricing aligns incentives when the provider can observe and influence a meaningful result. It can be compelling for buyers reluctant to pay for activity. However, the definition must be precise enough to support an auditable invoice.

The Intercom pricing page illustrates why definitions matter: its observed outcome terminology is broader than a simple count of explicitly confirmed resolutions. Buyers should read the current conditions rather than assume that all vendors use the same meaning.

For a new service, define the reopen window, disputed outcomes, and human contribution. If the agent prepares a draft that a person substantially rewrites, decide whether that is the outcome being sold. If the customer ignores a valid recommendation, decide whether the provider should bear that risk.

When seat pricing fits

Seat pricing can be understandable when the product supports individual workers and usage roughly follows active users. It is easy to budget but may discourage broad adoption or shared operational use. A background agent that performs work across a team may not map naturally to seats.

Define what counts as a seat and how temporary users, reviewers, and administrators are handled. Charging every occasional approver can create friction that reduces the product's usefulness. Conversely, a single shared login can undermine access control and accountability.

Consider whether the seat is a value proxy or merely a convenient billing unit. If usage varies dramatically by role, a hybrid allowance may better reflect economics without making the invoice unnecessarily complicated.

Test a hybrid against real workloads

A base fee plus usage can cover fixed readiness costs while keeping growth proportional. A minimum commitment can support reserved capacity, but it should reflect genuine obligations. Explain what the customer receives even when usage is low.

Use historical or pilot data to simulate several invoice structures. Compare typical months, peak months, and unusual exceptions. Usage-based billing tools can implement the chosen model, but they cannot resolve an ambiguous definition of value.

Ask customers to review sample invoices before launch. Confusion at this stage is useful feedback. If a buyer cannot predict the bill from ordinary operating data, simplify the model or improve the explanation.

Choose the dispute you are willing to manage

Every pricing model creates a different conversation. Subscriptions raise questions about ongoing value. Task fees raise questions about scope. Outcome fees raise questions about attribution. Seat prices raise questions about who needs access.

Select a model whose evidence you can produce and whose incentives support good service. Review it as the product matures. A transparent, sustainable hybrid is often better than a fashionable outcome promise that neither party can reliably verify.

Sources and research note

Intercom pricing. Stripe usage-based billing. Research checked October 26, 2026. Product behavior and prices can change. Business examples and proposed prices in this article are illustrative unless explicitly attributed.