Per-Interaction Pricing: Examples & Companies

5 companies in the corpus Updated stub analysis
Definition

Per-Interaction Pricing is a billing unit where each patient-agent or user-agent interaction is metered, common in healthcare AI and customer engagement platforms.

Also known as: Interaction-Based BillingPer-Patient-Interaction Pricing

What is it

Per-interaction pricing is a billing unit where each patient-agent or user-agent interaction is metered, common in healthcare AI and customer engagement platforms.

The term “interaction” sits between the granularity of a single message and the finality of a resolved ticket. An interaction is a bounded engagement — one patient phone call, one support session, one workflow run — that begins and ends regardless of whether the AI produced the outcome the buyer hoped for. That makes it intuitive to operational buyers (healthcare administrators, contact-center managers) who already count interactions in their SLAs, and it sidesteps the definitional disputes of outcome-based billing.

Hippocratic AI uses the interaction as the basis for a per-agent-hour staffing model: health systems pay roughly $9 per hour of active patient-interaction time — benchmarked against the ~$39/hour median wage of a registered nurse. Uniphore layers a per-interaction consumption charge on top of a platform license and per-seat fee, so the meter captures volume growth independent of headcount. Yellow.ai bundles interactions and sessions on its Free tier (1 AI agent, 500 chat sessions/month) before customers commit to its outcome-based $0.99-per-resolution overage.

That flexibility is also the definitional challenge. “Interaction” is deliberately broad — a real-time voice call of any duration, a multi-turn chat session, or a structured workflow run — and vendors draw the boundary differently. Confirm exactly what starts and ends an interaction in any contract before modelling spend.

What counts as "one interaction" — four boundaries
"Interaction" — billed per engagement, or only on success AGENT-HOUR ~$9 Hippocratic AI · /hr vs ~$39/hr RN wage active patient time USAGE LAYER + per-int Uniphore · on license + per-seat fee volume ≠ headcount SCOPED VOLUME chat · voice Maven AGI · contract conversation + resolution voice carries premium PER RESOLUTION $0.99 Yellow.ai · /close after 500 free sessions only on success ← BILLED EVERY ENGAGEMENT BILLED ONLY ON SUCCESS →

How it works

The interaction meter works by detecting a defined trigger event (a patient calling in, a customer opening a chat, an employee initiating a workflow) and counting that as one unit. The meter closes on a corresponding end signal: the call terminates, the chat session times out, or the workflow reaches a terminal state. Everything that happens in between — messages exchanged, tokens consumed, tools invoked — is absorbed into the flat per-interaction charge.

DimensionWhat it controlsExample
Interaction definitionWhat counts as “one” interaction — a call, a session, a workflow runHippocratic AI: one agent-patient call of any duration
Duration handlingWhether time within the interaction affects the chargeHippocratic AI bills per agent-hour, so longer calls cost more within the same interaction frame
Outcome linkageWhether the interaction must succeed to be billableYellow.ai charges per resolution (a successful close), not per raw interaction
Volume tiersWhether the per-interaction rate is negotiated at scaleUniphore negotiates per-interaction consumption inside custom enterprise contracts
ChannelsWhether voice, chat, email, and SMS interactions carry different ratesMaven AGI scopes voice and chat separately; voice typically carries a premium over chat

Unit math: For a flat per-interaction model — Total bill = interaction_count × per_interaction_rate. For Hippocratic AI’s time-weighted variant — Total bill = total_active_agent_hours × hourly_rate, where total_active_agent_hours = sum of all interaction durations in hours and hourly_rate ≈ $9.

The per-interaction model diverges from per-token or per-API-call billing by absorbing backend complexity. A patient call might trigger dozens of LLM inferences, several EHR lookups, and a post-call summary; per-interaction pricing presents one invoice line. That abstraction is a selling point in regulated industries where the buyer (a hospital CTO, a bank’s operations team) does not want to reason about token consumption. See the introduction to usage-based pricing for the framework, and choosing the right usage metric for when interaction-level meters outperform token- or output-based ones.


Companies using this

Four corpus companies use interactions as a primary or secondary billing unit, spanning healthcare AI (Hippocratic AI), customer-support automation (Maven AGI, Yellow.ai), and enterprise conversational AI (Uniphore). The cluster reflects where agent-human exchanges carry enough compliance and quality weight to make the interaction a natural billing anchor.

Company Product Pricing modelBilling unitsFree tier Verified
Hippocratic AISafety-focused healthcare LLM — patient-facing AI agents for non-diagnostic clinical tasksNo2026-06-10
Maven AGIEnterprise AI agent platform for customer supportNo2026-07-21
Observe.AIAgentic CX platform — contact-center AI agents, conversation intelligence & auto-QANo2026-07-23
UniphoreBusiness AI Cloud — enterprise conversational AI & agentic automationNo2026-06-09
Yellow.aiConversational CX automation platformYes2026-06-11

Explore this theme in the knowledge graph

FAQ

What is per-interaction pricing?

Per-interaction pricing is a billing unit where each discrete exchange between an AI agent and a human — a patient call, a support conversation, or an employee workflow session — is counted and charged as one interaction. Unlike per-message billing (which meters individual turns) or per-resolution billing (which requires a verified outcome), an interaction is typically a bounded engagement with a clear start and end, regardless of how many turns it contains or whether it resolves successfully.

How does per-interaction pricing differ from per-resolution pricing?

Per-resolution pricing charges only when the AI closes a ticket without escalation — you pay for successful outcomes. Yellow.ai, for example, charges $0.99 per resolution after 500 free sessions. Per-interaction pricing charges for the engagement itself regardless of outcome: Hippocratic AI meters patient interactions as active agent-hours, and Uniphore layers a per-interaction consumption charge on top of its platform license. The distinction matters for buyers: per-interaction costs accrue on every attempt, while per-resolution pricing puts some financial risk on the vendor to deliver a result.

Which companies use per-interaction pricing?

In this corpus, four companies meter interactions as a primary or secondary billing unit: Hippocratic AI (~$9 per active agent-hour of patient interaction), Uniphore (a per-interaction consumption charge on top of a platform license and per-seat fee), Yellow.ai (interactions and sessions on its Free tier), and Maven AGI (contracts scoped to conversation and resolution volume across chat and voice).

How is an interaction different from a message or a session?

A message is a single turn — one prompt or one reply. An interaction is the whole bounded engagement that a set of messages belongs to: one patient phone call, one support chat session, or one workflow run. Because the boundary is defined by the vendor, buyers should confirm exactly what starts and ends an interaction — a call of any duration, a chat that times out, or a completed workflow — before modelling spend.

Related billing units

Back to companies