Skip to main content
Consumption is the billing-angle view of your whole organization. Where the Organization dashboard reads the fleet through adoption and quality, this page reads it through what you actually consume: billable interactions, the workspaces and apps that generate them, LLM requests split by type, and a per-user breakdown you can charge back.
Consumption page with the interactions billing unit card, entry points, and per-workspace counts

​
Interactions, the Prisme billing unit

The headline block is Interactions (the Prisme billing unit). One interaction is one correlationId created on the Prisme side. It is the unit your consumption is counted in. The subtitle rolls the totals up as N interactions over 30 days (M over 7 days), and an info tooltip explains what feeds the count.
Interactions are not the same as messages. For a single member, the count adds up: each message sent to an agent, the automatic internal tasks each message triggers (title generation, summary, finalization), and interface browsing (agent listing, opening conversations). This is why the interaction figure is typically 6 to 7 times higher than the raw message count.

​
Where interactions come from

The left column, Where interactions come from, ranks the entry points that generate interactions, each with a horizontal bar and a count. The bar color encodes the kind of interaction:

​
Organization workspaces

The right column lists your organization’s workspaces with their interaction counts over the last 7 days and 30 days, so you can see which workspaces drive the most consumption.

​
Workspaces and apps inventory

Consumption inventory table listing workspaces, published apps, automations, pages, and interactions, with the per-user consumption table below
The Workspaces and apps table is a full inventory of the organization’s spaces. Each row is one workspace, and the header shows the total number of spaces. Per workspace you get:

​
LLM requests

The LLM requests card charts model activity per day, stacked into two series:
  • Completions: generation requests.
  • Embeddings: vectorization requests.
A period selector scopes the chart, and a toggle switches the unit between:
  • Requests: raw request counts.
  • Cost ($): the same series expressed in dollars.
The cost toggle only appears when cost data is available for the period. Without it, the card shows request counts only.
Below the chart, a per-user table breaks the same window down by member, with completions, embeddings, tokens, and cost per user.

​
Per-user consumption

The Per-user consumption card is the chargeback view: one row per member, so you can attribute spend across your organization. Columns cover completions, embeddings, messages, interactions, tokens, cost, and estimated CO2 (kg). Use the Export CSV action in the card header to download the full breakdown, for example to feed an internal cost-allocation or reporting workflow.

​
Aggregation freshness

Consumption analytics are computed from a backend aggregation. Blocks render only when their data exists, so a section stays hidden until the corresponding metrics are available for your environment.

​
Where to go next

Organization dashboard

Fleet-level KPIs for adoption, cost, and quality.

Adoption

User segments and recommendations to grow each one.

Users

The per-agent users table: who used a given agent and how much.