Analytics Boards

​

All analytics

​

Switching boards

​

The boards

• Usage and cost
• Flow activity
• Agent activity
• Knowledge base activity
• API usage
• Models
• Users
• My usage

What did this workspace use and cost?

Widget	Shows
AI Operations used	The AI Operations the workspace used, shown exactly so it compares with a limit
AI model usage cost	What its model calls cost
AI models total tokens	Tokens sent to and received from models
Active workspace users	Members who did something in the workspace in the period
AI Operations over time, Model tokens over time	The same, per hour, day, week or month
AI Operations per tool, Model tokens per tool	Which flows, agents and knowledge bases used them
Usage by model	Calls, tokens and cost per model
AI Operations per user, Model tokens per user	Who used them

How are flows running?

Widget	Shows
Total flow runs, Success rate, Avg. flow run time	How many runs, how many succeeded, and how long they took
AI model usage cost	What the flows’ model calls cost
Runs by how they started	Runs over time, split by API, editor, trigger, app, agent or test
Runs per flow, Tokens & cost by flow	Which flows run most and cost most
Runs per user, Tokens & cost per user	Who runs them
Run time	Run time over time, on average or at P25, P50 or P90
Tokens & cost by model	Which models the flows call
Failed runs over time, Failed runs per flow	When runs fail, and which flows fail

How are agents doing?

Widget	Shows
Conversations started, Messages sent	How much agents are used
Failed replies	The share of agent replies that failed
AI model usage cost, AI Operations used	What conversations cost and used
Chats started, Messages sent	The same, over time
Chats started per user, Messages sent per user	Who chats
Usage by agent	Conversations, messages and cost per agent
Tokens & cost by model, Tokens & cost per user	Which models answered, and for whom
Failed replies over time	When replies fail

What happened to documents?

Widget	Shows
AI Operations used	What ingesting files used
Documents uploaded	Documents added over time
Documents per Knowledge Base	Where they were added. Click a row to filter the board to that knowledge base.
Documents by user, User activity	Who added them

How is the API used?

Widget	Shows
API Calls	Calls over time
API Endpoint calls	Calls per endpoint
API Logs	Each call with its endpoint, method, status and duration, a page at a time

Filter it by the API key that made the calls. Each key is listed by name, with the last four characters of the key.

Which models are used, and what do they cost?

Widget	Shows
AI model usage cost, AI models total tokens	What models cost and how much they processed
Model calls, Average cost per call	How many calls, and what one costs
Model calls over time	Calls per hour, day, week or month
Usage by model	Calls, tokens and cost per model
Full answer latency, Time to first token	How long models take to answer, and how soon they start; each picks Average, P95 or P99
Request success rate	The share of calls that succeeded
Model response time	Both timings over time, or one of them by model

A model call is a call to a language, decision or voice model. Each retry is a call of its own, and a call answered from the platform’s cache is not counted as a call. Voice calls count in calls and cost, not in tokens. Latency and success rate cover language and decision calls.Filter › Model type keeps the models of the types you pick, Speech, Decision or LLM, historical and custom models included. Pick several to combine them, and narrow further with Model. The organization’s Models board has the same filter.Model response time draws each timing as a weighted average across the selected models. Only completed calls not served from cache count, and a call without a first token time is a gap, not zero.

Who uses the workspace, and how much?

Widget	Shows
Active members, Idle members	Members who ran a flow or talked to an agent in the period, and members who ran nothing
Active members over time	Active members per day, week or month
Model usage per user, AI Operations per user	What each member used
Flow runs per user, Messages sent per user, Documents added per user	What each member did

What did I use and what did it cost? For every member.

Widget	Shows
My AI model cost, My AI Operations, My runs	Your own cost, AI Operations and runs in the workspace
My AI model calls over time	Your model calls over time
My usage by model	Your calls, tokens and cost per model

My usage needs no analytics permission and never shows anyone else’s usage. Members without View workspace analytics land straight on it.

​

Reading a number

• Change badge: the number’s change against the period before. Green is good and red is bad, by what the number measures: fewer failed runs is green, fewer runs is red. Hover over it to see which period it compares with.
• Period before: today compares with yesterday up to the same hour, the current billing cycle with the last one up to the same day, the previous billing cycle with the whole cycle before it, and any other period with the one of the same length just before it.
• Cost: calculated from the tokens each call used, at the model’s catalogue rate, so it can differ from your provider’s invoice. An asterisk means some calls have no token count yet and are counted at their expected cost. Hover over it to see how it adds up. See AI model usage cost.
• Info icon: explains what the number counts, and links to its documentation.
• Not applicable: a widget whose number can’t be narrowed by an active filter says so, rather than showing a number that ignores the filter.

​

Exporting