Team performance
Analytics

Team performance

How your human agents perform once a conversation is handed to them — response times, SLA attainment, per-agent results — and how reliably the AI agent itself runs.

The Team page at https://app.asks.app/analytics/team answers "how do humans perform after the AI hands off?" It uses the same date range as every analytics page and is available on the Business plan and above Business and up. On other plans it shows a sample preview with an upgrade prompt.

Team page with the median and p90 First response after handoff tiles, and the SLA performance card showing first-response, next-response and resolution attainment, the live at-risk line, and the per-policy table, above the Agents table
Analytics → Team

First response after handoff

Two tiles show how long customers wait for a person: the median and the p90 (9 in 10 customers heard back within this time). Each measures the time from a conversation's first assignment to a teammate — a takeover or an assignment — to your team's first public reply after it. Internal notes don't count.

Response-time stats need at least 5 handoffs in the period; smaller samples show "Not enough data" rather than a misleading number.

SLA performance

When your workspace uses SLA policies, this card shows:

  • attainment per metric — First response, Next response, and Resolution — with met and breached counts
  • a live line of open conversations at risk or breached right now
  • a per-policy table: policy, attainment, met, and breached

The card stays empty until a policy is active.

Agents

A table of every teammate's activity in the period:

ColumnWhat it counts
MessagesPublic replies the teammate sent.
ResolvedConversations that closed while assigned to them.
CSATThe average rating on those conversations.

AI health

Four tiles describe the reliability of the AI agent itself across every AI turn in the period:

TileWhat it shows
Success rateThe share of AI turns that completed successfully.
Failed turnsThe share of AI turns that failed.
Model failoverTurns served by the backup model when the primary one was unavailable.
Reply latencyMedian time to reply, with the p90 underneath.

A rising failed-turn rate is worth a look even when CSAT is steady — check Escalation for what happens to customers when the AI can't answer, and the status page for incidents.

Chat on WhatsApp