Skip to navigation

Release notes

November 10, 2026

Updated automated resolution assessment

Starting November 10, 2026, Ada is gradually updating how it assesses automated resolution to improve accuracy. Ada starts using the updated assessment for each AI Agent at a different time, and the AI Agent’s automated resolution rate may go up or down from that day. Conversations from before that day keep their original classification.

What isn’t changing

Ada’s definition of a resolved conversation stays the same: a conversation is Resolved only when the AI Agent’s help is relevant, accurate, safe, and contained. Conversations handed off to a human agent are still Not Resolved. Your AI Agent’s behavior doesn’t change.

What’s changing

1. A newer, more capable AI model. It better understands complete conversations.

2. More complete business context. The assessment now sees more of the information your AI Agent used in the conversation, so it can check answers against your business rules and data:

  • Variable values saved during the conversation, such as the customer’s plan or account status.
  • Glossary definitions.
  • Playbook step instructions, and results from Code tools and connected systems.
  • Custom Instructions that applied in the conversation.

It still uses your knowledge, Processes, API tools, Coaching, and company description.

3. CSAT survey responses are no longer used, because a customer can get the right help and still leave a low rating.

4. A new classification: Excluded. This covers conversations where the customer didn’t ask for help, such as spam, a greeting on its own, or a message that isn’t about your products or services.

5. A clearer test for whether the customer got help. A conversation can be Resolved when the AI Agent answers the question correctly or gives steps that solve the issue, even if the customer doesn’t reply. It’s Not Resolved if the AI Agent only collects information and never answers, leaves a mistake uncorrected, or doesn’t do what it said it would.

6. Failed or offline handoffs. If a handoff through an app fails or live chat is offline, the conversation returns to the AI Agent. Ada now checks whether the AI Agent resolved the request, instead of marking it Not Resolved. All other handoffs are still Not Resolved.

How Excluded affects your automated resolution rate

The formula stays the same: Resolved / (Resolved + Not Resolved). Excluded conversations aren’t counted in either number.

10 conversations 6 Resolved 2 Not Resolved 2 Excluded Automated resolution rate6 / (6 + 2) = 75% Not counted in the rateStill counted in conversation volume

What you’ll see in your reports

Once Ada uses the updated assessment for your AI Agent:

  • The Automated resolution and containment report shows a banner explaining the change, and a marker on the chart for the date Ada started using the updated assessment.
  • Conversations show Resolved, Not Resolved, or Excluded, with a gray icon for Excluded. You can also filter conversations by Excluded.
  • If no conversations in a date range count toward the rate, the report shows ”—” instead of 0%.
  • Comparisons with the previous period only appear when both periods use the same assessment.

A date range that includes the change combines both assessments, so for a clean comparison, choose a date range that starts after the marker.

Data Export API and Ada MCP Server