The blocker
The call ends, but the work does not. The agent writes a note that explains what the customer wanted, what was done and what happens next. They pick a disposition code from a long list, copy a booking number into another system, and maybe create a follow-up. All the while, the queue grows.
Teams try to fix this by capping wrap-up time. Agents then rush, notes get shorter, and the next person to speak with the customer has nothing to work from. The customer explains everything again.
Why it happens
- Notes are typed from memory. The agent has to recall and rewrite a conversation that just happened.
- The same data is entered twice. Details the customer already said out loud, like an order number, are typed into one or more systems by hand.
- Codes are hard to choose. Long disposition lists slow agents down and lead to “other” being picked far too often.
- Wrap-up is spread over several screens. Each extra system adds clicks and time.
How Amazon Connect solves it
Amazon Connect conversational analytics (Contact Lens) already has the full conversation. Generative AI turns it into the notes and data that agents used to write by hand.
1. Let generative AI write the summary
Generative AI-powered post-contact summaries give the essential information from a conversation in a short, structured format. Agents see them in the Contact Control Panel for voice, chat and email contacts, and AWS describes using them to complete after contact work (ACW) more quickly. Managers see the same summaries on the contact details and contact search pages, so they understand a contact without replaying it.
You turn summaries on in the Set recording and analytics behavior block. For voice, that means call recording on, speech analytics on, and post-contact summary selected under generative AI capabilities. Agents also need the matching security profile permissions.
2. Extract the details, not just the story
A summary is prose. Your systems need fields. Information extraction uses generative AI to pull specific information out of a conversation. It can be verbatim, such as a preferred name, invoice number or reservation ID, or derived, such as the reason for contact, the resolution provided or the next steps promised.
You define what to extract, and a rule decides when to extract it. The results are stored with the contact as structured data and are available in the Contact Control Panel during after-call work, on the contact details page, in contact search, in Amazon S3 output files, through the API and in Amazon Kinesis.
3. Give agents a prefilled wrap-up form
Amazon Connect provides a smart default guide for after contact work. It opens automatically when the agent enters ACW and shows disposition codes, the contact summary and follow-up tracking. AWS notes that new instances created after May 8, 2026 include it by default, and existing instances can import it.
To go further, prefill the guide with conversational analytics. Build a custom view, and the ACW guide opens with the contact summary and extracted fields already filled. The agent reviews, edits if needed and submits. AWS notes this works for voice and chat contacts, in views shown during ACW.
4. Send the notes where they need to go
Extracted information can be passed into rule actions such as sending emails, creating tasks, creating cases or sending notifications. If you use Amazon Connect Cases, cases can be created when contacts come in, and agents can use generative AI-powered case summarization to summarize a case’s fields, comments and related contacts. AWS notes this feature is available to unlimited AI customers.
If your system of record is somewhere else, a rule can generate an EventBridge event. AWS lists joining the data with your CRM as one use. A small integration listens for the event and writes the summary and fields to the right record, so nobody copies and pastes.
5. Measure the result
The After contact work time metric in the metric definitions measures the total time an agent spent doing ACW for a contact. Track it by queue before and after the change.
What good looks like
The call ends and the wrap-up screen is already filled in: a summary, the booking number, the reason for the call and the promised next step. The agent reads it, fixes anything that is wrong, picks a code and moves on. Notes are more complete than before, because they come from the whole conversation and not from memory. The next agent who speaks with the customer knows the story.
Where to start
- Measure after contact work time by queue, and sit with a few agents to see where the minutes go.
- Turn on post-contact summaries for one queue and ask agents to rate them for two weeks.
- List the three to five fields agents type most often, and create information extraction definitions for them.
- Prefill an ACW guide with the summary and those fields, then connect the output to your case or CRM system.










