Reduce support ticket handle time with session summaries

Accelerator Customer Support Architect

Your customer already had a rough moment. Don't make them explain it twice. Add an AI summary of their Fullstory session right in the support ticket, so your agent knows what happened before they say hello.

The business outcome

When a ticket comes in, your support agents start from scratch. They don't know what the customer clicked, what broke or what the customer already tried. So the customer explains it all again, already frustrated, while the agent works out what went wrong.

That costs you three ways:

  • Longer handle time. Minutes go to "Can you walk me through what happened?"
  • Slower time to resolution. The fix can't start until the problem is understood.
  • Lower CSAT and NPS. Nobody likes repeating themselves to get help.

With a session summary in the ticket, your agent opens with "I can see you hit an error at checkout" instead of "What happened?" The customer feels heard, and the fix starts sooner. The summary also shows what the customer already tried, so your agent skips the "have you cleared your cache?" round trip that turns one ticket into two.

Measure by comparing tickets with a summary against tickets without, on first contact resolution rate, average handle time, time to resolution and CSAT.

How it works

A ticket comes in

Your ticketing tool finds that person's recent sessions.

AI writes the brief

Your prompt tells it what to pull out: their goal, errors, rage clicks and last action.

Support takes over

The summary is in the ticket, so they reply already knowing what happened.

Customer story

A leading gaming company measured average handle time across multiple contact reasons. Tickets where Fullstory sessions were provided ranged from a 7% to 27% reduction in handle time. Registration and login questions decreased handle time by nearly two minutes.

Before you start

Start with a session from a recent support ticket. There are a few ways to find one; this approach uses a Segment and the customer's email address. If you don't store email addresses in Fullstory, adjust the Segment to match whatever you do store.

  1. Copy a customer's email address from your support ticketing system.
  2. Open the Segments page, set USER FILTERS to Email is the customer's email address, and click Apply Filters.
  3. Click the matching sessions link to see every session for that person, not just the most recent.
  4. Click any play icon to open the Session Replay, then click Show More in the upper right.
  5. Copy the Client ID.

Create a prompt that summarizes sessions

A prompt governs how a session gets summarized — what the AI pulls out and how it writes it up. Fullstory stores each saved prompt as a profile, and the Profile ID is how you call it from the API later.

  1. Click your org name in the upper right and open Settings in a new browser tab, then click Anywhere Prompt Manager.
  2. Open the FCR Support Brief template. To find it manually, click New prompt, then Start with template, choose the Universal industry, and select the FCR Support Brief use case.
  3. Paste the Client ID into the Session Summary Output field and click Run Prompt.
  4. Edit the Prompt until the summary tells an agent what they need. Give it a meaningful Prompt Name and click Save.

The result is a brief an agent can read quickly. Here's the shape of it — this example is made up, but it's what the prompt produces:

WHAT THEY WERE TRYING TO DO
Complete checkout on a two-item cart using a saved credit card.

WHAT WENT WRONG
Payment step returned "We couldn't process your card. Please try again."
after each submit. Three attempts, same message, no card change.

WHAT THEY ALREADY TRIED
Re-entered the card manually. Switched to a second saved card.
Opened the "Payment methods" help page, returned to checkout, retried.

CONTEXT FOR THE AGENT
Chrome 142 on Windows. 9-minute session. Rage clicks on Place order
before navigating to Contact us. Customer is frustrated.

SUGGESTED OPENING LINE
"I understand your payment isn't going through at checkout — let me
take a look at what's happening with that card."

Send user details to Fullstory

So far you've searched for a session by hand. To make this automatic, Fullstory needs to hold the same identifier your ticketing system has, so your integration can look the person up.

  1. Attach the UID to sessions with the Identify Users API after your app knows who the user is (e.g. login).
  2. Attach the email address with the Set User Properties API, either along with the UID or instead of it.

How you wire this up depends on your application, so you may need your engineering team here.

Retrieve sessions using the API

Once the UID or email is on your sessions, you can pull a customer's sessions programmatically.

Run the List Sessions API. Replace <API_KEY_VALUE> with the API key you created earlier and update the email query parameter to be the customer's email address. Encode the @ symbol with %40, as shown here:

curl -L 'https://api.fullstory.com/v2/sessions?email=user%40domain.com' \
  -H 'Accept: application/json' \
  -H 'Authorization: Basic <API_KEY_VALUE>'

Each object in the results array is one session. The id property is the same Client ID you copied earlier, and created_time lets you pick the session closest to when the ticket was filed.

{
  "results": [
    {
      "id": "50e26159-914f-4b59-a19a-f1a72180993c:3a8a00ca-81e5-4bbb-b164-dcb7289f504a",
      "app_url": "https://app.fullstory.com/ui/org/session/6456738715912035756:4279788225117469114",
      "created_time": "2026-11-09T16:25:33Z"
    }
  ]
}

Summarize a session using the API

Now summarize those sessions with the prompt you saved.

  1. Get the Profile ID for your prompt. It's listed for each prompt on the Prompt Manager page, and there's a Copy button on the prompt's edit page. You can also list every prompt and its ID with the List Profiles API.
  2. Run the Generate Summary API, swapping in the session's id from the previous step and your Profile ID for the config_profile query parameter. Encode the colon in the session ID as %3A, as shown here:
curl -L 'https://api.fullstory.com/v2/sessions/50e26159-914f-4b59-a19a-f1a72180993c%3A3a8a00ca-81e5-4bbb-b164-dcb7289f504a/summary?config_profile=PROFILE_ID' \
  -H 'Accept: application/json' \
  -H 'Authorization: Basic <API_KEY_VALUE>'
  1. Compare the output to what you saw in Prompt Manager. It should be mostly the same.

Tailor the summary to the customer's complaint

A saved prompt is the same for every ticket. The Generate Summary with Overrides API lets you append the customer's own words to it, so the AI focuses on the part of the session that matches their complaint.

  1. Go back to the FCR Support Brief template.
  2. Append the following text to the default prompt and add the reason the customer opened the ticket by replacing [PASTE REASON].
The customer has stated the following reason for opening the ticket:
[PASTE REASON]
  1. Click Configuration to reveal advanced prompt options and expand the Preview JSON Request toggle switch. Copy the text value from the postPrompt property.
  2. Post the text as the post_prompt override. Replace the SESSION_ID used in the previous section. The example below already has steps 2 and 3 applied:
curl -L -X POST 'https://api.fullstory.com/v2/sessions/SESSION_ID/summary' \
  -H 'Accept: application/json' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Basic <API_KEY_VALUE>' \
  -d '{
    "config_profile": "PROFILE_ID",
    "llm": {
      "post_prompt": "Write an FCR brief under 200 words. Use section headers. Phrases over full sentences.\n\n**WHAT THEY WERE TRYING TO DO**\nOne sentence — the customer'"'"'s primary goal.\n\n**WHAT WENT WRONG**\nThe specific failure point. Quote any visible error messages exactly.\n\n**WHAT THEY ALREADY TRIED**\nSelf-service steps taken: help pages visited, retry attempts, settings changed.\n\n**CONTEXT FOR THE AGENT**\nBrowser, device, session duration. Note if the customer appears frustrated.\n\n**SUGGESTED OPENING LINE**\nOne sentence the agent can use that acknowledges the issue without revealing session access.\n\nThe customer has stated the following reason for opening the ticket:\nMy card keeps getting declined at checkout but my bank says there is no problem."
    }
  }'
  1. Review the output. It should be more aware of the customer's complaint than the static prompt.

The post_prompt override is a single JSON string, and the text copied from Prompt Manager already has the needed \n characters. In production, build this body with your language's JSON library rather than by hand, which handles the new line escaping for you.

What's next

You've now done the three things an integration does:

  1. Create and store a summary prompt.
  2. Retrieve sessions for a customer.
  3. Summarize a session with that prompt.

The last step is a widget in your ticketing system. When an agent opens a ticket, it looks up the customer's email, calls List Sessions, picks the session closest to the ticket's creation time, and posts the summary into the ticket as an internal note.


Was this article helpful?

Got Questions?

Get in touch with a Fullstory rep, ask the community or check out our developer documentation.