p↗ PAM-E
Workspace / Escalated cases
Cases Runs on your computerTechnical lab ↗
HANDOFF CHECK

What will the next agent know?
Make every detail count.

Pick an escalated case or paste your own conversation. See what each handoff passes on, what it leaves out, and what the next agent does with it.

Each case was escalated from first-line support (L1) to second-line (L2).

Both methods always get the same space. Smaller = more gets left out.

Preparing your workspace…
START HERE

Watch one conversation become two handoffs.

Generic PAM and PAM-E receive the same conversation and the same space. See what each one passes on, what it leaves out, and what the next agent does with it.

Browse the case library
01

Pick a case

Escalated support conversations across different industries, or paste your own.

02

Compare the handoffs

Side by side: every detail carried forward or left out, with the exact words.

03

See what happens next

A simulated next agent works the case with each handoff, and the result is checked.

Start with a story

All cases ↗
NEW HANDOFF

Check a handoff from your own conversation

Paste a support conversation or import a conversation JSON file. This preview compares what each method transfers; it does not score whether the case is solved.

Offline preview accepts new conversations, but can miss unfamiliar wording. AI extraction needs OPENROUTER_API_KEY in the server environment and a restart. No model calls run automatically.

Start each message with Customer: or Agent:. Up to 100 messages and 20,000 characters.

One phrase per line, copied from the conversation. These checks never influence selection.

Download example format ↓
Offline mode uses local rule-based extraction. If you enable real AI models, conversation text is sent to the configured model provider. Handoff text is saved in the local audit log; exported evidence also contains your text. Use anonymised conversations.
EXAMPLE LIBRARY

Start with a customer story.

Internal scenarios, not published-dataset evaluation. Every example runs through the same comparison pipeline.

A clearer view of the handoff.

01

Read the conversation

The first agent gathers information. We show the actual recorded messages and system activity.

02

Compare what travels

Generic PAM and PAM-E get the same conversation and space limit. Green means tracked words are present; coral means some are missing. This is a text-preservation check, not a guarantee of meaning.

03

See the next step

Example cases replay a simulated next agent and check its actions against the case requirements. Your own conversations show transfer only; no outcome is invented.

This is a local research and demonstration workspace. A company deployment needs authentication, access controls, data handling policies and independently validated live-model performance.
FOLLOW THIS DETAIL