Assembled AI Agents

It would be very helpful if the top panels on the QA pages adjusted for search. Right now, I believe it only adjusts for filters? And then moving the search above the panels once that's functional would be cool. Let me know if I'm missing something on this! It would also be helpful to be able to adjust # of conversations on each page, and see total number of pages with a way to jump between them more readily.

As a support user in Co-pilot, I want a button that will read my KB and guide me through any troubleshooting steps, process, or operational process that is published. This should initiate an interactive conversation with the agent where they can reply to CAL with the outcome of their investigation to guide their next step. Example would be having a guide on how to trouble-shoot a purchase failing. If a guide has clear steps, the Guide Me button should reference that information, and ask them to perform the steps in sequence with results to be given back.

We are looking to have conversation_id added to the payload of the message_sent payload event. Currently we only get the below, but the conversation ID would really help with our reporting! { message: string; role: 'user' | 'assistant' | 'agent'; created_at: number; }

Metrics such as Error count and Handoff rate should be clickable to show the specific tickets behind that number, instead of manually hunting for the relevant cases in the lower tables / other views.

Today, you can view the version history for workflows, but it would be great if we had the ability to revert back to any previous version. This ensures changes can be safely reverted if an issue is introduced.

We want to use MCP to analyze what is happening across our AI Agent conversations, but today we don't have access to the underlying conversation data. We can see workflow monitoring data and access the summaries that get passed into our cases, but the summary alone doesn't contain enough detail for the types of analysis we want to do. Ideally, we'd be able to use MCP to query the full transcripts across AI Agent conversations, filter or segment them by customer/case attributes, and use an LLM to identify themes and trends. This would allow us to do much deeper voice if customer and QA analysis without having to export conversation data into another system first.

Example use cases: Checking whether a user is signed in, for example, user_id is not blank Referencing values captured in previous workflow steps, or previous runs of the workflow such as an invoice ID Comparing the current value with a previously handled value, such as invoice_id is not the same as the last processed invoice Preventing a workflow from re-triggering when the same value already exists in the conversation history Allowing a user to re-enter the workflow after signing in or supplying a new value

We would like our chat bot to be more personal and human-like and to do so would like "AI Agent" to be removed from the UI beside the bot's name.

Unsure what criteria determine which items appear in the Top issues section. It can surface workflow categories or common intents that do not represent issues, and it does not populate consistently across workflows. Suggesting showing the most frequent issues that exist, rather than only items selected as especially important, or providing an explanation for why each item appears.

The current static chart is not clear. Recommend adding trend lines to make performance changes over time easier to interpret. The workflow summary metrics are useful, but want the view to have clearer investigation paths into the underlying performance and tickets.