Find the questions that stump your team. Joy reads the support conversations where agents couldn't find an answer, ranks the topics by how often they come up, and shows you which knowledge-base gaps to close first.
For
Knowledge Managers
Support Leads
CS Operations
Enablement Teams
Sample prompt to Joy's Assistant
Look through the support conversations from [time period] where the agent couldn't find an answer, and rank the topics that stump us most often so I know which knowledge-base gaps to close first.
What it can look like
An illustrative example. Your output will reflect your own sources, data, and prompts.
Look through the support conversations from the last 60 days where the agent couldn't find an answer, and rank the topics that stump us most often so I know which knowledge-base gaps to close first.
I went through the 430 conversations from the last 60 days where the agent had no knowledge-base answer, grouped them by topic, and checked each against what's in your Knowledge Center. Five topics drive most of the misses and have nothing documented behind them.
I read the 430 support conversations from the last 60 days where the agent couldn't find an answer in the knowledge base. Either they escalated or told the customer they'd follow up. Five topics account for 372 of them, and none has a current article behind it.
Conversations with no knowledge-base answer (last 60 days)
SSO / SAML setup118
Data export limits94
New billing tiers77
Webhook retries52
Mobile offline mode31
Here's where the volume sits, and what I'd write first.
SSO / SAML setup: 118 conversations. Agents piece the answer together from memory each time; a single setup guide covering the common identity providers would deflect most of these.
Data export limits: 94. Customers keep hitting row caps they can't find documented anywhere, then ask support to confirm them.
New billing tiers: 77. Since the pricing change, agents have no reference for what each tier includes, so answers vary agent to agent.
Webhook retries: 52. A short developer-facing note on retry timing and back-off would cover nearly all of these.
Mobile offline mode: 31. A quick FAQ on what does and doesn't work offline handles the recurring confusion.
Start with the top two: SSO and data export limits are 212 of the 372 conversations on their own, so documenting them clears the most deflectable volume for the least writing. The other three are worth queuing next, but they can wait a sprint.
Want me to draft the SSO setup guide first, or map all five gaps into a documentation backlog you can work through?
From dead-end chats to a writing list
Knowledge Gap Identifier asks Joy to read the support conversations where an agent escalated or couldn't find an answer, group them by topic, and rank them by how often each one comes up. The result is a ranked list of the questions that stump your team and have nothing documented behind them.
Point Joy at the conversations
Make sure your support chats, tickets, and escalations are connected to the workspace or in your Knowledge Center. Joy reads what's there when you ask.
Say what counts as a gap
Tell Joy the window to look at and what a miss looks like (an escalation, a “let me follow up,” or a case with no article attached), then ask it to rank the topics.
Review the ranking
Get a ranked chart of the topics that stump the team, each with its conversation count and examples of what customers actually asked.
Turn it into a plan
Ask for a prioritized writing list, then copy it into your documentation backlog or hand each topic to whoever owns it.
Make it one click for your team
Save this ask as a custom command on the assistant your team already uses, so anyone can run it in one step.
Make it yours
Ranked by Real Volume
Topics are ordered by how many conversations they stall, so you write what customers ask about most.
Reads the Dead Ends
Joy focuses on the conversations that escalated or went unanswered, exactly where the gaps live.
Evidence per Gap
Every topic comes with its conversation count and example questions, not just a label.
A Ready Backlog
The output is a prioritized list you can drop straight into your documentation queue.
One Product Area
Narrow the analysis to a single product or team to plan its documentation.
Month Over Month
Compare gaps across periods to see whether new articles are actually deflecting.
Customer vs. Internal
Split gaps by who's asking to know which knowledge base needs the work.
Straight to Draft
Have Joy draft the top article as soon as the biggest gap is identified.
Frequently Asked Questions
How do I find knowledge-base gaps?
Joy reads the support conversations where an agent couldn't find an answer (escalations, follow-ups, and cases with no article attached), groups them by topic, and ranks them by volume. The topics with the most misses and no documentation are your biggest gaps.
What counts as a question agents can't answer?
Any conversation that escalated, ended in a “let me check and get back to you,” or closed with no knowledge-base article attached. Joy compares these against your Knowledge Center to confirm the topic is genuinely undocumented.
How is this different from analyzing all support volume?
This focuses on the dead ends (the conversations where the answer wasn't available) rather than every ticket. That surfaces gaps that hurt deflection, which high-volume-but-already-documented topics would otherwise hide.
How does it decide what to prioritize?
Gaps are ranked by how many conversations each topic stalls, so the questions that stump your team most rise to the top. Each gap comes with its count and example questions so you can sanity-check the order before you write.
Can Joy write the articles too?
Yes. Ask Joy to draft the top gap as an article, then review and copy it into your knowledge base. Joy drafts the text in chat; publishing it to your help center is your step.
