Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/arainey2022/myaskai-docs/llms.txt

Use this file to discover all available pages before exploring further.

Your AI agent can only answer questions it has the knowledge to answer. The Improve section of My AskAI is a dedicated studio for finding what your agent can’t answer and giving it the tools it needs to do better. From automated knowledge gap detection to manual custom answers, behavior-shaping guidance, and deep conversation inspection — Improve gives you full visibility and control over your agent’s performance. Access everything in this section by navigating to Dashboard → Improve.
Improving your AI agent is an ongoing process, not a one-time setup. We recommend revisiting the Improve section regularly — especially after your first few weeks live — to catch and address gaps as your customers reveal them through real conversations.

Improve Sub-Features

Knowledge Gaps

Knowledge Gaps shows you every question the agent couldn’t answer and that resulted in a human handover. Questions are grouped by similarity so you can identify recurring patterns at a glance.For each gap, you can fix it immediately by writing a custom answer, generating a draft from your existing documentation, or generating a knowledge snippet from your Zendesk or Intercom ticket history. Once a gap reaches 3 questions, My AskAI auto-generates a suggested answer for you to review and publish.

Guidance

Guidance lets you create rules that shape how the agent behaves across three categories:
  • Communication Style — define tone, vocabulary, and writing style applied to every response.
  • Context & Clarification — instruct the agent to ask follow-up questions in specific scenarios before answering.
  • Handover & Escalation — define exact situations that should always trigger a human handover.
You can add up to 30 guidance rules per category (90 total). Each rule can be tested immediately using the adjacent chat window in your dashboard.

Custom Answers

Custom Answers (also called knowledge snippets) let you manually define the exact answer the agent should give for a specific question. Use them to correct inaccurate responses, fill gaps that don’t warrant a full documentation article, or lock in precise answers for high-stakes questions.For best results, enter the question in three different phrasings — this maximizes the chance the agent matches the custom answer when real customers ask.

Test

Test is a batch evaluation environment that lets you assess your agent’s performance before going live or after making changes. Import test questions from your Zendesk or Intercom ticket history, upload a CSV of up to 50 questions, or add questions manually.For each test question you can score the response, add notes on what to improve, and mark whether you’d send the response to a customer. Results can be exported and shared with your team as a CSV.

Content Usage

Content Usage shows you which knowledge sources and content articles are being used most frequently to answer questions, along with the AI resolution rate for each piece of content.Low resolution rates on a specific article can signal that the content needs updating or that customers using it are likely to need human support — useful for prioritizing documentation improvements.

Inspect & Logs

Inspect & Logs is a searchable log of every conversation your AI agent has had. Search by date, email address, ticket ID, or keyword to find specific conversations. Click into any conversation to see:
  • The knowledge sources used for each answer
  • How the AI interpreted the question
  • Which guidance rules were applied
  • A button to ask Echo (your AI copilot) to explain why the agent answered the way it did
From the same view, you can create a Custom Answer, add Guidance, or take other corrective actions without leaving the conversation context.

The Improvement Workflow

Use these tools together in a repeatable cycle to keep your agent getting better over time:
1

Identify Gaps

Start with Knowledge Gaps to see what your agent couldn’t answer in real conversations. These are your highest-priority improvement opportunities because they represent actual customer needs.
2

Fix Gaps

For each knowledge gap, choose the fastest fix: generate a snippet from existing docs, draft one from historic tickets, or write a custom answer from scratch.
3

Shape Behavior

Use Guidance to define communication style rules, clarification workflows, and escalation triggers that align the agent with your brand and support policies.
4

Test Changes

Use Test to validate how your changes affect performance — run your question set before and after updates to measure improvement.
5

Monitor Content

Check Content Usage regularly to see which parts of your knowledge base are being used and how well they’re resolving conversations without escalation.
6

Inspect Edge Cases

Use Inspect & Logs to dig into individual conversations where the agent’s response wasn’t ideal — understand the root cause and apply the right fix.
Self-learning runs automatically in the background and complements everything in the Improve section. It generates new knowledge articles from your human agent replies each week — so even without manual effort, your agent is getting smarter. See Self-Learning for details.

Build docs developers (and LLMs) love