Axero Copilot | Axero Documentation

What Is Axero Copilot?

Axero Copilot is a chat assistant built on OpenAI's GPT models. It runs inside your intranet, so the content it reads and the conversations it has stay private to your organization.

A search engine hands back a list of documents. Axero Copilot reads the matching pages and answers the question in a few sentences, and you ask it the way you would ask a colleague.

Axero Copilot Features & Capabilities

Search Results: Axero Copilot searches your published content and reads the matching pages in full before answering. 

It searches these content types:

• Articles
• Wikis
• Blogs

Excluded content types are:

• Events
• Tasks
• Cases
• Recognition

Its main job is searching documentation content, but it can also read and understand:

• A subset of user profile data
• Text on Page Builder pages

Access to Intranet Content: Axero Copilot reads your organization's intranet content and nothing outside it. That includes articles, presentations, and files. In the chat, it can read and quote text from supported document files, including PDF, DOCX, PPTX, and XLSX, as well as image-based or scanned PDFs.

Conversational Interface: You ask in plain words and get an answer in plain words, with no search syntax to learn.

Personalization and Permissions: Each person gets answers drawn only from content they are allowed to view, so nothing leaks through the chat that would not show in search.

Content Analysis and Relevance Scoring: Each page Axero Copilot draws on gets a relevance score, and the most relevant details come first.

Mentions: Use @mentions to point Axero Copilot at a person, space, or piece of content and pull information about it into the conversation. Typing the @ symbol in the Axero Copilot chat box opens a dropdown list of matches, and only items the user has permission to view appear in it. Selecting an item inserts a formatted reference (example: @Laura Carter), and Axero Copilot returns a profile summary, content details, or related links, depending on what is mentioned. Mentions are enabled by default.

Conversation History: Axero Copilot saves conversations, so users can return to previous chats and continue where they left off. In full-screen mode, prior conversations can be viewed and searched for the selected Axero Copilot profile.

Customization Options: You can change the name, the logo, and the look of the chat window to match your intranet. Three other options change how people use it:

• Full-Screen Mode: The chat fills the window, which suits a longer back-and-forth.

• Suggested Questions: Ready-made questions, each with an icon, appear at the start of a chat. Clicking one asks it, so people have somewhere to begin.
• Inline placement: Axero Copilot can sit inside a page as part of its layout, rather than in a floating window.

How Axero Copilot Works

When a user asks a question, Axero Copilot finds relevant information and answers with a short text summary. Under the summary, a divider separates the answer from the pages it drew on. Each of those references carries a relevance score, and only pages the person is allowed to open are listed.

• Search and Retrieval: When someone asks a question, Axero Copilot searches the intranet for content that answers it.
• Content Analysis: Axero Copilot reads the documents it found and pulls out the details that answer the question.
• Conversational Interaction: The person asks in plain words and gets a short answer drawn from that content.
• Relevance Scoring: Each source gets a relevance score, and the most relevant comes first.
• Feedback Option: Users can give feedback on each answer. If it wasn't helpful, they can pick the reason and add details before submitting.

Are Customizations Possible?

Axero Copilot can be customized through the Axero Copilot Manager, which lets you adjust it for your organization:

Appearance and Branding
Update the Axero Copilot name, avatar, launcher icon, and loading animation to match your intranet’s branding.

Suggested Questions
Add, reorder, and style Suggested Questions to point users toward common tasks and give them a place to start.

Knowledge Access Controls
Control what Axero Copilot can search: limit it to specific spaces, content types, or tags, or use a dedicated Bot User for consistent, permission-aware results. 

Multiple Profiles (if enabled)
Create separate Axero Copilot profiles for different audiences or use cases, such as HR, IT support, leadership, multilingual sites, or regional teams.

Axero Copilot Analytics

Analytics shows administrators what people asked, what Axero Copilot answered, and how they rated it, with sorting, filtering, a choice of timeframe, and the detail of each session.

Read Axero Copilot Analytics for the filters, timeframes, and session details available in the dashboard.

Axero Copilot Best Practices

These practices help a rollout go well.

Content Size Optimization

Axero Copilot reads a lot of content, but there is a limit to how much it reads before it answers. If it struggles with large files, contact your Client Success Manager. The usual remedies are adjusting the prompts, moving to a GPT model that reads more, or splitting a long PDF into parts.

You might keep the full document where people read it and put the parts in the profile's private space. If you do, answers drawn from those parts have no reference a reader can open. Add a redirect back to the full document at the end of each part's Description field, so every answer that uses a part links to the original. See Axero Copilot Manager for how redirect lines are written.

Formulating Effective Queries

Ask specific questions in plain language, and give context: the period, the department, the product. A precise question gets a precise answer.

• Here is an example of formulating an effective query for Axero Copilot:
  • Ineffective Query: "Budget report"
    • The ineffective query, "budget report," is so general that it could match many different documents.
  • Effective Query: "Can you provide the budget report for Q1 2024 for the Finance department?"
    • The effective query specifies the period (Q1 2024) and the department (Finance), which gives Axero Copilot clear limits to search within.

Narrowing/Focusing Responses

If the first answer is too broad, ask a follow-up question or add context, and the conversation narrows.

• Here's an example of narrowing or focusing a response after receiving an initial answer from Axero Copilot:
  • Initial Query
    • User: "How do I set up a virtual meeting using our company's video conferencing software?"
  • Initial Response
    • Axero Copilot: "To set up a virtual meeting using our company's video conferencing software, follow these steps: 

1. 1. 1. 1. Sign in to the software using your company credentials.
         2. Navigate to the 'Schedule Meeting' option.
         3. Enter the meeting details, including the date, time, and participants.
         4. Choose the desired settings, such as audio and video preferences.
         5. Click 'Schedule' to confirm the meeting."

• • Narrowing/Focusing Response: 
    • User: "Thanks for the steps. Could you also clarify how to invite external participants to the meeting?"
      • A follow-up question about one part of the first answer gets the user information aimed at what they actually need. Here, the user asks how to invite external participants to the virtual meeting, so Axero Copilot can give instructions or tips for that specific step. 

Understanding Responses

Tell people to look at the relevance score on each source: the higher it is, the closer that page matched the question.

• Here's an example of the relevance score (r:90) from Axero Copilot:

Providing Feedback and Types of Useful Feedback

Ask people to rate answers, and to say why: what was wrong, what would have helped, or what worked. That is what improves the answers over time.

Note: Axero Copilot can be configured to require feedback after each interaction, so input is collected consistently for ongoing improvements.

Teach users to tell apart feedback on what an answer said (accuracy, relevance) and feedback on how Axero Copilot itself works (ease of use, speed).

The feedback panel looks like this:

• Here's an example of good and bad feedback:
  • Helpful Feedback: "The response provided by Axero Copilot was accurate and helped me quickly find the information I needed. It precisely addressed my query and saved me a lot of time."
    • Why it's helpful: It confirms the answer was accurate, says how it helped, and notes the time Axero Copilot saved.
  • Unhelpful Feedback: "Axero Copilot's response was inaccurate. It didn't address the main point I was asking about. This tool is useless."
    • Why it's unhelpful: It says the answer was inaccurate but not what went wrong, and gives no detail about the question. It also uses a strong adjective ("useless") without suggesting any improvement.

What Your Axero Copilot Can't Do

Knowing what Axero Copilot cannot do keeps expectations realistic. It answers from what is written on your site, so these are the situations where it falls short:

Lack of Documented Procedures

• User Query: "How can I troubleshoot an issue with my Lenovo ThinkPad P14s workstation?"
• If your Axero site does not document how to troubleshoot the issue, Axero Copilot cannot give detailed or complete guidance. Incomplete or outdated documentation limits the troubleshooting steps it can find, so people asking for help with technical issues get unsatisfactory answers. 

Insufficient Coverage of Specialized Topics

• User Query: "What are the best practices for communicating with the IT department?" 
• If your Axero site has little on a specialized topic such as communicating with IT, Axero Copilot will struggle to answer best-practice questions about it. Limited or outdated content on the topic leaves it less relevant information to find, so its answers are less helpful.

Highly Contextual or Personalized Queries

• User Query: "What tasks should I prioritize for my role as a marketing manager in the upcoming quarter?" 
• Axero Copilot will struggle to recommend priorities based on a user's specific role and responsibilities in your organization. That takes an understanding of the person's own context and priorities, which Axero Copilot cannot work out from your intranet content alone. 

Ambiguous or Unclear Queries

• User Query: "Tell me about the latest content."
• The query "latest content" does not name a topic, so Axero Copilot cannot tell what the person wants to know. Without more context, it gives a general answer that may not cover what they meant. 

Sensitive or Confidential Information Queries

• User Query: "What is my salary?" 
• Axero Copilot answers only from intranet content the person asking is allowed to view. It cannot retrieve sensitive or confidential information, such as salary details, unless that information is stored on the intranet where the person can already view it. 

In these situations, a person or an outside source is the better route.

Related pages

• Axero Copilot Manager (to create or change a profile, its branding, and its knowledge access)
• Axero Copilot End-User Experience (how to place the widget on a page and what people see)
• Axero Copilot Analytics (to review the questions people ask and the answers they rated)
• AI Assistant for Content Editors (the assistant inside the content editor)
• Axero AI-Powered Assistants (the other AI assistants)