Introduction
This article offers guidance on how to interact with conversational AI chatbots for more accurate solutions. By treating the chatbot as a fast librarian searching the Pima Community College (PCC) knowledge base, users can provide clear and detailed information that helps the AI locate the most relevant answers.
Detail
Be Specific and Avoid Vague Language
The AI chatbot uses internal articles to search for keywords and specific symptoms. Specificity helps the AI immediately reference the correct KB article.
Providing the exact application name and error message allows the AI to narrow down the search logic.
Avoid This: "The system isn't working for me today."
Use this instead: "I am getting an Error 404 when trying to access D2L."
Address One Issue at a Time
Mixing multiple problems can confuse search logic. Solving issues sequentially ensures the AI provides a clean, step-by-step resolution for each unique problem.
Avoid This: "My email is not receiving messages, my cellphone screen is broken, and my laptop keeps dying."
Use this instead: "The laptop will not hold a charge for more than 10 minutes even when plugged in."
Follow the Conversation Flow
If the chatbot asks a question, the chatbot is triaging or narrowing down the issue. Providing the requested data allows the AI to find the fix faster.
Tip: Answer the specific prompt (e.g., "Yes, I have already tried restarting the router") before moving to the next detail.
Caution: Do not skip ahead. If the chatbot asks for a department and the user replies with error details, the chatbot may fail to categorize the ticket correctly.
Balance Information Provided
Providing the right amount of information ensures the AI does not get distracted by irrelevant details.
| Input Level |
Result |
The Better Way |
| Too Little Information |
The chatbot will ask multiple follow-up questions, slowing down the resolution process. |
- Include the application name.
- Provide the exact error message.
- Specify when the issue started.
|
| Too Much Information |
The chatbot might focus on minor, irrelevant details and miss the primary issue. |
- Stick to facts regarding the current problem.
- Avoid providing history that is not relevant to the current fix.
|
Identify and Manage Hallucinations
AI hallucinations are confident but incorrect or fabricated outputs. These can include instructions for buttons that do not exist or mixed instructions from different processes.
- Verify via the Source: The chatbot will usually link the KB article used. Always click the link to verify the steps.
- Redirect the Chatbot: If the AI is incorrect, state: "That button does not exist. Is there another way?"
- Trust Training: If the AI suggests something that contradicts college policy, stop and consult a human.
Quick Tips
- Always copy and paste exact error codes or messages.
- Use plain language and avoid jargon or slang.
- Spell out an acronym the first time the acronym is mentioned if there is a chance the reader or AI will not recognize the acronym.
- Use the % symbol instead of spelling out "percent".
- Use a hyphen to indicate a range of numbers (e.g., 20-30 minutes).