Answering Questions with AI from Website Content

Wouldn’t it be amazing if you could visit a website and ask a question that was answered using AI based only its curated content related to a certain subject matter? Ta da – that’s exactly what you can do with this unique feature of the Product Manager Handbook!

In short, this is how it works:

1. A question is entered that is related to product management.
2. That question is passed to an AI engine along with all of the content from this website.
3. An answer is returned that includes the source(s) from which it was based.

Behind the scenes, there is bit more that happens, including logging questions that could not be answered and error handling.

The screenshots below show the user experience.

1. The feature appears at the top of the page in the header with placeholder text to explain what it does.

2. Clicking on the info icon will display a popover with more of an explanation and link to learn more (this page).

3. Enter a question and submit it (or hit the [enter] key).

The results are displayed:

4. The question that was submitted.
5. The answer generated by the AI engine.
6. The list of sources from which the answer was derived, that is linked to the respective page.
7. If the length of the answer falls within a range that considered to be a bit long or longer, options will be presented to try to shorten or lengthen the answer.

Other Things to Note

While it is presumed that the feature is used as expected and produces the desired results (AKA the “happy path”), edge cases were accounted for:

• If something was submitted that was not in the form of a question, the response will indicate such (with no answer). For example, sentences or fragmented words will not produce an answer.
• If a question cannot be answered from content in the website, the response will indicate that’s the case. All unanswerable questions are logged so that they may be reviewed and incorporated into plans for content updates.
• If an error occurs, which hopefully doesn’t happen, a user-friendly message is displayed and it is logged so that it can be analyzed and addressed as quickly as possible.

What Do You Think?

In the spirit of good product management, feedback is always welcome, of course!