We've put together a starting guide for you to get the best results on Ask AI. You'll find a list of prompts below. This is not an exhaustive list. We encourage you to explore Ask AI and tailor it for your research.
Here are a few points to remember while writing an Ask AI prompt:
1. To find specific insights, filter by projects, file tags, and labels
For example, you may want to collate feedback about a specific feature in your product. Such as "What do users think of Quant Analysis?".
Marvin can identify recordings and documents where "Quant Analysis" is clearly specified in conversations. However, it may not have enough context to identify files that mention Quant Analysis when it is described in generic terms, such as "that feature" or "the most-used feature".
Instead of using a prompt to surface all the information on a topic, we recommend that you filter by project, file tags, and labels. Marvin can then search within specified files to surface all the insights.
2. You can specify the format in which you want to read the answer
Marvin can arrange information as a list, table, quote, podcast script, and even a song!
3. You can ask follow-up questions
If you need to refine the initial answer, just ask a follow-up question. You can ask Marvin to summarize, shorten to a specific word limit, dive deeper, etc. Learn more in our help center article.
Detect themes and trends
Ask AI is a great way to determine user sentiment or find ideas for feature improvements. For example, you can search for, ‘What are five queries we receive most often in our support tickets?’
You're likely to get targeted and accurate results if you ask Marvin to consider a specific data set. Build more context by using filters for:
Projects (select one or more projects)
File tags (search among files that have tags)
Research areas (categories into which you've classified published reports)
Research panel data (filter for cohorts, revenue, company size, etc.)
Time period (find data for a specific month, year, quarter, or custom time range.
Recency or relevance (prioritize recent data or the most relevant insights across all-time data)
File type (files, Insights, notes, support tickets, surveys)
Participants (surface insights only from participants or interviewers)
Choose as many filters simultaneously to zero in on the right information.
Find people
You can ask Marvin to find people who gave specific feedback or mentioned a certain kind of detail on the call.
For example, "Find cusotmers who requested the new AI analysis".
Marvin will show you the name of speakers, what they said, and which call they said it on.
Compare two things
Marvin can distinguish between two topics and show you a table to easily compare them. For example, the difference between labelling and tagging data.
You can verify the data as Marvin lists all the sources it used to write this answer. Each numbered circle will lead you to a source in your repository. It’s a good practice to verify all AI-generated answers. Read more about this in our help center article.
You can also rate the answer to let us know how Marvin’s AI did. We use your feedback to make Marvin better.
Find quotes
This is a handy feature when you want to build out the reports showcasing your research. Hearing it straight from the users as clips, or a string of clips organized into a playlist is a great way to back up your research hypothesis.
You’ll notice shortcuts at the end of the answer to add the quote to an Insight report and copy the text. You can also ask Marvin to regenerate the answer if you are not satisfied with it.
*We recommend that you always check the accuracy of all AI-generated results before sharing the information or acting upon it.