CatalyseAI · Ask the study (AI analyst)
Ask your results a question. Get the why.
Type a question about your survey in plain English, such as which group liked an idea most and why. The AI analyst works it out from your study's own data and answers with the numbers, a chart and the quotes behind it.
- Plain English
- Report-engine numbers
- Real quotes
- Number check
The next question
Charts tell you what. You still need the why.
A report shows every question one at a time. The questions your team asks next cut across them: why one group answered differently, what the people who disliked the price wrote, whether a gap is big enough to act on. Answering those usually means building crosstabs by hand and reading open answers one by one. Ask the study does that work when you ask, and shows how it got there.
Why did one group answer differently?
What did the people who disliked the price write?
Is that gap big enough to act on?
Which plan do weekly cooks prefer?
Ask the studyOne question, five steps
What happens when you ask
- 01
Ask in your own words
Open the AI insights tab of your report and type a question. Filters you have set on the report page are taken into account.
- 02
It plans the steps
The analyst works out what it needs: a group to define, such as people aged 18 to 24, a crosstab, the share for each answer, or open answers to read.
- 03
The report engine does the math
Every number is calculated by the same code that draws your report, with the same weighting. The AI explains the result; it does not do the arithmetic.
- 04
You get an answer you can check
Charts or tables, quotes copied word for word from respondents, and a check that lists any number or quote it could not trace back to your data.
- 05
Keep the conversation going
Follow up in the same chat, for example “now split that by gender”. Chats are saved with the survey, so your team can read them later and see who asked.
Which plan do people under 35 prefer, and is the gap with older subscribers real?
Report filters: none
Your words, plus any filters already set on the report.
Which plan do people under 35 prefer, and is the gap with older subscribers real?
- Define group “under 35” from Q2 (age)
- Define group “35 and over”
- Crosstab: preferred plan by the two groups
- Look up open answers about the plan choice
Which plan do people under 35 prefer, and is the gap with older subscribers real?
| Under 35 A | 35 and over B | |
|---|---|---|
| Classic | 27% | 38%A |
| Veggie | 44%B | 29% |
| Family | 29% | 33% |
| People | 212 | 388 |
The AI explains the result. It does not do the arithmetic.
Which plan do people under 35 prefer, and is the gap with older subscribers real?
The Veggie plan: 44% of under-35s prefer it, against 29% of those 35 and over. The gap is significant at the 95% level.
“Cheaper than the Family box, and I don’t cook meat at home.”
Number check: 4 numbers and 1 quote traced to your data
Now split that by gender
Under 35, the Veggie plan leads with women (49%) and men (39%). Based on 118 women and 94 men.
- Plans by ageAsked by you, saved with the survey
- Delivery commentsAsked by Sam on Monday
Ask it for
What you can ask it for
- 01
Groups defined on the spot
“Gen Z”, “people who chose Pack B”, “weekly buyers”: it builds the group from your questions and tells you how many people are in it.
Make three groups: Gen Z, people who chose the Veggie plan, and weekly cooks.
- “Gen Z”Aged 18 to 29, from Q2131 people
- “chose the Veggie plan”Answered Veggie at Q4205 people
- “weekly cooks”Cook at home 4 or more times a week, Q6248 people
- 02
Crosstabs and shares
Any question by another question or by a group, with significance testing from the report's crosstabs.
Plan choice by age group. Is the gap real?
Under 35 A 35+ B Classic 27% 38%A Veggie 44%B 29% Family 29% 33% Letters mark significant differences
- 03
Charts that fit the data
Bars, stacked bars, pies, tables and more, drawn in the report's own style. Ask for a chart that doesn't suit the data and it picks one that does.
Show plan choice by age as a line chart.
A line chart is for change over time. Shares by group read better as grouped bars.
- 04
Quotes from open answers
It searches typed answers and the transcripts of video and audio answers, and shows matching quotes as cards.
What did people say about delivery?
“Delivery came a day late, twice.”
“The ice packs were warm when the box arrived.”
- 05
Fieldwork facts
Completes, drop-outs and time taken, for everyone or for one group.
How many people finished each day?
Day 1Day 2Day 3Day 4Day 5Completes 600Drop-outs 7%
- 06
A plain “can't answer”
If your study can't answer the question, it says so and tells you what it can answer instead.
What profit will the Veggie plan make?
This survey can’t answer that: it has no cost or sales data. I can show which plan each group prefers.
Checks
Checks behind every answer
- Same numbers as the report
Shares, crosstabs and weights come from the report's own calculations, so the chat and the report agree.
- Number and quote check
Any number or quote in the answer that can't be traced to a result in the chat is listed under the answer.
- Group sizes shown
It tells you how many people each result is based on, so a big-sounding share from a small group is easy to spot.
- Your study only
The analyst reads only the survey you asked from. It cannot see your other surveys or anyone else's account.
Which plan do people under 35 prefer?
- Veggie, under 3544%base 212, traced
- Veggie, 35 and over29%base 388, traced
- Significance95%from the crosstab
- Quote1copied word for word, traced
Nothing untraced
Shown under every answer
Example
Example: four questions for a meal-kit survey
Asked of a survey of 600 US meal-kit subscribers: “Which plan do people under 35 prefer, and is the gap with older subscribers real?” “What do people who rated delivery 1 or 2 out of 5 say in their comments?” “Show the share who would recommend us by how long they have subscribed, as a chart.” “How many people finished the survey each day?” For each one, the analyst names the groups it used, shows a chart or table, and adds quotes where they help.
Which plan do people under 35 prefer, and is the gap real?
Veggie plan: 44% against 29%, significant
What do people who rated delivery 1 or 2 out of 5 say?
Show the share who would recommend us by how long they have subscribed.
How many people finished the survey each day?
FAQ
Questions, answered
Can it make things up?
Every number comes from the report engine, not from the AI's memory. A check under each answer lists anything it can't trace back to your data. If the data can't answer the question, it says so.
Will the chat's numbers match my report?
Yes. It uses the same calculations and weighting as the report page, so a share in the chat is the same share you see on the card.
Who can see the chats?
Chats are saved with the survey. Anyone in your account who can open that survey can read and continue them, and each question shows who asked it.
Which AI does it use, and what does it see?
It runs on Anthropic's Claude Opus model. To answer, it sees your survey's questions, the calculated results and the open answers it looks up for that question. Ask us for the data processing details.
Related
Goes well with
Bring the question you keep meaning to ask your data
We will show you the analyst working on a sample study, and talk through how it would read yours.
Nearest office: Orlando, FL (corporate office)