From 300 answers to one page: inside an AqiloPulse readout
A walk through one AI survey summary, part by part: how each answer is read, how themes are counted, where quotes come from, and how it's translated.
Ashik Salman P
· 5 min read
An AqiloPulse readout is an AI survey summary your team can forward. Every number on it is counted from your responses, every quote is copied from an answer word for word, and the AqiloPulse engine writes only the wording around them. Here's how one comes together, part by part, using a sample pulse from a partner summit.
Sample answers written for this article; the analysis is the engine's unedited output.
What you open when the answers are in
Open a pulse's Insights tab and the summary sits at the top: a few sentences on what people said, the points they praised and the ones they didn't, what they asked for, suggested next steps, and the themes. Below it, the charts for every question, counted from every response.
For our sample, 30 people answered "What's the main reason for your score?" in English, Malayalam, Manglish, Hindi and Arabic. The summary opened:
Respondents praised the practical sessions, great speakers, and lovely venue, particularly highlighting the networking dinner and workshops. However, widespread frustration was directed at technical and logistical issues, most notably the failing registration process, long food queues, and unreliable wifi.
Step 1: every answer is read as it lands
Each written answer is read when it's submitted, not when you click "Generate". The engine notes what it's about (up to six topics), how the person felt about each one, whether it's praise, a complaint, a suggestion or a question, and how urgent it sounds. "Practical sessions kurachu koodi venam, theory mathram aanu kooduthal" is read as a request for more practical sessions, not as noise, because it's read in the language it was typed.
Answers to email and phone questions are never sent to the engine, and very short answers like "ok" are skipped rather than guessed at.
Step 2: themes, counted from your data
Topics that mean the same subject become one theme: "registration" and "signup" in one, a conference "track" with the sessions. The count beside each theme is how many responses raised it, counted in the database, so it's the same every time you look.
| Theme | People | Feeling |
|---|---|---|
| Registration | 8 | negative |
| Sessions | 7 | positive |
| Food | 4 | positive |
| Workshops | 3 | positive |
| Wifi | 2 | negative |
The biggest themes get a card each. Smaller ones, such as parking and networking in this sample, sit in a row underneath, labelled "Also mentioned", still counted and a tap away, so a long tail of one-off subjects doesn't bury the main ones.
Step 3: quotes, copied, in the language they were written
Open a theme and you see answers from the people who raised it, exactly as they wrote them. Under Wifi:
Wifi kittiyilla, slides download cheyyan pattiyilla.
That's Manglish: Malayalam in English letters, "no wifi, couldn't download the slides." The quote is labelled with the language it was written in and is never rewritten or translated, because the tone is part of the finding.
Step 4: what's working, what's not, what they asked for
The engine also lists the points themselves, grouped by how many people made them:
- What's not: "The registration form kept freezing, timed out, and required multiple attempts" (some) · "Food and lunch queues were too long" (some) · "The wifi did not work in the main hall, preventing slide downloads" (some)
- What they asked for: "Want more of those workshops" · "Fewer panels"
- Suggested next steps: "Upgrade the registration system to handle load without timing out or throwing errors" first, then the wifi, the air conditioning and the lunch queue.
"Some" and "few" are worked out from how many responses made each point, not chosen by the model, so a point one person made is never presented as what most people think.
Step 5: the same page in your team's language
If your team reads Malayalam, Hindi or Arabic better than English, the summary is translated into the languages you choose. Here's the opening line for a Malayalam-reading team:
പ്രാക്ടിക്കൽ സെഷനുകൾ, മികച്ച പ്രഭാഷകർ, മനോഹരമായ വേദി എന്നിവയെ പ്രതികരിച്ചവർ പ്രശംസിച്ചു…
| On the page | Translated? |
|---|---|
| Summary, points, next steps, theme names | yes |
| Counts and percentages | never: they're numbers from your data |
| Quotes | never: they stay in the respondent's own words |
When the readout is up to date
Because every answer is read as it lands, the themes are ready the moment you generate the summary, with no waiting for a batch to run overnight. If more answers arrive afterwards, the summary tells you how many came in since, and Regenerate writes a new one over the full set. Earlier versions are kept on record rather than overwritten.
Two things are always current regardless: the charts below the summary, and the count beside each answer list, which are worked out from the responses each time you open the page. Only the wording waits for you to regenerate, since rewriting it on every new answer would change the text under people's feet while they read it.
A readout also tells you what it couldn't read. If a few answers were too short to mean anything ("ok", "nil"), the page says how many, rather than quietly leaving them out of the counts.
What the readout won't do
- Invent numbers. A figure the data didn't contain is treated as an error, and the summary is written again.
- Rewrite a quote. Quotes are copied from answers, never paraphrased.
- Read contact details. Email and phone answers never reach the engine.
Frequently asked questions
Does the AqiloPulse engine make anything up?
It writes the wording, so read it as a summary, but every count is counted from your responses and every quote is copied from one. You can open any theme to see the answers behind it.
Which plans include the readout?
Every plan includes it: per-answer reading and the summary, on Starter, Pro and Scale.
Can I export it?
You can export every response, with its answers, to CSV for Excel or Sheets from the Downloads tab.
Related reading: how to analyze open-ended survey responses and Manglish, Hinglish and the comments translate-first tools miss.