Skip to content
quisquilia, home
Versione italiana: ITEN
Try it an interview, in Italian (opens in a new tab)

The Quisquilia method, told by Michele

Example · made-up data

One interview.

Michele talks with a maintenance technician. Four stages, twenty minutes, by voice.

×12 Twelve people, three interviewers.

Carlo, Michele and Catherine run interviews in parallel. Three styles, one method.

×120 The whole plant in four days.

Everyone picks their moment. Shift A in the morning, the night shift on a break, the warehouse from home.

Six agents, one chain.

Every sentence is transcribed, anonymised and coded twice. Similar sentences move closer together.

The tensions come out.

The reviewer checks every quote. Then a person looks at where the stories don’t add up.

The reading.

Results first, then the data. Every number has its sentences behind it.

All very nice, but how does it work?

Michele walks you through it in under three minutes. He speaks Italian, with English subtitles.

  1. One interview. Michele asks a maintenance technician about yesterday’s work, the last time everything stopped, what only he knows how to do and how he would teach it.
  2. Twelve people, three interviewers. Carlo, Michele and Catherine run interviews in parallel with the same method.
  3. A hundred and twenty people in four days. Each one picks their moment.
  4. Six agents in a chain transcribe, anonymise and double-code 1,562 sentences. Similar sentences form six skill areas.
  5. The reviewer checks 214 quotes and discards 3. Three tensions come out. Skills held by one person, a manual that says something else, new hires who learn on their own.
  6. The reading. The know-how is there but it sits with a few people. Three actions to discuss together.

Under the hood.

From the goal to the analysis, twelve specialised agents hand over work and check each other. Three steps need a person. Follow the process here with illustrative data.

workflow · run 041214 people · 12 agents · 3 human checkpointsExample · made-up data

Interview

personGoal with youbrief.goalwaiting
agentDesignerdesign.promptidle
agentSimulatorsim.personasidle
agentCarlovoice.liveidle
agentMichelevoice.liveidle
agentCatherinevoice.liveidle
agentNotespost_call.extractidle
dataLogtranscript.jsonempty

Analysis

agentPseudonymspii.redactidle
agentCodercode.multiidle
personHuman checkpointHALTwaiting
agentExtractorextract.taskidle
agentTheme analystqual.themesidle
agentNumbers analystquant.aggidle
agentReviewerqa.citeidle
agentWriterwrite.reportidle
agentChartsrender.vizidle
personReading with youreviewwaiting

In plain words. It starts from a goal agreed with you.

Zoom in on one turn. About a second.

It listens, transcribes and checks the answer against the goal. Then it chooses a follow-up, without judging, and answers in its own voice. If you interrupt, it stops and listens.

One turnlive

  1. Listeningvad
  2. Transcriptionasr
  3. End of turneot
  4. Directionpolicy
  5. Replyllm
  6. Voicetts
  7. Logwebhook

From your pause to its voice812 ms

session 7f3a·e21Michele · team leader, night shiftExample · made-up data

confidential. Prompts, weights and criteria are our craft.Everything else is right here.

Behind every conversation there is a direction.

Years of field interviews, human intelligence techniques (HUMINT) and knowledge of AI agents: these experiences shape our approach to listening.

  • Goal first, then questions.

    Every interview starts from a stated goal. What we want to understand and with which framework. We choose who runs it before deciding what to ask.

  • An interviewer designed like a real one.

    Personality, style, goals, quality markers for the conversation. Plus the context of the case, to draw on while the conversation happens.

  • Adaptive, person by person.

    Indirect questions, follow-ups that build on the answer. It slows down, digs deeper, changes course when the story opens up and offers scenarios instead of direct questions.

  • The method is built into the platform.

    Interview number 500 has the same quality as the first. A reviewer checks every interview.

Carlo AI interviewer

Listening · 04:12

Example · made-up data
  1. Tell me about your last night in the warehouse. Where did it start?
  2. With a round of the cold rooms. I check the temperatures by hand even though there’s an alarm.
  3. The last time the alarm went off, what happened?
  4. It went off late. I had already moved the stock out of cold room three myself.

In the reportManual cold room checks are not in any procedure.

Human direction from start to finish.

No technical skills needed, from you or from the people interviewed.

  1. We define it together.

    What you want to understand, from whom and what shape the result should take.

  2. We design the interview.

    We build the interviewer, the scenarios and the quality criteria around your case. This is the part that sets the quality of everything else.

  3. The interviewers listen to everyone.

    Voice or written interviews, in parallel. Each person gets a link and takes part when it suits them. On a lunch break, in the car, at home. You get updates while the interviews run.

  4. We read the results with you.

    Reports, maps and dashboards, discussed together and not just sent over. No output reaches you without an expert reading. The setup stays, so next time you start ready.

Three styles, not three avatars.

The same method and the same kind of result, with three ways to open, explore and close a conversation. Choose the style that suits the people being interviewed.

  • Carlo Zanella, AI interviewer

    Carlo Zanella Practical

    Quick and polite, he gets to the point. Simple questions that go straight to the facts. For people who prefer to describe how they work and watch the result take shape.

  • Michele Picella, AI interviewer

    Michele Picella Informal

    Curious and direct, the kind of person you enjoy talking to. On first-name terms. For people who prefer a chat to a formal interview.

  • Catherine Marshall, AI interviewer

    Catherine Marshall Challenging

    Direct and fast. She pushes back on generic answers and tests the story. The tone stays fair, never aggressive. She focuses on the points where the story drifts from the facts. For people who are happy to be challenged, even when the question gets uncomfortable.

Made to measure. If none of the three fits, we design an interviewer for your case. Goal, tone, context and the shape of the output.

Relax. We take care of the legal side.

We comply with GDPR and the EU AI Act, and we put it in writing before we start.

All about data and compliance
See what comes outReports, maps, dashboards

Rather try it yourself? Try an interview (in Italian) (opens in a new tab)