For AI agents
Give your agents the ability to run their own studies.
Create a population, put proposals to it, dig deeper into the answers and compare several options: build synthetic population research into your agents’ work.
How a study unfolds
Start from the mission
Your agent frames the question to explore: understanding how something is received, comparing proposals or testing a hypothesis. It sets the constraints and what it will look for in the answers.
- What your agent prepares
- The question, the constraints and what to look for.
- What it gets from Imagine All The People
- Nothing at this stage: framing stays with your system.
Build the population
It creates the population relevant to the study, based on the parameters and reference data available in your configuration.
Not every population can be represented with the same reliability: it depends on the reference data available. Some characteristics are calibrated on that data; others are modelling assumptions.
- What your agent prepares
- A description of the target population and its parameters.
- What it gets from Imagine All The People
- A synthetic population built from the available reference data.
Submit questions and scenarios
The agent presents its questions or proposals. It can organise a comparison of several options and keep their context.
Supported formats depend on the capabilities exposed in your configuration.
- What your agent prepares
- The questions, the proposals and how the comparison is organised.
- What it gets from Imagine All The People
- The population’s answers to each question or proposal.
Review the answers
Synthetic answers become material to analyse: understanding, objections, stated conditions and differences between groups.
Individual answers, aggregation and interpretation are three distinct levels. The interpretation is produced by your system.
- What your agent prepares
- Analysing and interpreting the answers in light of the mission.
- What it gets from Imagine All The People
- Individual answers and their aggregation by group.
Dig deeper
An answer is ambiguous? A divergence is worth examining? Your agent can continue its investigation with follow-up questions.
How follow-ups are organised depends on the capabilities exposed and on your system’s orchestration.
- What your agent prepares
- The follow-up questions and the decision to ask them.
- What it gets from Imagine All The People
- The answers to those new questions.
Use the results
Your agent resumes its work with what it has gathered: comparing an option, reworking a wording, preparing a recommendation or identifying a question that is still open.
The final recommendation remains your system’s, not Imagine All The People’s.
- What your agent prepares
- The rest of the work and the final recommendation.
- What it gets from Imagine All The People
- The material gathered during the study, which your system builds on.
Two ways to organise the questioning
A study defined in advance
Your system prepares the questions and scenarios, runs the questioning and then uses the results.
A progressive investigation
Your system adapts the next questions to what it discovers, within the limits of the authorised mission and usage.
Two ways of organising the work, not two separate plans.
Examine a run
An execution dossier will make it possible to follow a real study end to end. It will present:
- the brief;
- the population and its reference data;
- the questions and scenarios;
- the calls actually made;
- the answers;
- how they were used in the rest of the work;
- the deliverable and its limitations.
Illustrative outline. No execution dossier has been published yet.
An answer, a result and a conclusion are not the same thing.
An individual answer keeps its context. A comparison brings answers together according to a defined indicator. A recommendation interprets these elements in light of the mission.
Your system needs to preserve these distinctions when it uses the results.
The answers are synthetic. Whether they match real human behaviour is assessed separately.
Understand validation →