Runa Intelligence is a social behaviour company developing behavioural intelligence for Latin America. We build synthetic agents grounded in real data, and test how they read, react to and act on your decisions, so you can rehearse them before they are made.
The effectiveness of a message depends on how different audiences interpret it. Knowledge, beliefs, experiences, and context all influence that interpretation, creating a gap between intended meaning and actual reception.
Traditional audience research can sometimes be time-consuming and costly, making it difficult to evaluate every decision before execution.
Runa bridges this gap by bringing audience intelligence into your everyday workflow. Through a dedicated digital workspace, teams gain on-demand access to synthetic populations, allowing them to test messages, explore audience interpretations, and make more informed decisions before execution.
Test a message on a simulated population and measure the gap between what you meant and what people took away.
Measure how much a message, or a sequence of messages, moves a variable such as candidate approval over the following weeks.
Four steps. A clear method you can repeat and stand behind.
Weeks per cycle and a new cost each time, so testing tends to happen once — if at all — and rarely after the message has changed.
Minutes per cycle against a set of agents that stays constant, so a rewrite can be compared with the version before it. Same agents, same baseline, same method: every result is reproducible.
The ambition is not to replace the research you already trust. Runa works the way good pre-testing always has, asking the audience before you commit, only faster, repeatable and grounded in a model of the population that is calibrated against observed outcomes. Fieldwork stays for the questions that need it.
We state the limits of the method plainly, and we keep testing it against reality. See how on the technology page.
Wherever a communication has to produce a specific response from a specific audience, it can be rehearsed first — so the decision is made with confidence, not hope.
Which proposition produces the response you want from the customers you want it from.
Whether an announcement is read the way it was written — before it is published.
Which response to a fast-moving situation lands best with each audience you have to hold.
Who reads a policy as protection and who reads it as a cut.
Which framing moves people to act — and which segments it fails to reach.
Which message, framing and call to action to commit resources behind.
Runa does not prompt a language model to imagine an audience. Every simulation starts from structured evidence about the real population, organised through ARDA, our knowledge infrastructure of people, places, institutions and the relationships between them. ARDA is being developed around Latin American populations, beginning with Argentina, so the evidence underneath each agent belongs to the market you are addressing. Every attribute carries its source, so you can see exactly what a result rests on.
Explore the technology →The intelligence underneath Runa is being developed around the demographic, socioeconomic, geographic, political, cultural and behavioural contexts of Latin American populations, not adapted to them afterwards.
Where ARDA and the products are being built, tested and validated first, on a population with rich public statistics and strong regional variation.
Each additional market receives its own evidence base and population model on the same architecture. The region is not one population, and we do not model it as one.
Agents are built from structured evidence about a real population, not from the assumptions inside a prompt. You can trace every reading back to its source.
The agents are held constant, so every rewrite is compared against the same baseline. Run it twice, get the same answer.
Calibration against observed outcomes is part of the architecture, and every result ships with its range and its blind spots — so you know how far to lean on it.
