E1. Conversing with AI Twins
This is where quantitative audience data becomes something you can actually interrogate. A Behavior, Intent, or Mentions report tells you what a pattern is. A conversation with the AI Twin built from that pattern lets you ask why it exists, pressure-test a message against it, or run something close to a focus group without waiting weeks to recruit one.
Starting a Session
Open the Converse icon to reach the conversation workspace. Ongoing Conversations holds your active research threads, so you can pick a session back up without losing momentum. Archived Conversations is the historical record, useful whenever you need to revisit a past finding or audit an earlier persona session. When you're ready for something new, start a fresh conversation with whichever Twins you've deployed.
If you're managing several segments at once, it's easy to forget which demographic or market definition sits behind a given persona name. The Segment/Twin toggle switches the view from creative persona names to the underlying segment definitions, so you can confirm you're talking to the right audience before you type a word. Click a profile to add it to the chat panel; you can add or remove multiple Twins at once to run a cross-segment conversation rather than a single-persona one.
Before you ask anything, click the Twin's chip to open its profile card. This shows the audience segment and data behind the persona, so you're grounding your questions in the same reality the Twin is drawing from, not a name alone.
Interview vs. Reflective
From your profile menu, you can set the conversation's framework:
Interview mode is built for broad category understanding and direct feedback. Responses come back succinct and objective, useful for quick benchmarks or straightforward reactions.
Reflective mode is built to surface the reasoning underneath a response, emotional triggers, motivations, lifestyle context, the kind of detail that explains not just what an audience prefers but why. The same question asked in both modes will often return genuinely different answers, not just a different tone, because Reflective mode is pulling on a different layer of the Twin's grounding.
Start from a suggested prompt if you want a fast, high-intent question, or type your own to steer the conversation exactly where your research needs it to go.
Validating a Response with Trace
A response is only as useful as your ability to check it, so every message carries a path back to its source. Open More Options on a response and select See Trace to view it.
Trace highlights the specific parts of the response tied to real data, each with a Score indicating how confident and statistically stable that particular data point is, plus a short line explaining the reasoning that connects the raw signal to what the Twin said. From there, View Report takes you out of the conversation and into the actual Behavior, Intent, or Mentions report the citation was pulled from, so a claim about, say, search intent traces directly back to an Intent report, and a claim about organic sentiment traces back to a Mentions report. Nothing in a Twin's response should be unaccounted for; Trace is how you confirm that.
It's worth comparing this directly against a general-purpose model. Compare with ChatGPT sends the same question to ChatGPT so you can see the difference side by side: an untraceable, generic answer next to one you can verify down to the data point. That gap is the entire argument for grounding research in observed behavior rather than an unaudited model's best guess.
Capturing What You Find
When a response captures your audience's voice precisely, whether it's a phrase, an insight, or a piece of reasoning, click it and select Save Nugget. Nuggets collect in a dedicated tab as your team's curated repository of validated consumer insight, ready to feed directly into copy and campaign workflows. You can also generate a shareable card from any response, useful for pitch decks or alignment meetings, or use the follow-up suggestion tool when you want the next logical question generated for you rather than written from scratch.
Running Research on Autopilot
Auto Mode hands the interview over to an autonomous research engine. Set how many follow-up questions deep you want it to go, define a Conversation Objective (uncovering conversion barriers, mapping the customer journey, testing messaging, or iterating on a concept), and start it. The system conducts a structured interview on its own, surfacing verified insight without needing a live researcher present for the whole session.