Your First Run

Get hands-on with existing work before you start producing your own.

Step 1: Do not create new segments yet

Segments are a means to an end, not the product itself. Building a good segment requires understanding the pipeline end to end. Until you have that understanding, use segments that already exist in the account.

The platform comes with pre-built segments (for example, a GEICO segment was set up for onboarding). Use those.

Step 2: Converse with an existing twin

Open the Converse section. You will see a list of AI Twins: ones you have had conversations with and ones you have not. Pick one that looks relevant to a brand or category you know.

Before you ask your first question, pick the right role:

• Brand representative (default when you are logged in under a portfolio): the twin knows you are asking on behalf of the brand. It will frame all answers through the brand lens. Use this when you want brand-specific consumer insight.
• Category researcher: the twin does not know who you represent. You are a neutral surveyor. Use this when you want unbiased, open-ended insight about the category, not just the brand. This mode mirrors how a real market researcher introduces themselves: “I’m conducting a survey on credit cards, can you help?”, not “I’m from Amex, what do you think of us?”

Ask as a neutral researcher when you want the most honest, unfiltered answers.

For more on roles, see Core Concepts.

Step 3: Run one or two surveys

Go to the research section and create a new research study. Choose an existing segment as your respondent base.

Start with a custom short survey: keep it to five or six questions with a clear objective. Do not paste strategy documents or long essays into the question field. The platform is a conversation tool, not a strategy-review engine. Ask the kind of questions a real surveyor would ask a real person.

You can also try a standard brand track survey; it has preset questions and will show you what a completed brand health output looks like.

Step 4: Run a digital study (creative test)

Go to the digital study section. Pick any brand that has a creative asset: a banner, a social post, a video thumbnail. Upload it and let the platform analyze it.

You will get:

• Feedback on the creative from the cohort’s perspective
• Action items: specific things to change or improve

Compare the variation of creatives and note what the platform flags. This is one of the most immediately useful outputs for marketing teams.

What good looks like

Do not be over-generic. If your segment is broad (18–55, US, all genders), the output will be broad. You will get results that sound like an LLM answered them, because at that level of generality the twin has nothing specific to draw on. Good output comes from well-scoped segments.

Ask as a neutral researcher. The twin will give you more honest answers when it does not know which brand you represent. Brand-context questions are useful for brand-specific insight; neutral-researcher questions are useful for category insight.

Read the emotional layer. When you converse with a twin, look at the emotional/reflection response as well as the rational response. The rational response tells you what a consumer says. The emotional response tells you what they actually feel, the part they would not share in a focus group.