Scout Agent

Scout is atypica.AI's social-media observation module. You give it a topic or audience, it reads posts from public social platforms and builds a profile of who these people are — their lifestyle, values, identity, and the language they use to express themselves.

The framing Scout works under: like an anthropologist doing fieldwork, not a data analyst doing statistical research. Scout reasons about why people think and speak the way they do, not how many of them posted.

How Scout works

Scout runs in three phases. You don't manage them directly — they happen in order during a single Scout run.

1. Observation. Scout reads posts across the supported platforms. It looks for content that feels interesting, resonant, or conflicting, and pays close attention to tone, wording, and rhetorical style. This phase is deliberately unhurried — Scout is looking for patterns, not collecting data.

2. Synthesis. Once Scout has enough observation, it forms hypotheses about the audience. These users care most about X, not Y. They use Z framing because... This step is where Scout differs from a social-listening dashboard — it doesn't just count mentions.

3. Validation. Scout goes back to social media to test its hypotheses. If "core motivation is X" holds up, the profile strengthens. If not, Scout revises.

A full Scout run typically takes about 15 minutes of wall time, depending on topic breadth and how much validation the synthesis demands.

What Scout pays attention to

Scout watches for four dimensions specifically:

Lifestyle — how people describe their daily lives, what tone they use, what values surface in shared moments.
Social interaction — how they interact with others, how they respond to different viewpoints, how they define "us" and "them."
Values and identity — value priorities in consumption, ideal-vs-reality tradeoffs, identity labels they embrace or resist.
Expression and language — vocabulary, sentence patterns, slang, memes, emojis, and how style shifts by context.

These aren't data dimensions — they're lenses. Scout reasons about the meaning of what it sees through these lenses.

Supported platforms

Scout reads publicly visible posts on:

Xiaohongshu
Douyin
TikTok
X / Twitter
Instagram

Private accounts, gated groups, and paywalled content are not accessible.

What Scout produces

The output is a set of AI personas built around a 7-dimension framework (demographics, geography, psychology, behavior, needs & pain points, tech acceptance, social relations). These personas can be used immediately in Interview Chat, Discussion Chat, or any other research that needs a persona pool.

A typical Scout run gives you 3–10 personas depending on how much variation the audience shows.

When Scout is the right tool

Scout shines when:

You're entering a market you don't know well.
You want fresh trends, not a stale industry report.
You're repositioning a brand and need to see how people actually talk about it.
You want to build personas without paying for primary research first.

Skip Scout when:

You already have quant data on this audience and need to fill qualitative gaps — go straight to Interview.
Your target audience doesn't post publicly on the supported platforms.
You need statistically significant numbers — Scout is qualitative.

Limits

Scout is qualitative. It does not produce statistically significant numbers.
A profile from this month may not match next quarter — social-media trends are volatile.
Scout reasons, but it's still an LLM. Treat its conclusions as hypotheses to validate, not findings to ship to the CEO.
Some platforms have weak or no public data for certain demographics. Don't trust a profile built on 12 posts — ask Scout to broaden or pick a different topic.
Scout can't tell you why something is missing. If the audience isn't on the supported platforms, you won't see them.

A worked example

You ask Scout: "Understand young Chinese women who drink specialty coffee."

Scout reads Xiaohongshu and Douyin posts, observes recurring patterns, forms a hypothesis (this group cares more about aesthetic experience and shareability than caffeine function), validates by looking for counter-evidence, and produces a set of personas — say, 4 personas ranging from "the aesthetic café-goer who photographs everything" to "the practical home-brewer who doesn't post."

You can now use those personas directly in a Discussion Chat about a new coffee product concept.