> ## Documentation Index
> Fetch the complete documentation index at: https://agencykey.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Recon

> Cross-creator AI briefing with clear sharing scope, limits, and safety guidance

**AI Recon** builds a tactical fan briefing from eligible conversations shared by other models. Open the **AI Recon** section in the right-side **Insights** panel of an active chat.

It is intentionally different from [Chat Summary](./chat-summary): Chat Summary covers only the current model's conversation, while AI Recon can use conversations from several models within the selected sharing scope.

## Before you generate

AI Recon is available only when all of these conditions are met:

* your role has **View AI Recon** for the selected model;
* the selected model's **AI Recon Sharing** is not **Don't share**;
* at least one eligible shared model conversation contains **20 or more parsed messages** for this fan;
* to create or update a briefing, your role also has **Generate AI Recon**.

If sharing is disabled, the section offers a shortcut to the model's settings when you are allowed to open them.

## Sharing scope

The selected model's effective sharing level controls both what it contributes and what it can receive.

| Level             | Data AI Recon can use                                                   |
| ----------------- | ----------------------------------------------------------------------- |
| **Don't share**   | Nothing is shared or received; AI Recon is unavailable for this model   |
| **Within agency** | Eligible participating models in the same agency                        |
| **Entire CRM**    | Eligible participating models across agencies on the AgencyKey platform |

Participation is reciprocal: a source is included only when its sharing setting is compatible with the selected model's scope. New agencies default to **Entire CRM**, but an agency default or a model-specific override can change the effective level.

See [Data Sharing](../accounts/data-sharing) for changing these settings.

<Note>
  AI Recon does not identify the source model or agency and does not give you the source conversation. The briefing can still repeat personal details or short phrases the fan stated, so treat it as sensitive, source-deidentified information rather than anonymous data.
</Note>

## Analysis limits

One generation can analyze:

* up to **10 eligible models**;
* up to **50,000 messages per model**.

If more sources are eligible, one result still uses no more than 10 models. Very long conversations are capped at 50,000 messages per model.

The same fan can therefore produce a different briefing when opened under another model with a different sharing scope.

## Generate and regenerate

1. Open the fan's chat under the correct model.
2. In the right panel, open **Insights** and expand **AI Recon**.
3. Click **Run recon**.
4. Read the result and verify important claims against the live chat.

When AgencyKey marks a saved result as stale — for example, after the eligible sharing pool changes — it stops showing the stale content and prompts you to **Run recon** again with the current eligible sources. An up-to-date result cannot be regenerated repeatedly on demand.

An existing result is restored for authorized teammates in the same AgencyKey workspace. It remains tied to the selected model's effective sharing scope; do not assume every model or agency sees the same version.

## What a briefing may include

Depending on the available conversation history, AI Recon can organize findings such as:

* personal details the fan stated;
* communication style and preferred tone;
* interests and content preferences;
* approaches associated with purchases;
* approaches the fan rejected or ignored;
* phrases that may indicate buying intent.

Sections without enough supporting conversation are omitted.

<Warning>
  AI Recon is AI-generated. It can miss context, confuse people or time periods, overstate patterns, and misinterpret jokes or role-play. Treat personal and psychological details as hypotheses, not confirmed facts. Verify important claims against the live chat, notes, and payment history before acting on them.
</Warning>

## Language and translation

AI Recon is generated in English. To translate it:

1. Open the chat's **Translation settings**.
2. Choose **Translate chat to**.
3. Enable **Translate AI fan profile**. You can also enable **Translate recon** from the gear in the AI Recon section; both control the same setting.

Changing ordinary incoming-message translation alone does not enable AI Recon translation.

## Permissions

| Action                                         | Permission                        |
| ---------------------------------------------- | --------------------------------- |
| Show the section and read an existing briefing | **View AI Recon**                 |
| Generate or regenerate the briefing            | **Generate AI Recon**             |
| Change a model's sharing level                 | Account/sharing management access |

Permissions are assigned independently of job title. If the section or button is missing, ask the Owner or someone who can manage roles and creator access to check your permission for the selected model.

## Troubleshooting

* **AI Recon is disabled for this model** — Change **AI Recon Sharing** from **Don't share** to **Within agency** or **Entire CRM**.
* **Not enough data for cross-creator analysis** — No eligible shared model conversation has reached 20 parsed messages for this fan yet.
* **Scope changed** — Wait for the sharing update to finish, then click **Run recon** again.
* **Run recon is missing** — You may be able to view results but not generate them.
* **The result looks thin or inaccurate** — Check the sharing scope and verify the underlying chat. A current result has no on-demand regenerate action; **Run recon** appears again if AgencyKey marks it stale.