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

# Synthetic Customer

> Talk to any customer as an AI persona grounded in real data: call transcripts, feedback, CRM profile, and public research. Every reply names its source.

Talk to any customer or segment as an AI persona grounded in real data. The persona is built from their actual call transcripts, product feedback, CRM profile, and public research, not made up. Every response includes source attribution so you know what's fact vs. interpretation.

## Usage

```
/closedloop-synthetic-customer Acme Corp
/closedloop-synthetic-customer Enterprise customers
/closedloop-synthetic-customer frustrated customers in Payments
```

Type a company name, segment, or industry: the skill figures out what you mean and builds the persona.

## Modes

### Company Mode

Type a company name. The skill loads all their data (call transcripts, product feedback, support tickets, CRM profile) and role-plays as that company. Multiple speakers are blended into one voice that represents the company's collective perspective.

### Segment Mode

Type a segment or industry name. The skill creates a composite persona from all companies in that group. Uses "we" language and gives counts ("8 of us have raised this"). Verbatim quotes are attributed to specific companies.

## How It Works

The persona is built in three layers, using multiple parallel agents:

**Layer 1: External Research**

* Company website, news, industry context, competitors
* CRM data: deals, contacts, revenue, segment
* Public voice: how they present themselves, recent announcements

**Layer 2: Voice Extraction**

* Reads ALL their call transcripts (not summaries, the actual conversations)
* Extracts speech patterns, recurring themes, emotional moments, questions they ask, how they make decisions
* Analyzes support tickets for what breaks, how they report issues, their patience level
* Captures verbatim quotes that define their voice

**Layer 3: Synthesis**

* Merges all research into a persona file
* Verifies every claim is grounded in data
* Builds a knowledge boundary: what they have and haven't discussed

The persona file is saved locally and reused. Next time you invoke the same customer, it loads instantly and only updates for new data.

## During the Conversation

**Source attribution on every response.** Every in-character response ends with:

```
---
Sources: what data grounded this response
Interpretation: anything inferred rather than directly stated
```

**Real-time data lookup.** For specific factual questions, the persona searches live MCP data rather than relying only on the persona file summary.

**Three-tier responses:**

* **Direct evidence**: speaks confidently, uses verbatim quotes
* **Adjacent evidence**: qualifies with "I think" or "probably"
* **No evidence**: declines naturally: "That hasn't really come up for us."

## Commands

During a persona conversation, you can use these commands to inspect the data or switch personas. Invoke them the same way you started the skill:

```
/closedloop-synthetic-customer /sources
```

| Command                                            | What it does                                       |
| -------------------------------------------------- | -------------------------------------------------- |
| `/closedloop-synthetic-customer /sources`          | Show detailed data points behind the last response |
| `/closedloop-synthetic-customer /confidence`       | Rate evidence strength on the current topic        |
| `/closedloop-synthetic-customer /raw`              | Show the relevant section of the persona file      |
| `/closedloop-synthetic-customer /switch Acme Corp` | Drop this persona, build a new one                 |
| `/closedloop-synthetic-customer /refresh`          | Force rebuild the persona from scratch             |

To exit the conversation, type **"stop"** and the persona will end immediately.

## Use Cases

**Validate a feature idea**

Pitch a feature to the synthetic customer and get pushback grounded in what they've actually said. "What if we added dynamic pricing with AI-driven rules?" The persona responds based on their real conversations about pricing, not generic LLM opinions.

**Prepare for a customer call**

Before meeting a customer, talk to their synthetic version first. Understand their priorities, pain points, and how they communicate so you walk in prepared.

**Test sales messaging**

Try different positioning with a synthetic customer to see which resonates. The persona's reactions are grounded in their actual decision criteria and competitive context.

**Explore a segment's perspective**

"What do Enterprise customers think about our API?" Get an answer that aggregates real feedback from all Enterprise accounts, not one company's view.

**Understand internal dynamics**

Ask about their team structure, decision-making, and internal tensions. The persona knows who the decision-maker is, who the power user is, and what frustrates each role, all from actual call transcripts.