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

# Building Tips & Techniques

> A grab-bag of small, copyable techniques for building, refining, and remixing roleplay scenarios

Small, practical techniques for common building moments. Find the one that matches what you're trying to do, copy the example, and adapt it to your scenario. For ready-to-paste agent prompts, see the [Prompt Library](/prompt-library). For full walkthroughs, see [Creating Scenarios](/roleplays/ai-agent) and [Advanced Roleplay Building](/roleplays/advanced-building).

| I want to...                   | Jump to                                                                                                                    |
| :----------------------------- | :------------------------------------------------------------------------------------------------------------------------- |
| Make a character sound human   | [Natural speech](#natural-speech-patterns), [Response length](#control-response-length)                                    |
| Fix how a word is said         | [Pronunciation rules](#pronunciation-rules)                                                                                |
| Stop the character over-asking | [Don't ask superfluous questions](#dont-ask-superfluous-questions)                                                         |
| Keep reps from getting stuck   | [Exit conditions](#define-exit-conditions), [Concession points](#define-a-concession-point)                                |
| Make discovery harder          | [Multi-level depth](#add-multi-level-discovery-depth)                                                                      |
| Write better grading           | [Observable chains](#write-criteria-as-observable-chains), [Scenario-specific notes](#add-scenario-specific-scoring-notes) |
| Spin up a variation            | [Hard/easy variants](#make-a-hard-or-easy-variant), [Dynamic variables](#reuse-with-dynamic-variables)                     |
| Figure out why it broke        | [Diagnose from a transcript](#diagnose-a-stuck-conversation)                                                               |

***

## Character & Voice

### Make the persona specific

Give the character a personality, a motivation, and a communication style, not just a job title. The more human they feel, the less robotic the conversation.

> Do: "VP Finance, skeptical and efficiency-oriented, wants to pre-qualify on price before investing time." Don't: "VP Finance."

Use a real personality type (for example DISC: Dominant, Influential, Steady, Conscientious) to keep buyers distinct. More on character fields in [Advanced Roleplay Building](/roleplays/advanced-building#shape-the-character).

### Write the opening line verbatim

Write the character's first line in exact words so the AI doesn't improvise an off-base start. It sets the tone for the whole conversation.

### Natural speech patterns

Make the character sound human with occasional filler and pauses.

```text theme={null}
Use occasional filler words like "hmm," "well," "um," and natural pauses
when thinking through a response. Don't overdo it.
```

### Control response length

Stop a character from monologuing or over-sharing.

```text theme={null}
Keep responses to 1-2 sentences unless asked to elaborate. This is a busy
manager on a call, not someone who monologues or overshares.
```

### Pronunciation rules

If the character mangles a product name or acronym, add a text guideline that spells out how to say it.

```text theme={null}
Always pronounce "NCR" as "N-C-R" (three separate letters), never "nicker."
Pronounce the product "Naviga" as "nuh-VEE-guh."
```

### Segment variants

The same buyer type behaves differently by company size. Spell out how an Enterprise buyer's motives differ from SMB or mid-market, or use [conditional context](/roleplays/conditional-context) to vary the character by the learner's profile.

***

## Conversation Behavior

These live in the character's [Conversation Guidelines](/roleplays/conversation-guidelines), built from **text sections** (free-form rules like response length, speech, pronunciation) and **trigger sections** (a trigger plus leveled responses).

### Be explicit about objections

List the specific pushbacks you want and when, or the AI invents objections that may not match your product. If you only want pressure on price and timeline, say so.

### "Don't ask superfluous questions"

One of the highest-impact lines you can add. Without it the AI over-asks and creates dead ends.

```text theme={null}
Don't ask superfluous questions. Accept what you're told unless it touches
a point of resistance I've defined.
```

### Define exit conditions

If the rep must do something specific to advance (ask for the meeting, present ROI), say so, or reps get stuck waiting for a "magic phrase" the character is silently holding out for.

### Define a concession point

Tell the AI when to soften, or it stays in objection mode forever.

```text theme={null}
After the rep addresses the budget objection with at least two supporting
points, become more open to discussing next steps.
```

### Add multi-level discovery depth

Make reps dig instead of accepting the first answer. Ask the agent to add levels to a response guideline.

```text theme={null}
Level 1 (initial): guarded, deflects.
Level 2 (if pressed well): shares generalities.
Level 3 (only with excellent probing): shares specific numbers.
```

### Add or edit text and trigger sections

Reach for a **text section** for a general rule ("keep answers short") and a **trigger section** for a specific moment ("when asked about budget, respond like this"). You can add your own triggers and as many response levels as a moment needs. Full editor: [Conversation Guidelines](/roleplays/conversation-guidelines).

***

## Evaluation Criteria

### Write observable behaviors, not vague qualities

> Do: "Asks at least three open-ended questions about the prospect's workflow before presenting." Don't: "Demonstrates good discovery skills."

### Write criteria as observable chains

Format a criterion as "rep does X, then Y, without Z."

```text theme={null}
Rep does not quote a specific number, explains pricing structure only if
pressed, and redirects to discovery within the first minute.
```

### Add scenario-specific scoring notes

Drop a note right into the Good / Fair / Poor text to override or supplement the default logic.

```text theme={null}
Good: Rep establishes credibility with a relevant customer example.
Scenario-specific: Rep can only earn Good if they confirm the timeline
before discussing price.
```

Align the weighting to what the scenario is actually testing, and associate a [skill](/skills/overview) with each criterion so the results feed your analytics. Editing criteria: [Advanced Roleplay Building](/roleplays/advanced-building#fine-tune-the-evaluation-criteria).

***

## Editing & Variants

### Edit in place; Remix for variants

Everything in a published scenario is editable in place, either directly or by telling the agent what to change: context, criteria, character identity and behavior, response guidelines, voice, personality, session settings. Reach for **Remix** only when you want a separate variant and the original kept intact. See [Edit a Scenario](/roleplays/edit-scenario) and [Remix a Scenario](/roleplays/remix-scenario).

### Make a hard or easy variant

Remix and describe the change, keeping the original intact.

```text theme={null}
Make a hard version of this scenario: the character is more resistant to
sharing information and the grading is stricter. Keep the same character
and context.
```

### Reuse with dynamic variables

Set a placeholder like company name once and have it update everywhere in the scenario, so you can reuse a scenario across teams or accounts without hand-editing every mention.

### Clone vs. Remix

Use **Clone** to adapt someone else's scenario (for example from a shared collection) into your own; use **Remix** to create a variation of your own scenario while keeping the original. To change a scenario itself, use **Edit**. If you get stuck while editing, use **Revert** to return to a previous version.

***

## Testing & QA

### Test with Try Now before publishing

Click **Try Now** for a quick conversation with the character. Check the opening line, whether resistance feels right, whether it reveals too much, and whether the flow gets stuck.

### Preview with a simulated transcript

Don't always run it live. Ask the agent for a sample run to review in a couple of minutes.

```text theme={null}
Generate a simulated transcript of a mid-tier (Fair) performance of this
scenario so I can see where the conversation falls short of Good.
```

### Diagnose a stuck conversation

Paste a real transcript back to the agent and ask what went wrong.

```text theme={null}
This is a conversation that happened. The rep got stuck here: [paste].
Why did this happen, and what in the scenario design caused it?
```

To fix it, open the scenario in **Edit** and paste the transcript with a note on what felt off. The agent makes targeted changes based on what actually happened. You can also ask the builder about its own blind spots before publishing.

***

## Planning Shortcuts

* **Start with the moment, not the topic.** "Prospect pushes back on price during a procurement demo after seeing the proposal" beats "objection handling." Specific moments make realistic practice.
* **Split long conversations.** Keep a scenario under about ten minutes; split longer ones at natural stopping points and match the character name, company, photo, and voice across parts so it feels continuous.
* **Start simple, then layer.** Build the easy, cooperative version first, confirm it works, then edit in tougher objections, a more resistant persona, or time pressure. Remix instead if you want to keep the easy version as its own scenario.
* **Feed a long brief in pieces.** If you've worked up a detailed briefing document, don't expect one large paste into the chat to come through intact. Add the document to the Knowledge Hub and point the agent at it as a [source](/roleplays/advanced-building#build-from-your-own-materials), prioritize the background, objectives, and context that matter most, then layer the rest in with focused instructions ("add this background to the character," "make sure the context includes these market conditions"). Anything that must appear in the scenario is worth naming explicitly.

For the foundational planning model (the Four Pillars) and the build methods, see [Creating Scenarios](/roleplays/ai-agent).

***

## Getting Help

**Need help?** Contact us at [hello@exec.com](mailto:hello@exec.com) for guidance on building scenarios or any questions about the Scenario Studio.