BrandWise Scenarios — Creation, Configuration, Basic & History Modes

What is a scenario

A scenario is the core evaluation unit in BrandWise. It combines three components:

• Intents — questions and prompts sent to AI models
• Models — the set of LLMs to test (ChatGPT, Claude, Gemini, DeepSeek, and more)
• Brand — the profile used to evaluate responses

Each scenario answers a specific business question: "How does our brand appear in AI responses about the best CRMs?", "Do models recommend our restaurant when people search for dinner spots?", or "What do AI models say when comparing us with competitors?"

Running a scenario produces an aggregated report with scores across 6 metrics, a model heatmap, and evidence quotes — verbatim excerpts from model responses backing each score.

Scenario list

All project scenarios are displayed in the Scenarios section in the project sidebar. Each scenario shows:

• Name and linked brand
• Mode (Basic or History)
• Latest run status
• Creation and last run date

You can create a new scenario, open an existing one for editing, or duplicate it.

Creating a scenario

Click Create scenario on the scenarios page. The editor opens with several sections.

Core fields

Basic mode — single-turn queries

In Basic mode, each intent is sent to the model as a standalone question. The model replies with a single message, and that response is evaluated across 6 metrics.

Best suited for:

• Direct questions: "What's the best CRM for small businesses?"
• Comparison queries: "Compare Slack vs Microsoft Teams for remote work"
• Recommendation requests: "Suggest a restaurant in downtown Manhattan for a business dinner"
• Overview queries: "Top 5 cloud storage services for enterprises"

History mode — multi-turn conversations

In History mode, the model engages in a conversation with a virtual user (persona) who pursues a specific goal. The dialog runs for 1 to 10 turns and ends when one of these conditions is met:

• task_completed — the persona achieved the goal (e.g., chose a restaurant)
• early_stop — the persona decided to end the conversation early
• max_turns — the turn limit was reached (default: 10)

When you select History mode, additional fields appear:

History Task examples:

• "You're choosing a restaurant for a business meeting and need to pick 2 options"
• "You're looking for a CRM for a 10-person team with a budget under $50/month and need to make a decision"
• "You're comparing cloud storage services for a corporate data migration"

History mode is ideal for evaluating how models behave in realistic conversations — when a user refines requirements, asks follow-up questions, and gradually narrows down options. This produces a more realistic picture than single-turn queries.

Intents — questions for models

Intents are the specific questions and prompts that BrandWise sends to the selected AI models. Each scenario contains 1 to 20 intents, each between 10 and 1,000 characters long.

Best practices for writing intents

Phrase intents as real user queries to AI:

Tips:

• Mix broad and specific questions for a complete picture
• Include queries that real users actually ask ChatGPT or Claude
• Add queries with constraints (budget, location, team size)
• Use different phrasings of the same question to test response consistency

Managing intents

In the scenario editor, intents appear as a numbered list. You can:

• Add a new intent with the Add intent button
• Edit the text of an existing intent
• Delete an intent
• Drag and drop intents to reorder them

Selecting AI models

BrandWise provides a model catalog grouped by vendor. At least 1 model must be selected to run a scenario.

Available vendors and models

Models are grouped by vendor in the catalog. Select the models you want using toggle buttons. Inactive models and those marked "coming soon" are visually distinguished and cannot be selected.

Tip: for your first run, select 3–5 models from different vendors to see how your brand is represented across ecosystems. After reviewing the first report, expand the model set based on results.

Saving and launching

The scenario editor offers four actions:

If you navigate away with unsaved changes, a warning dialog will appear — you can choose to save or discard changes.

Duplicating a scenario

You can duplicate any existing scenario. The copy inherits all settings from the original:

• Name (with a "Copy" label)
• Linked brand
• Mode (Basic / History)
• Persona and History Task (for History mode)
• All intents
• Selected models

Duplication is useful when you want to test variations — for example, the same set of intents with a different model selection or a different brand.

Execution matrix

When a scenario is launched, BrandWise generates execution items using this formula:

Number of items = intents × models × temperature variants

For example, a scenario with 5 intents and 4 models produces 20 execution items (with one temperature variant). Each item is a separate model call evaluated across 6 metrics.

Run cost in AI credits depends on the number of items, selected models, and response length.

What's next

• Running an evaluation: process, statuses, actions — how to launch a scenario and track progress
• Brand profile — setting up a brand for accurate evaluation
• Personas — creating user profiles for History mode
• Create an account and get started