Using Brando

This guide walks through how to use Brando in practice, from first experiments to a fully operational Brand Knowledge Graph:

Get oriented.
Model a first brand.
Choose how you’ll store and version the graph.
Validate and iterate.
Integrate with LLMs, MCP, and APIs.
Grow into portfolios, categories, and automation.

It assumes you’ve read:

and now want to put Brando to work.

1. Decide your initial scope

Start with a narrow, high-leverage scope, not the entire brand universe.

Typical starting points:

One brand, one channel:
- e.g. “Corporate brand, public website chatbot”.
One brand, two key contexts:
- e.g. “Support chatbot” and “Marketing landing pages”.
A single regulated flow:
- e.g. “Savings account recommendations in UK”, “Health information Q&A”.

For that scope, identify:

the brand (top-level brando:Brand),
the context(s) (brando:Context),
the tone & language requirements (Brando:VerbalIdentity),
the minimum policies you need (brando:Policy).

You’ll add visuals, audio, categories, and automation later.

2. Model your first brand graph

Use JSON-LD or YAML as your authoring format (YAML is usually easier for human editing).

At minimum, create:

A brando:Brand node
A brando:Context node
A Brando:VerbalIdentity node
A brando:Policy node

2.1 Brand node

Capture identity and intent:

@context:
  schema: https://schema.org/
  brando: https://brandoschema.com/

@graph:
  - @id: https://example.com/brand/acme
    @type: brando:Brand

    schema:name: ACME Tools
    brando:missionStatement: >
      Make high-quality tools accessible to every builder.
    brando:visionStatement: >
      A world where anyone can build with confidence.
    brando:coreValues:
      - Reliability
      - Practicality
      - Straight-talking support

2.2 Context node

Capture where and for whom behaviour changes:

  - @id: https://example.com/brand/acme/context/support-chat
    @type: brando:Context

    brando:audienceSegment:
      - Existing customers seeking product support
    brando:audiencePersona:
      - Hands-on builder
    brando:domainContext: >
      {"channel":"chat","surface":"support-portal","region":"global"}

    brando:usesVerbal:
      @id: https://example.com/brand/acme/tokens/verbal/support

2.3 Verbal node

Capture tone, style, and vocabulary:

  - @id: https://example.com/brand/acme/tokens/verbal/support
    @type: Brando:VerbalIdentity

    brando:toneOfVoice: >
      Straightforward, encouraging, and practical.
    brando:dialogueStyle: >
      Talk to the user like a colleague on the job site.
    brando:writingStyle: >
      Short, direct sentences. Concrete instructions first; theory later.

    brando:approvedTerms:
      - heavy-duty
      - reliable
      - safety first

    brando:prohibitedTerms:
      - indestructible
      - lifetime-guaranteed for anything

    brando:mustDo:
      - Offer a clear next step or action in every answer.

    brando:mustNotDo:
      - Mock or belittle the user’s level of expertise.

2.4 Policy node

Capture guard rails and refusal patterns:

  - @id: https://example.com/brand/acme/policies/basic-brand-safety
    @type: brando:Policy

    brando:enforcementLevel: mandatory
    brando:riskTag:
      - brand-safety

    brando:guardRails:
      - Do not make claims about capabilities that are not documented.
      - Do not provide safety-critical advice without including an appropriate caution.

    brando:refusalStrategies:
      - >
        If the user asks for unsafe modifications, explain why it is unsafe
        and suggest a safer alternative.

    brando:retrievableInLLM: true

Link it from the brand:

  - @id: https://example.com/brand/acme
    @type: brando:Brand
    # ...
    brando:hasContext:
      @id: https://example.com/brand/acme/context/support-chat
    brando:usesVerbal:
      @id: https://example.com/brand/acme/tokens/verbal/support
    brando:hasPolicy:
      @id: https://example.com/brand/acme/policies/basic-brand-safety

Now you have a minimal working Brand Knowledge Graph.

3. Choose where this lives

You need a single source of truth.

Common options:

Git repo
- YAML or JSON-LD files under brands/<brand>/...
- Reviewed via PRs
- Ideal for early pilots.
Graph database / triple store
- Store the JSON-LD as RDF/graph.
- Useful when you already have a knowledge-graph stack.
Document store
- JSON-LD documents per brand.
- A simpler option if you don’t need full graph DB features yet.

Non-normative suggestion:

Start with Git-managed YAML → generate JSON-LD as a build artefact.
Once stable, mirror into a graph store if your infra supports it.

4. Validate and iterate

Before wiring into production:

Schema validation
Check that:
- each node has a valid @type,
- properties are used on appropriate domains,
- required properties are present (e.g. schema:name on brando:Brand).
You can use:
- JSON Schema or similar for structural checks,
- custom scripts to enforce your own usage rules.
Brand review
- Have brand/marketing teams review:
  - tone of voice text,
  - approved/prohibited terms,
  - guard rails and refusal strategies.
Treat this as a living extension of your brand guidelines.
Prompt simulation
- Build a simple script or notebook that:
  - loads the Brando graph,
  - generates a system prompt,
  - runs a few test questions through your LLM,
  - lets humans review the outputs vs the Brando rules.

Use this loop to refine Verbal, Policy, and Context contents.

5. Integrate with your chatbot / LLM

The basic integration pattern:

At request time, your backend receives:
- a brandId (brando:Brand @id),
- a contextId (brando:Context @id), or enough info to resolve one.
Brand OS / middleware:
- loads Brand + Context + Tokens + Policies from the Brando graph,
- constructs a RuntimeBrandConfig (see Runtime Integration),
- builds a system prompt and policy configuration for your LLM.
LLM adapter:
- prepends the system prompt,
- optionally applies output checks using Brando Policy and Verbal data,
- logs which Brando nodes were used.

If you’re using MCP or a shared tools layer, you can expose this as:

get_brand_config(brandId, contextId)
list_policies(brandId)
get_prompt_scaffold(brandId, contextId)

so any AI client can call into your Brand OS.

6. Add other modalities and structure

Once text is working, extend:

6.1 Visual identity

Add Brando:VisualIdentity nodes for:

colour palette (brando:colourPalette as JSON string),
typography (brando:typography),
design tokens (Brando:VisualIdentity),
logo and imagery rules (Brando:VisualIdentityUsageGuidelines, Brando:VisualIdentityReferenceLink).

Link them:

brando:usesVisual:
  @id: https://example.com/brand/acme/tokens/visual/master

These can drive:

design systems,
generative image tools,
theming for UI.

6.2 Audio identity

Add Brando:AudioIdentity nodes for:

sonic logos,
voice signature,
pronunciation guide,
audio usage guidelines.

Link with brando:usesAudio.

6.3 Product categories

Define brando:BrandedCategory nodes for key categories and link to:

GS1 / UNSPSC / Google Product Taxonomy via:
- brando:gpcCategoryCode
- brando:unspscCode
- brando:googleProductCategoryId

Then link from Brand via brando:hasProductCategory.

This enables category-aware prompts and RAG filters.

7. Introduce governance & automation

As you mature:

7.1 Governance detail

Use policy-level properties:

brando:enforcementLevel – e.g. "mandatory", "advisory", "conditional".
brando:riskTag – e.g. "brand-safety", "financial-regulation".
brando:refusalStrategies, brando:riskScenarios – for explainability and training.
brando:reviewWorkflow, brando:updatePolicy – for lifecycle management.
brando:retrievableInLLM – to control what enters high-trust RAG or prompt corpora.

7.2 Automation rules

Add brando:AutomationRule nodes to tie metrics to behaviour:

triggers (brando:triggerType),
metrics (brando:monitoredMetric JSON string),
sources (brando:dataSource),
actions (brando:automationAction JSON string).

Your Brand OS can then:

detect rising unsafe output rates,
tighten enforcement levels or add extra checks,
notify governance teams automatically.

8. Common adoption patterns

Typical sequence organisations follow:

Pilot (single brand, single use case)
- Minimal Brand + Context + Verbal + Policy.
- YAML in Git; manual review; small-scale LLM integration.
Operational rollout for text
- Multiple contexts per brand (support, marketing, sales).
- Richer policies and refusal patterns.
- Brand OS service/API introduced.
Portfolio & categories
- Multiple brando:Brand nodes (house-of-brands, regional variants).
- BrandedCategory plus GS1/UNSPSC/Google alignment.
Multi-modal and automation
- Visual and audio tokens.
- Automation rules driven by safety/quality metrics.
- Integration with observability and risk platforms.

Brando vocab v1.3 is designed to support each stage without forcing you to adopt everything at once.