Skip to content

Configure Brand Guidelines in Opal

⏱ 20 minutes intermediate
πŸ“œCoreOpal

Without guidelines, Opal generates content that is well-written but sounds generic. Every organization has a voice β€” formal or casual, technical or accessible, bold or measured. Instructions tell Opal how to adapt its output so that generated content is consistently on-brand, whether a content author uses it for a blog post or a marketer uses it for campaign copy.

Opal’s instructions system goes beyond simple brand voice. Instructions are foundational context, rules, and behavioral guidelines that shape every response. They can include brand tone, product details, personas, term-bases, and compliance requirements. A key differentiator is Opal’s ability to dynamically select and apply instructions based on the context of each request.

Opal separates instructions into two types that work together:

Set by administrators, these enforce standards across every user in the organization. Use them for:

  • Brand voice and tone β€” Define how your organization communicates
  • Compliance rules β€” Enforce regulatory language requirements
  • Approved terminology β€” Set preferred and banned terms
  • Templates and formats β€” Standardize output structure
  • Tool usage rules β€” Specify when certain tools should or should not be used

Set by individual users, these customize Opal’s behavior for your own workflows without affecting anyone else. Use them for:

  • Response format preferences β€” Bullet points vs. prose, short vs. detailed
  • Tone adjustments β€” More casual or more formal than the org default
  • Expertise level β€” Technical depth appropriate for your role
  • Language preferences β€” Default output language
  • Output defaults β€” Standard length, structure, or style you prefer

When you send a request, Opal does not blindly apply every instruction. It follows a selection process:

  1. Gathers all active instructions (both organization-wide and personal)
  2. Matches tool-based instructions first β€” if an instruction specifies it should activate when a particular tool runs, that match takes priority
  3. Evaluates remaining instructions for relevance using AI, based on the β€œWhen to use” conditions you defined
  4. Injects matched instructions into the context for the current request

The relevancy model is intentionally inclusive. Opal errs toward applying an instruction rather than skipping it, so writing clear β€œWhen to use” conditions is important to avoid conflicts.

Every instruction contains five elements:

ElementPurpose
NameA brief, descriptive identifier for easy recognition
ActiveToggle that controls whether Opal considers this instruction
Core InstructionThe behavioral guidance: objectives, steps, format rules, constraints
Where to useWhich Optimizely products and instances the instruction applies to
When to useConditions that determine when Opal activates this instruction
  • Anchor to specific workflows rather than vague intent (β€œWhen the user asks about Q4 campaign planning” not β€œWhen the user asks about marketing”)
  • State keywords explicitly to prevent semantic mismatches
  • Add negative conditions (β€œDo NOT use for technical documentation requests”)
  • List synonyms for narrow concepts to maintain coverage
  • Draw clear boundaries between similar instructions to avoid overlap

Optimizely provides nine prebuilt instructions (turned off by default):

  • Campaign Brief
  • Industry Marketer
  • Keyword Research
  • Marketing Planning Meeting
  • Marketing Researcher
  • Personalization Advisor
  • Task Brief
  • Tone of Voice
  • Video Transcription

Enable any prebuilt instruction, then customize its core content to match your organization’s specific needs.

After creating instructions:

  1. Open Opal Chat from any connected product
  2. Ask it to generate content that should trigger your instruction (for example, β€œWrite a 100-word product description for our analytics dashboard”)
  3. Compare the output against your brand guidelines
  4. If the tone, terminology, or format is off, return to the instruction and refine the core content or β€œWhen to use” conditions
  5. Test edge cases β€” requests that are similar but should trigger different instructions

Instructions need regular review:

  • After a brand refresh or rebrand
  • When entering new markets or audiences
  • After team feedback on AI output quality
  • When you add new Optimizely products
  • Quarterly, as routine practice