Skip to main content

How to Set Up Prompt Instructions for Better Replies

Prompt instructions shape how AI responds in different scenarios. By creating multiple prompts and applying filters, you can ensure every reply matches the tone, context, and intent of the original message.

Updated over a week ago

Why This Matters

Not all replies should sound the same. Not every comment deserves the same reply style. A frustrated customer expects empathy and problem-solving. A happy fan expects warmth and celebration. Multiple prompts let you give AI clear rules for each situation, while filters make sure the right prompt is applied automatically.

Tips for Building Effective Prompts

Strong prompts are the difference between generic replies and responses that feel tailored to your client or prospect. Always build a custom prompt instead of relying on the default.

Best Practices

  1. Fill every field: Brand name, industry, and tone of voice should never be left blank.

  2. Turn on AI post analysis: Always enable this. It grounds replies in the actual post context.

  3. Refine continuously: Test replies often. When something feels off, add it as an instruction additional details.

💡 Tip: Great prompts are built iteratively. The more you test, refine, and update them, the closer AI replies will match your brand’s expectations.

Create Multiple Prompts

  1. Go to Control Panel > Engage Library > AI Prompts.

  2. Click Add New Prompt.

  3. Name your prompt and add a short description so your team knows its purpose.

    • Example: “Complaints Prompt” for customer escalations.

    • Example: “Positive Sentiment Prompt” for compliments and fan messages.

  4. Customize how the AI will respond:

    • Tone of voice: Define if replies should be formal, approachable, concise, or celebratory.

    • Emoji use: Decide if emojis should be avoided, limited to one, or encouraged.

    • Language: Set replies to match the original message or lock to a specific language.

  5. Fill in Context fields:

    • Brand name

    • Industry

    • Additional details such as target audience, values, or product guidelines

Apply Filters for Automatic Selection

  1. In the AI prompt modal, turn on Advanced Default Settings.

  2. Add filters that define when each prompt should be applied. Example:

    • Positive Sentiment Prompt → Filter on Positive Sentiment

Save the prompt.

Once filters are set, the platform automatically detects signals, matches them with your filters, and applies the correct prompt when you click Generate AI Reply.

Examples of Prompts You Can Build

You can create prompts for different situations and let filters handle when they’re applied. This keeps replies sharp, relevant, and tailored to both the user and the business objective.

By sentiment

  • Positive sentiment: Use a warm, celebratory tone, add light emojis, and thank the customer.

  • Negative sentiment: Use empathy, keep tone calm and concise, avoid emojis, and offer a clear next step.

  • Filter: Sentiment analysis (positive, neutral, negative).

By tags

  • FAQ tag: Draft replies that are clear, factual, and reference knowledge sources.

  • Purchase intent tag: Draft replies that highlight product benefits, link to purchase pages, and encourage action.

  • Filter: Message tags applied through automation or manual tagging.

By region

  • US profiles: Replies can use American English, casual tone, and region-specific offers.

  • UK profiles: Replies can use British English, a more formal tone, and local references.

  • Filter: Asset-level filters based on the social profile the comment comes from.

By content type

  • Ads: Prompts can be optimized for conversion. Replies should reinforce the offer, create urgency, and include strong CTAs.

  • Organic posts: Prompts can focus on community-building. Replies should feel conversational, supportive, and low-pressure.

  • Filter: Post type (ad vs organic) or message type (comment vs direct message).

Other ideas

  • By product line: Create a prompt for each category (e.g., skincare vs makeup) so replies always include product-specific value.

  • By escalation level: One prompt for first-line complaints, another for escalated cases that need apology plus a support handoff.

  • By campaign: Build prompts tied to seasonal campaigns or launches to ensure messaging stays consistent across replies.

Did this answer your question?