Skip to main content

Practical Examples

Think of these examples as recipes you can cook from — each one lays out the whole dish, from what sets it off to the steps your agents follow. They show complete automations you can copy: the trigger (the event that starts the automation), the agent steps, and the exact instructions you'd give each agent.


How to use these

Pick the example closest to what you're trying to do, then adapt it. Each example explains why it uses a particular trigger so you can recognize the same pattern in your own work:

  • A Schedule trigger runs the automation on a clock — every morning, every Monday, and so on.
  • A Webhook trigger runs the automation the moment something happens in an outside service — like a form being submitted (a webhook is the automatic alert that outside service sends Nirvai).

You don't have to recreate these by hand. The fastest route is described in Building these examples at the bottom of this page.


Example 1: Daily News Digest

Goal: Get a summary of industry news in your inbox every morning.

Why a Schedule trigger? You want this to run automatically at the same time every day, with no manual action needed.

StepAgentInstructions
1Research Agent"Search the web for the latest AI industry news published in the last 24 hours. Focus on product launches, funding rounds, and major partnerships. List each article with its title, source, and a one-sentence summary."
2RouterIf articles found → go to Writer Agent. If no articles → go to Notification Agent with "No notable news today."
3Writer Agent"Take the list of articles and write a concise 5-bullet summary of the key developments. Highlight the most impactful story first. Keep each bullet to 2 sentences max."
4Notification Agent"Send the news summary to my email with the subject line 'AI News Digest — {today's date}'."

Example 2: Customer Feedback Processor

Goal: Automatically categorize incoming feedback and route it based on urgency.

Why a Webhook trigger? The automation needs to run every time a customer submits a feedback form, which is an event coming from an outside service.

StepAgentInstructions
1Analyst Agent"Read the customer feedback from the webhook data. Determine: 1) Sentiment (positive, negative, neutral), 2) Category (billing, product, support, feature request), 3) Urgency (1-5 scale, where 5 is critical). Output a structured summary."
2RouterIf urgency ≥ 4 → go to Priority Handler. Otherwise → go to Standard Handler.
3aPriority Handler"This is urgent feedback. Draft an immediate, empathetic response addressing the customer's concern. Flag this for human review by adding 'URGENT' to the subject line."
3bStandard Handler"Draft a thoughtful response thanking the customer for their feedback. Acknowledge their specific point and mention that the team will review it."

Example 3: Content Publishing Pipeline

Goal: Create a weekly batch of social media content based on trending topics.

Why a Schedule trigger? Content planning happens on a regular weekly cycle, always on the same day.

StepAgentInstructions
1Trend Researcher"Search for the top 5 trending topics in the digital marketing industry this week. For each topic, include: the topic name, why it's trending, and 2-3 relevant hashtags."
2Content Creator"Using the trending topics, write 5 social media posts (one per topic). Each post should be under 280 characters, engaging, and include a call-to-action. Write versions for Twitter and LinkedIn."
3Editor Agent"Review the 5 social media posts for tone, grammar, and brand consistency. Make sure each post is professional but conversational. Fix any issues and return the polished versions."
4Formatter Agent"Organize the final posts into a content calendar format: for each post, list the platform, suggested publish day (Tuesday through Saturday), the post text, and the hashtags."

Example 4: Lead Qualification from Form Submissions

Goal: When a potential customer fills out a contact form, automatically research their company and qualify the lead.

Why a Webhook trigger? Each form submission is an event from an outside service that needs immediate processing.

StepAgentInstructions
1Research Agent"The webhook contains a form submission with name, email, company, and message. Research the company: find their website, industry, estimated size, and recent news. Compile a lead profile."
2RouterIf company has 50+ employees AND is in a target industry → High quality. Otherwise → Low quality.
3aSales Prep Agent"This is a high-quality lead. Write a personalized outreach email referencing their company's recent news. Include 3 specific ways our product could help them based on their industry."
3bNurture Agent"This lead needs nurturing. Draft a friendly welcome email with links to our best resources and case studies relevant to their industry."
4Notification Agent"Send the lead profile and drafted email to the sales team channel. Include the qualification result (high/low) and the recommended next action."

Building these examples

The fastest way to recreate any of these examples is with AI Setup:

  1. Go to the Automations page and click "AI Creation".
  2. Describe the example in your own words — you can paste the goal description as a starting point.
  3. The AI proposes which agents to use — confirm or swap them.
  4. Pick from the workflow options the AI generates.
  5. Test the automation from the setup page.
  6. Click Activate after a successful test.

You can also build these manually with the Manual Builder if you prefer full control over every step.

tip

These examples use generic agent names like "Research Agent" and "Writer Agent." In practice, you can use the default Nirvai agent for all steps, or create custom agents with specialized knowledge and tools for better results.


Common mistakes

MistakeWhy it's a problemBetter approach
Using a Schedule trigger for something that should react instantlyA scheduled run only fires on the clock, so a form submitted at 9:05 waits until the next scheduled timeUse a Webhook trigger when the automation should react the moment an outside event happens
Writing vague agent instructions like "summarize the news"The agent guesses at length, format, and focus, so results vary run to runSpell out the format, length, and what to focus on — like the instructions in the examples above
Skipping the Router stepEvery item gets the same treatment, even when urgent and routine cases need different handlingAdd a Router to send each item down the right path based on what the previous agent found

What's next