Trellis, Sprout's AI Agent, transforms social listening data into actionable insights, helping your team work smarter. By directly analyzing social insights, Trellis surfaces trends, conducts competitive analysis, and provides contextual answers to inform your brand's processes, products, and services, all without manual report sifting.

Trellis is designed to help your team work faster, smarter and with deeper insights. It empowers your team to leverage social insights to enhance the quality of your brand’s processes, products and services. You can ask questions to Trellis directly and receive direct, contextual answers rooted in social insights. It can also surface trends without the need to sift through reports and configurations. Right now, Trellis works with Listening data to help you extract insights and recommendations.

Getting started with Trellis (Sprout AI Agent)

It’s easy to get started using the Trellis in Sprout. To begin:

Navigate to the global toolbar on the left side of the screen, and then click the Ask Trellis icon:

(Note: If you aren't seeing Trellis in the left side menu, you might still be on our legacy navigation. To update your navigation experience: In the lower left side click on your initials and select "Turn on Unified Navigation")
Trellis will appear on the right hand side of your session.

Enter your conversation prompt into the input box.
Press Enter.
Note: Trellis is limited to only using Listening data at this time.

Getting the most out of Trellis (Sprout AI Agent)

The following section provides common use cases and guidelines for creating effective prompts to help you extract the signals you’re after.

Competitive Analysis with Trellis

Use Trellis to discover the following insights:

Trend Identification - “Fetch the 50 most viral posts in the [topic] from the past week. What trends or themes can we use to inform our future campaigns? Please explain your thought process.”
Brand Health - “Are there any spikes in sentiment or unusual volume over the past week in the [topic]? Don’t include today as the data may be incomplete.”
Product/Event Feedback - “Are there recurring concerns about [product/event] over the past three months? What themes show up most in negative feedback, if so?”
Daily Performance Monitoring - “Get me the top 30 most engaging posts from the [topic] in the past 3 days and summarize what’s resonating with our audience. Don’t include today as the data may be incomplete.”

Trellis Threads

Trellis can now start and store several conversations. As with other AI tools, you may want to come back to the conversation later and not lose your place. You will see a ‘New Chat’ button in the upper right hand corner of your chat window:

From here you can start a new conversation. Conversation names are dynamically created but can be changed.

To rename, delete, or switch between chats, click the history button:

Trellis Threads are useful when you have a good conversation going and want to use the same line of questioning with new data over time. Additionally, you can keep a thread open and come back after a meeting with new information to dig deeper.

How to write a good prompt with Trellis (Sprout AI Agent)

Writing a good prompt is key to extracting the insights that matter most to your brand. Use the following tips to craft the best prompts:

Be clear and specific about the goal - state what you want Trellis to achieve, not just the data you want to look at.
- For example, “What are the most common themes in Starbucks mentions this month and how are they trending?”
Provide the right context - include the timeframe and topic.
- For example, “Summarize sentiment around Nike in the brand health topic in July.”
Ask for actionable insights, not just data - frame prompts so Trellis provides meaning or recommendations, not raw numbers.
- For example, “Which competitors gained the most positive attention this quarter, and what should we learn from their content?”
Guide Trellis with framing - use question words like what, why, how, or should for deeper analysis.
- For example, “Why did sentiment spike for Tesla last week?”
Break down complex tasks - for multi-part needs, guide Trellis step-by-step or ask it to structure results clearly.
- For example, “Identify top themes for Starbucks this month, then rank them by engagement. Suggest three content opportunities based on those themes.”
Leverage Trellis’s tools - keep in mind the functional capabilities it has (sentiment analysis, trend charts, message summarization).
- For example, “Plot volume and sentiment over time for Adidas in the last 3 months and summarize emerging topics.”
Stay open-ended for exploration - if you’re looking for patterns you don’t already know, phrase prompts to encourage exploration.
- For example, “Are there any new slang terms or hashtags emerging around Starbucks this month?”
Think of Trellis as a template - if you’re having trouble getting the answer you want, ask it to help you create a useful prompt.
- For example, “How should I ask this question, given your capabilities, if I want to know about volume and top themes of social conversations about sustainability?”

FAQs

Is my data secure when using Sprout’s AI features, and does it train external AI models like OpenAI’s?

Yes, your data is secure. Sprout’s AI features, including those that leverage OpenAI, such as Trellis, are configured to ensure your data is never used to train external AI models. Your proprietary information remains private at all times. If you’ve already approved Sprout’s security protocols during onboarding, enabling AI features does not introduce any additional security risk beyond what was previously approved.

What is the smallest time period I can analyze or view data for?

The agent is designed to work with time periods no smaller than one full day. You can’t analyze data at an hourly or minute level.

How can I use Trellis to analyze product or event feedback?

You can prompt Trellis with a query like, "Are there recurring concerns about [product/event] over the past three months? What themes show up most in negative feedback, if so?"

What are the key elements for writing an effective prompt for the Sprout AI Agent?

Good prompts should (1) be clear and specific about the goal, (2) provide the right context (timeframe and topic), (3) ask for actionable insights, (4) guide Trellis with framing (using 'what,' 'why,' 'how'), and (5) break down complex tasks.

Providing feedback

You make Trellis better. Providing feedback is a great way to ensure it gets it right.

Use the 👍 / 👎 reactions on agent responses.

Thumbs down is most valuable. If Trellis falls short, click 👎 and tell us why using the checkboxes and free-text input to provide suggestions. The “why” is critical to improving Trellis.

Was this article helpful?

0 out of 0 found this helpful

Trellis (Sprout AI Agent) Overview and Best Practices

Table of Contents