AI can be a powerful tool, but it can be difficult to know where to start when trying to use it effectively. Given LLMs like Claude or ChatGPT are designed to be people pleasers, it can be hard to depend on the insights these LLMs provide.

Prompting strategies such as the Chain of Thought or COSTAR can be great starts to help build the muscle to start thinking in the way of building a prompt, and in this article we will show some prompts that can help serve as starting off points.

Feel free to take bits and parts of these prompts to craft one that matches what you need!

While your company's operations may seem obvious to you, the benefits or even what your company does may be completely alien to an AI model. These details can matter to help guide the LLM to reach an outcome relevant to your organization.

To help illustrate this, check out the example scenarios here:

Initial approach:

We are an online clothing brand that has released a new jacket to our catalogue. We are hoping this can help drive revenue and bring more customers to our platform.

Improved approach:

We're a US-based online brand specializing in fashionable, high-quality swimwear and summer clothing. Our customer base largely comes from colder US regions, buying for warm-weather vacations.

The recent decline in remote work has led to fewer purchases from this demographic, impacting our quarterly revenue.

To address this, we are launching a versatile clothing line designed for year-round use.

These pieces function as light jackets for fall and cover-ups for cool summer evenings. This strategy aims to diversify our offerings and maintain customer engagement beyond the traditional summer season, helping to offset our recent sales dip.

Initial Approach:

We have identified that our customer base is not able to use our product effectively out in the ocean as a result of poor cellular service. This feature release will auto-download all charting course details to help our customers still have access to our tool even in unreliable network situations.

Improved Approach:

We're a Cargo Ship Route Optimization company. We help vessels:

Our efficient routes balance environmental compliance with operational savings.

The data we provide can be 1+ terabytes large, and we have commonly heard from our customers that they are:

To address this, we released a feature that will automatically download shipping route data in a compressed simplify for offline use focusing on helping the ship complete its route.

As mentioned in the introduction, but LLMs are people pleasers. They will deliver you information that you want to hear, and even if that means they have to lie to reach that conclusion.

Here's how to approach this:

Here is an example of how you could incorporate this instruction into a prompt:

As a CX analyst, I need to provide valuable insights to our product team about customer reactions to our recent major release, [FEATURE NAME]. To accomplish this effectively, I'll follow these steps:

1. Theme Identification:

1a. Read through the collected tickets, noting recurring issues or comments.

1b. Group similar feedback into broader themes.

1c. Identify the most prevalent themes.

2. Theme Analysis: For each of the most prevalent themes:

2a. Write a brief description of the theme.

2b. Select tickets and share comments shared by the customer to support the theme identified.

2c. Consider the underlying customer needs or pain points revealed by this theme.

3. Presentation Preparation:

3a. Organize the findings in a clear, logical structure.

3b. Use a professional, concise tone appropriate for the product team.

3c. Focus on presenting actionable insights that can guide future feature enhancements.

4. Final Review:

4a. Verify that the suggestions are specific and relevant to [FEATURE NAME].

4b. Make sure that you are referencing

Provide a comprehensive analysis of customer feedback on [FEATURE NAME], including the top recurring themes and customer quotes. Present this information in a format that enables the product team to make informed decisions about future feature enhancements.

Output results in a format like so:

[CX ANALYSIS: FEATURE NAME]

[THEME {X}]

Description:

{Theme description}

Customer Quotes:

1. "Quote 1"

2. "Quote 2"

3. "Quote 3" ...

(continue adding quotes as needed)

Underlying Need: {Identified need or pain point}

[ACTIONABLE INSIGHTS]

1. Insight 1

2. Insight 2

3. Insight 3 ...

(continue adding insights as needed)

ANALYSIS TASK:

Analyze the given customer service ticket based on the "Leading w/ Curiosity" criteria. Evaluate how well the agent demonstrated curiosity to understand the customer's needs, drive comprehensive troubleshooting, and show empathy.

DEFINING "LEADING W/ CURIOSITY":

"Leading w/ Curiosity" is a customer service approach that emphasizes:

This approach aims to resolve immediate issues while also improving overall customer satisfaction and preventing future problems.

ANALYSIS STRUCTURE:

1. PRIMARY ISSUE

1.1. Brief description:

1.2. Supporting quote:

1.3. Reasoning:

2. AGENT'S CURIOSITY

2.1. Key question 1:

2.1.1. Relevance:

2.2. Key question 2 (if applicable):

2.2.1. Relevance:

2.3. Key question 3 (if applicable):

2.3.1. Relevance:

2.4. Overall curious approach analysis:

3. OUTCOMES ACHIEVED

Customer sentiment explored: [Yes/No] Quote: Analysis:

Future issues addressed: [Yes/No] Quote: Analysis:

Larger goal/initiative revealed: [Yes/No] Quote: Analysis:

4. CURIOSITY ASSESSMENT

Breadth: [High/Medium/Low] Depth: [High/Medium/Low] Key example quote: Effectiveness analysis:

5. SCORE AND JUSTIFICATION

Score: [Great (3) / Good (2) / Coaching Opportunity (0)] Justification: Critical supporting quote:

6. IMPROVEMENT RECOMMENDATIONS (if not "Great")

Recommendation and Rationale 1:

Recommendation and Rationale 2 (if applicable):

7. SUMMARY

Brief overview:

Most representative quote:

Key insight: