Understanding Customers

AI for SMBs Weekly #9 tackles how to use AI to deeply understand your customers and improve their experience. 

No changes to the AI Models for SMBs Comparison Chart since last week, but check out my analysis of what's changed on LinkedIn. Drop a comment with what features you're hoping to see next; hopefully the frontier labs are listening. 

-j

~

Every business knows they should be doing more with the mountains of customer data they collect, but few have the time or resources to analyze it properly. There is gap between collecting feedback and actually using it to improve your customers' experiences.

With generative AI, you can transform both customer data and unstructured feedback into actionable insights—without a data science degree or expensive analytics tools. This isn't just about understanding what customers think; it's about creating a continuous cycle of improvement that drives real business results.

Here's how.

Step 1:

Gather Your Customer Data

Assemble all the customer feedback and experience data you already have. Don't worry if it's scattered across different systems or formats; AI can handle it. Collect these data:

Quantitative

• Customer survey; NPS scores
• Website analytics and user behavior
• Sales figures and conversion rates
• Customer service metrics (response times, resolution rates)
• App/product usage statistics
  
  

Qualitative

• Customer service emails and chat logs
• Call or meeting transcripts (remember our guide from issue #4?)
• Social media comments and messages
• Customer reviews (Google, Yelp, Amazon, etc.)
• Open-ended survey responses

Generative AI excels in combining data types. While traditional analytics might treat them separately, AI can find patterns across both structured numbers and unstructured text. The more the better, but note that most AI models will struggle with more than ~10-20 attachments at a time. 

Step 2:

Build an "Insight Generator" Jig

Just like with proposals, cash flow forecasting, and inventory management in previous newsletters, now it's time to create a jig specifically designed to extract customer insights.

Start by creating a new GPT (ChatGPT), Project (Claude), or Gem (Gemini) with custom instructions like this:

You are my Customer Insight Analyst. Your purpose is to analyze customer feedback and data to identify actionable insights that can drive business improvements.

When analyzing customer feedback:
1. Identify recurring themes and sentiment patterns
2. Highlight specific product/service elements mentioned frequently
3. Note unexpected or surprising comments that reveal blind spots
4. Detect language patterns indicating strong emotional responses

When analyzing quantitative data:
1. Identify significant trends and patterns
2. Spot correlations between different metrics
3. Highlight anomalies or outliers worth investigating
4. Calculate key statistics to understand customer behavior

When combining qualitative and quantitative data:
1. Connect sentiment themes with performance metrics
2. Identify disconnects between what customers say and what numbers show
3. Develop testable hypotheses about customer behavior
4. Prioritize action items based on potential business impact

Format your analysis with:
- Key Findings: Clear statement of the most important insights
- Supporting Evidence: Data points that validate each finding
- Actionable Recommendations: Specific, practical steps to address issues
- Confidence Level: How certain we can be about each conclusion

Tone: Data-driven, practical, and focused on business impact rather than technical analytics jargon.

Upload relevant materials from Step 1 to your jig's knowledge. The more context you provide, the deeper the insights your jig will generate. Make sure to include an explanation about the files you're providing (i.e. where you exported data from) and any other relevant context that might not be in the data itself, such as that week when a water main break in your parking lot cut foot traffic in half. 

Step 3:

Analyze Customer Language Patterns

Let's start with the qualitative data, the rich, unstructured text from reviews, emails, and support conversations. Try this prompt with your Insight Generator jig:

Analyze our customer feedback from the past quarter and identify:
1. The top 5 recurring themes or topics mentioned by customers
2. The sentiment around each theme (positive, negative, or mixed)
3. Specific product features or service aspects that receive frequent mention
4. Any unexpected comments that might reveal blind spots
5. Language patterns that indicate strong emotional responses

The power of this approach is how quickly it can process hundreds or thousands of comments, finding patterns that would take a human analyst days to uncover. Your jig will extract the most common themes (like product reliability or shipping speed), measure sentiment around each one, and highlight unexpected insights that might otherwise be missed.

This analysis becomes your foundation for deeper investigation. Once you've identified key themes, you can ask follow-up questions to drill down:

Based on our customer feedback about product reliability, can you:
1. Identify which specific products are mentioned most often in reliability complaints
2. Extract any patterns in how these products are failing
3. Analyze if reliability perceptions vary by customer segment or geography

This targeted approach transforms the overwhelming volume of customer feedback into clear, actionable insights. Instead of reacting to the loudest or most recent complaints, you're systematically understanding the entire customer experience.

Step 4:

Find the Story in Your Numbers

Now let's examine your quantitative data; the metrics, statistics, and structured information about customer behavior. Try this prompt with your jig:

Analyze our customer data from the past quarter to identify:
1. Significant trends in purchase behavior, usage patterns, or customer retention
2. Correlations between different metrics (e.g., NPS scores and retention)
3. Outliers or anomalies that warrant further investigation
4. Customer segments with notably different behaviors or preferences
5. Leading indicators that might predict future customer behavior

This analysis will reveal patterns that aren't visible in customer comments alone. You'll discover which products have the highest repurchase rates, how geographic location influences buying behavior, or what combination of products customers tend to buy together.

The real value comes from unexpected discoveries: perhaps customers who buy your spring-loaded boxing glove as their first purchase have a 68% repurchase rate, while those who start with rocket-powered roller skates only return 42% of the time. Or maybe canyon-dwelling customers are your most loyal segment, purchasing 4.3 times per year compared to the average of 2.8. (Sorry - can't resist the Wiley E Coyote / Acme Corp references.)

These quantitative insights will help you understand not just what customers say, but how they actually behave—often revealing disconnects between stated preferences and actual actions.

Step 5:

Connect the Dots

The magic happens when you integrate both types of analysis. Ask your Insight Generator to connect the dots:

Based on both our customer feedback analysis and quantitative data:
1. Identify connections between what customers are saying and what our data shows
2. Highlight any disconnects between customer feedback and actual behavior
3. Develop 3-5 key hypotheses that could explain these patterns
4. Suggest 3-5 practical actions we could take based on these insights

This integrated approach will reveal insights neither analysis could provide alone. You might discover that the high return rate for earthquake pills in February aligns perfectly with customer comments about effectiveness issues during unusual geological activity. Or that while customers frequently complain about price, your data shows limited price elasticity during promotions.

These connections lead to testable hypotheses: maybe new customers need clearer guidance on product usage, or perhaps your spring-loaded boxing glove creates a better first impression than your heavily marketed rocket products.

From testing these hypotheses come clear actions, like revising subscription timelines based on actual product longevity data or testing the boxing glove as a discounted entry-point product. Now you're not just understanding customer behavior, you're developing a roadmap for improving your customers' experience.

Step 6:

Design Experiments

With clear insights established, it's time to move from understanding to action. Your jig can help design experiments to test your hypotheses:

For our hypothesis about subscription retention dropping due to product durability, help me design an experiment to test if changing our subscription cadence would improve retention. 

Include:
1. The exact change we should test
2. How we would measure success
3. Potential risks and how to mitigate them
4. What resources we'd need
5. How long we should run the test

This design thinking-based approach transforms insights into testable actions. Instead of making company-wide changes based on hunches, you'll create structured experiments with clear success metrics.

For example, you might test offering a flexible delivery schedule option to customers approaching their 3rd shipment, randomly assigning 50% to receive this offer while keeping the rest on the standard monthly schedule. Your primary success metric would be comparing the 4-month retention rates between the test and control groups.

By running this experiment for 90 days, you'll gather enough data to make an informed decision about changing your subscription model. If successful, you might evolve from "automated equipment delivery" to "personalized pursuit support on your terms"—a subtle but powerful shift that addresses the actual reason customers are canceling.

Step 7:

Build a Continuous Improvement Loop

The final step is to establish a process for regularly updating your insights and measuring the impact of changes. Ask your jig:

Help me create a quarterly process for collecting, analyzing, and acting on customer insights. Include:
1. A timeline for data collection and analysis
2. How to track experiments and their outcomes
3. A framework for prioritizing new initiatives
4. How to communicate insights across the organization

This approach creates a sustainable cycle of improvement rather than a one-time analysis. By establishing a regular cadence, perhaps monthly for reviewing experiment results and quarterly for comprehensive analysis, you ensure insights remain fresh and relevant.

For prioritization, have your jig suggest a framework that balances potential impact, implementation cost, and alignment with business goals. This helps prevent the "shiny object syndrome" where businesses chase interesting ideas without strategic focus.

Insight = Advantage

I have never met a business that said, "We understand our customer too well."

Building a customer insight engine creates a competitive advantage that compounds over time. While your competitors rely on anecdotal feedback or surface-level metrics, you'll develop a deep, nuanced understanding of customer needs that informs every business decision.

For SMBs, this capability was once accessible only through expensive consultants or enterprise-level analytics teams. Now, with generative AI powering your insight engine, you can establish these sophisticated feedback loops with minimal investment. By transforming the mountains of feedback you already have into actionable insights, you can make customer-centered decisions that drive growth, retention, and long-term business success.

✨ ✌🏻 ✨

If you have ideas for future newsletters, I'd love to hear them.