The AI Edge: Building AI Automation for Deep Customer Insights – Dan Sanchez

I recently spoke with Marc Thomas, Senior Growth Marketer at Podia, who’s built something that makes me genuinely excited about the future of marketing: an AI-powered customer research system that’s changing how his company builds products.

What’s fascinating isn’t just the technology—it’s how practical and immediately useful this approach is for businesses of any size.

Let me walk you through how Marc built this system, why it works, and how you can create something similar for your business.

The Customer Research Problem Nobody Talks About

Every marketer knows they should be doing more customer research. But here’s the dirty secret most won’t admit: valuable insights get lost in the chaos of disconnected tools and platforms.

As Marc explained, the problem at Podia was simple but significant:

“In your business, there are so many sources of customer insight. There’s surveys, phone calls, in-app surveys, screen recording analysis, and it’s all over the place. Some of it was in one survey tool account like Typeform, some in Hotjar, some in our in-app survey tool, and then we’ve got Fireflies for recording phone calls.”

Sound familiar? The issue wasn’t a lack of data—it was that this data was scattered across a dozen tools, making it nearly impossible to see patterns or extract meaningful insights.

The result? Companies make decisions based on incomplete information, or worse, ignore valuable customer feedback entirely because it’s too much work to process.

Building a Central Repository (Without Breaking the Bank)

Marc’s first move wasn’t to buy an expensive enterprise solution. Instead, he started with something much simpler: Google Drive.

“We ended up putting all of this data into Google Docs and Google Sheets. They’re just incredibly effective,” Marc shared.

The key insight here is to use what your team already works with. For Podia, that was Google Drive. For your company, it might be Microsoft 365, Notion, or something else entirely.

What matters is creating a system that:

Centralizes all customer feedback in one searchable location
Creates a consistent format for feedback regardless of source
Makes it easy for anyone at the company to access insights

This foundation is critical before adding any AI capabilities. As Marc put it: “How can I make this system work with the behavior that I am already doing or my team is already doing?”

Where AI Changes the Game

Once Marc had established a basic system for collecting customer feedback, he started experimenting with how AI could enhance the process.

His first test was simple: could AI help summarize and classify individual pieces of feedback? He built a small custom GPT to test the concept, feeding it survey responses and asking it to:

Create an 8-word title summarizing the feedback
Write a 30-word summary of the key points
Categorize the feedback from a predefined list
Generate searchable tags for the response

This initial test showed promising results. The AI could effectively distill lengthy customer responses into searchable, scannable formats that made it much easier to identify patterns.

But the real breakthrough came when Marc automated this process using Zapier.

Automating Everything with Zapier + ChatGPT

“We built steps into Zapier which used ChatGPT directly,” Marc explained. “That really is when everything went from good to this is honestly brilliant and has made our lives 100 times easier.”

Here’s the basic workflow he built:

When a new survey response comes in (from Typeform, for example), Zapier creates a record in a spreadsheet
Zapier sends the response to ChatGPT for analysis
ChatGPT returns a title, summary, category, and tags
Zapier updates the spreadsheet with this analysis
Zapier creates a permanent document record of the full response
The document is stored in the appropriate folder structure
The spreadsheet is updated with a link to the permanent record

This system processes thousands of pieces of customer feedback each week with minimal human intervention. Everything is automatically categorized, summarized, and made searchable.

The Secret to Better AI Analysis: Break It Down

One of the most valuable insights Marc shared was about handling longer content like call transcripts. Instead of feeding entire transcripts to AI in one chunk (which often hits token limits and produces mediocre results), he breaks them down:

“We basically summarize small chunks of text in the exact same process that we would summarize a whole survey response. And then we store those all in an array of responses. Then we use a parent Zap to basically summarize the summaries.”

This approach has two major benefits:

It avoids hitting token limits in AI models
It produces much higher quality summaries by preserving important details

As Marc noted: “The quality of the output from summarizing lines of a transcription is far superior to summarizing chunks. It’s going to cost you a little bit more, but the end result will be better by a lot more than the margin of cost.”

Using AI to Verify Your Own Analysis

Perhaps the most interesting application of AI in this system isn’t just automating busy work—it’s using AI as a check against human bias.

When Marc prepares his weekly research reports, he’ll export filtered data from the database and upload it to ChatGPT. Then he asks questions like:

“What are the most common feature requests grouped by category?”
“Give me specific examples of what people have said about design options”
“How do people talk about Podia when they mention competitors?”

Then he’ll compare AI’s interpretation with his own:

“I’ll say, ‘Here’s how I often interpreted this data. Compare and contrast our outputs to see who’s right.’ It’s almost like you’ve got a coworker there who you’re having a debate with.”

This process of “arguing” with AI helps surface insights that might otherwise be missed and ensures that the final analysis is as objective as possible.

Results: From Weekly Reviews to Continuous Insights

Beyond the efficiency gains, this system has transformed how Podia approaches customer research. They now have:

A Slack channel that updates the team with new insights in real-time
Weekly reports that identify patterns and trends across thousands of feedback points
A searchable database of customer feedback that anyone can access
A much deeper understanding of customer needs and pain points

As Marc put it: “If you want to level up your marketing career, becoming the person who understands the customer best is certainly the way to do that.”

Building Your Own AI-Powered Research System

The beauty of Marc’s approach is that it doesn’t require enterprise software or a dedicated data science team. You can build something similar with tools you probably already use:

Start with centralization: Create a simple Google Sheet or other database to track all customer feedback
Connect your feedback sources: Use Zapier to pull in data from survey tools, help desk software, etc.
Add AI analysis: Use Zapier’s ChatGPT integration to automatically summarize and categorize feedback
Build a reporting cadence: Set aside time weekly to review trends and share insights

Marc has even created a course that walks through this entire process, which you can find at positivehumanco.com.

The True Power of AI in Marketing

What I find most valuable about Marc’s approach is how it challenges our assumptions about AI’s role in marketing. As he noted:

“A lot of people are seeing AI uses in marketing as generative—’I’m creating an image for my social media.’ But marketing isn’t really about social media. There’s a lot of organizational, informational processing tasks. And AI is fantastic at that stuff.”

This is the kind of AI application that excites me most—not just automating content creation, but helping us understand our customers better and make smarter strategic decisions.

By combining human intuition and expertise with AI’s ability to process large amounts of information, we can build marketing systems that are both more efficient and more effective.

If you’re interested in learning more about implementing AI marketing systems like this in your business, I’d love to hear about your experiences and challenges.