Mastering Generative AI from Scratch

Chapter 2: Data Doesn't Lie, But It Sure Can Be Loud

PetGlam's office was unusually quiet, except for the rhythmic clack of Ivy's keyboard and the occasional slurp from her bubble tea. On her screen: a flood of Instagram comments, customer form responses, and screenshots of inbox messages—all part of her weekly content report.

She sighed.

“This is chaos,” she muttered.

Derek walked by with his second coffee of the day and arched an eyebrow. “Still wrestling with last week's engagement data?”

Ivy spun her chair around. “You mean the avalanche of emojis, random complaints, and one lady asking if she can feed our shampoo to cats? Yes. I don’t even know what counts as useful.”

Derek chuckled and leaned on her desk. “Welcome to the world of unstructured data.”

She blinked. “Un...what?”

“Unstructured data,” he repeated. “Basically, all the messy, free-form stuff—comments, emails, DMs, reviews. Stuff that doesn’t come neatly packaged in a spreadsheet.”

“So what's structured then?” she asked.

He tapped her screen. “That Google Form result you pulled? Click-through rates? Ad spend versus conversion ratio? That’s structured. It's labeled, organized, and the AI can process it directly.”

Ivy tilted her head. “And the messy stuff? How do we even begin?”

Derek grinned. “We ask the AI to help.”

✦✦✦

Fifteen minutes later, Ivy was on ChatGPT, pasting a batch of comments into the prompt window.

Prompt:

“Please summarize the overall sentiment of these customer comments. Identify common themes, complaints, or ideas that repeat.”

As the AI typed back:

“Most comments reflect positive sentiment toward product packaging, especially the pink holiday edition. Several users suggest offering a subscription model. A few negative comments express concerns over delivery delays.”

Ivy blinked. “This... is gold.”

“Keep going,” Derek encouraged. “Try clustering them by topic next.”

She fed in more data. Soon, she had themes: packaging appeal, shipping concerns, product scent, return policy confusion. All previously just noise—now organized.

✦✦✦

Two hours later, Ivy cross-referenced these themes with the latest Facebook Ad results.

“That ad with the unboxing video?” she said. “It had the highest ROAS—and people keep commenting about the cute bow on the packaging. That’s probably why.”

“Exactly,” said Derek, pulling up a stool. “You’re doing insight extraction. This is how we turn interaction into strategy.”

She crafted a short report summarizing the AI analysis and ad performance insights. It felt... real. Like she finally saw the connections.

And for the first time, when she looked at a messy comment section, she didn’t see chaos—she saw opportunity.