With LLMs talk is no longer cheap

By

Bartek Kuban

November 6, 2024

•

6

min read

What's inside

Imagine discovering that 15% of your abandoned carts happen because customers can't find your international shipping rates.

Or learning that customers absolutely love your product's sustainable packaging — a feature you've never even promoted.

These are the kinds of insights that traditionally required conducting customer interviews and surveys. But today, Large Language Models (LLMs) are transforming how businesses understand their customers through automated conversation analytics.

Instead of relying on traditional surveys and surface-level analytics, companies can now automatically analyze millions of customer interactions, reaching a new level of understanding of all our customers.

Uncovering the "Why" behind customer behavior

Traditional digital analytics tell you what customers do — they clicked this button, spent 2 minutes on that page, and abandoned their cart.

But they don't tell you why.

Consider these limitations:

A customer spending 25 minutes on your FAQ page might be thoroughly informed — or completely confused
High cart abandonment rates don't explain if it's price, shipping, or a broken “Checkout” button
Product page bounce rates don’t reveal if customers found better alternatives or simply couldn’t find information about the materials used to make the product

Website analytics leave these crucial questions unanswered.

We need to let customers tell their story — in their own words

In the AI Age, customers can express their needs naturally. Instead of forcing them through rigid feedback forms, they can simply say: "I'm looking for a non-stick pan that works with induction cooktops and won't scratch easily." This single sentence reveals:

Specific product requirements
Technical compatibility needs
Durability concerns
Implicit previous experience with scratched cookware

Like our Insights Module that reveals findings from each customer interaction

Imagine scaling this level of personal understanding.

For example, an online home goods retailer might notice a sudden increase in questions about “flexible storage solutions for small spaces.” Rather than just reacting to this trend, the company could leverage LLM-powered conversation analysis to uncover the underlying motivations — perhaps customers are adapting to remote work setups or downsizing due to economic shifts.

Similarly, an unexpected spike in complaints about a specific product feature could signal a quality control issue that requires immediate attention. Just as some companies use sensor data to proactively address vehicle problems, brands can leverage conversation analytics to spot emerging issues before they turn into a PR crisis.

LLMs can extract emotional insights that traditional reviews often miss

Take for instance a skincare brand launching a new eco-conscious moisturizer. They might find that 70% of conversations express excitement and relief about the sustainable ingredients and cruelty-free commitment. Customers aren’t just satisfied — they’re proud, sharing sentiments like:

“Finally a moisturizer that doesn’t make me feel guilty!”

Meanwhile, a smaller but vocal group raises concerns over product texture and absorption, revealing frustrations:

“I want to love it, but it feels too heavy on my skin,” or,

“The scent is a bit strong — wish it were gentler for sensitive skin.”

Curiosity and hints of skepticism also emerge as some users wonder:

“Does it really hydrate as well as my regular moisturizer?”

“I hope this isn’t just another greenwashed product.”

This level of emotional insight allows the skincare brand to go beyond surface-level sentiment and understand what drives customer loyalty, hesitation, and excitement. In response, they can emphasize the effectiveness of their ingredients, consider developing more skin type options, and ensure transparency.

The journey of LLM-powered conversation analytics doesn’t end with understanding. It culminates in delivering truly personalized experiences.

Just as Spotify uses listening habits to curate playlists, brands can leverage chat histories and other interaction data to provide tailored product recommendations, content, and support.

An electronics store using an LLM to recognize that a customer frequently asks about “gaming laptops with high refresh rates” and prefers “budget-friendly” brands. Based on this, the LLM could suggest specific models, accessories like cooling pads, or even notify the customer when a compatible gaming mouse goes on sale.

But let’s shift from exploring the possibilities to what’s proven.

Some real-life conversation analytics solutions we’ve implemented for our customers

Knowledge Gaps

One of the most valuable uses of conversation analytics is in identifying knowledge gaps — when customers ask questions that aren’t covered by existing FAQs, knowledge bases, or documentation. For instance, an LLM might detect that customers frequently ask, “What’s the return policy for items on sale?” even if it’s not explicitly listed on the website. Every unanswered question is an opportunity for improvement. LLMs can detect these knowledge gaps, helping businesses expand their content and ensure that future customers have the information they need at their fingertips.

User intents

LLMs can classify user intents such as inquiries about product features, complaints, or requests for support. This classification allows businesses to route conversations to the right support channels or automate responses for common queries, enhancing the overall customer experience. For example, an online electronics retailer might use LLMs to detect when users intend to “request a refund,” allowing the system to prioritize these cases.

Trending topics

With LLMs, businesses can monitor what topics or phrases are frequently appearing in user conversations. Are customers discussing a new feature, a common issue, or a popular product category?

Anomaly and issue detection

One of the most valuable aspects of conversation analytics is the ability to spot anomalies — unexpected patterns that might indicate a problem. A sudden spike in conversations around product returns or complaints about a specific feature can be detected early.

In the case of a tech product launch, an LLM might detect that users frequently mention “battery overheating” in support chats — a sign of a potentially serious issue. AI-driven anomaly detection enables businesses to catch product defects or emerging issues in real time, allowing them to address the problem before it escalates.

Sentiment analysis

Sentiment analysis allows businesses to understand the emotional tone of customer interactions. Are users frustrated, satisfied, or confused? You can even break down this analysis in various ways, as shown in the localization chart below.

Popular product categories and products

LLMs can analyze which products or categories are mentioned most frequently in conversations, shedding light on customer preferences. This insight allows businesses to understand which products generate the most interest. They can identify seasonal trends or shifts in customer preferences to optimize their offerings.

User language and localization

With LLMs, companies can identify the language preferences and cultural nuances of their users. Understanding the localization needs of different customer segments enables businesses to tailor their messaging and support to resonate with diverse audiences.

LLMs enable businesses to set up custom notifications based on specific conversation patterns, allowing them to stay in the know and respond quickly. If a significant number of customers mention a specific issue, like delayed shipments, the AI can trigger an alert, prompting the support team to investigate.

This real-time capability transforms customer support from a reactive function into a proactive one.

Essentially, your business can be placed in a continuous learning loop where customer interactions constantly feed back into the system, creating a virtuous cycle:

Customer expresses need or concern
LLM analyzes and categorizes the interaction
Insights drive product/service improvements
Improved offering leads to new customer feedback
Cycle repeats with richer data

Every user interaction, every message, becomes part of a growing body of insights that improve with time. As LLMs process more conversations, they become better at identifying patterns, predicting issues, and generating recommendations.

They make every customer conversation an opportunity to learn.

Bartek Kuban

Similar articles

Artificial Intelligence

How Model Context Protocol works. MCP Explained

Still unsure what MCP actually does? Let’s clear it up, one step at a time.

Artificial Intelligence

Conversation is the new distribution: On media and AI

Newsrooms once feared being "dumb pipes" for Google. Now, they risk becoming ghostwriters for AI chatbots.

February 23, 2025

Artificial Intelligence

What makes an LLM generative? AI copyright law and the Schrödinger's cat

Do Large Language Models create original text or just perform high-tech copy-paste? Lawyers may have to settle that debate before technologists and philosophers even get the chance.

Piotr Grudzień

February 15, 2025

Artificial Intelligence

Generative Search Optimization: How to rank higher in ChatGPT?

ChatGPT, Perplexity, and DeepSeek are becoming the new gatekeepers of online visibility. Let’s break down how this works and what you can do to show up.

Piotr Grudzień

February 10, 2025

Artificial Intelligence

We tested DeepSeek. Here’s what you need to know

Some say DeepSeek R1 is a game-changer, others say it’s impractical. Both are right.

Mateusz Jakubczak

January 29, 2025

Artificial Intelligence

What’s wrong with outcome-based pricing models?

How they can hold customers back and work against them.

December 16, 2024

Artificial Intelligence

On AI hallucinations

"Hallucinations” reveal more about us and our systems than the models themselves.

December 6, 2024

Artificial Intelligence

With LLMs talk is no longer cheap

How Conversation Analytics turn every customer interaction into a chance to learn something new about your business.

November 6, 2024

Artificial Intelligence

Rethinking Car Servicing With Carmen [Co-Authored Post]

The automotive service industry has long been bogged down by slow, manual processes — think long phone calls, missed appointments, and outdated scheduling systems. It's time for a change.

September 24, 2024

Artificial Intelligence

8 Real-World Examples of Conversational AI Use

Dive in to see how companies like Quickchat AI, Qualtrics, Ticketmaster, and others are using conversational AI in different parts of their business.

Artificial Intelligence

Reviewed: 5 Best Enterprise Search Software Right Now

Tired of cluttered intranets? Look no further — these enterprise search companies provide all the information you need in a chat interface.

Artificial Intelligence

5 Best Enterprise AI Chatbots (For Serious Business Applications)

Here are the best enterprise AI chatbots along with their differentiating features.

Artificial Intelligence

LLM benchmarks: What are they and can you trust them?

Zuzanna Dubanowska

Artificial Intelligence

What are AI Hallucinations? It’s a feature, not a bug

"What are they, what causes them, and what can you do to prevent that problem... but is it really a problem?"

Artificial Intelligence

AIaaS: The What, Why, And Wow of AI as a Service

We're witnessing the rise of AIaaS — Artificial Intelligence as a Service. Learn what AIaaS is and why should you care.

February 23, 2024

Artificial Intelligence

RAG vs Fine-tuning for your business? Here's what you need to know

Considering custom AI solutions for your business? Discover how Retrieval Augmented Generation (RAG) offers a flexible and affordable alternative to fine-tuning LLMs.

Zuzanna Dubanowska

February 22, 2024

Artificial Intelligence

AI Chatbot Buyer Guide: 6 crucial factors to consider

Looking for the perfect AI chatbot for your business? Read this and learn about the 6 unobvious factors that you should pay attention to.

January 31, 2024

Artificial Intelligence

Is ChatGPT a real polyglot? Problems with multilingualism of LLMs

Let's talk about the limitations of Large Language Models in terms of their multilingual capabilities, why things are the way they are, and what can you do about it.

Zuzanna Dubanowska

January 23, 2024

Artificial Intelligence

GPT Tokens Explained - what they are and how they work

Large language models generate text token by token. But what are tokens exactly? And how do they relate to entropy and data compression?

Piotr Grudzień

October 6, 2023

Artificial Intelligence

GPT-4: How Multimodal Learning Takes Us Closer to Human-level Performance

Let's investigate how the integration of multi-modality in the newly released GPT-4 will enhance its knowledge and capabilities.

Maciej Kowalski

September 29, 2023

Artificial Intelligence

One Thing all AI Breakthroughs have in common

All recent AI breakthroughs had one thing in common: a clever way of generating a huge amount of accurately labelled data.

Piotr Grudzień

September 29, 2023

Artificial Intelligence

ChatGPT: what it is, use cases and limitations

In today’s blog post we will introduce you to ChatGPT - the latest project published by OpenAI. We will explain what it is, suggest some use cases but also mention important limitations.

September 29, 2023

Artificial Intelligence

3 Hard Truths about Generative AI

At Quickchat AI, we think and talk about Generative AI a lot. There are some things that rarely get mentioned in popular content on the topic.

Piotr Grudzień

September 29, 2023

Succeed in the new world redefined by AI.
Don't get left behind.

‍7-day free trial · No credit card required

‍Start with a 7-day free trial to build and customize your AI Agent. Custom conversational AI solutions available.

Start building for free