AI-Powered Insights: How to Find the Needle in a Haystack of Customer Conversations
New generative AI applications help process vast stacks of conversations into business insight needles.
A business is a conversation. The company provides a product; customers call, email, or chat with questions, concerns, complaints, or praise. Business opportunities, risks, and threats are buried in those conversations like needles in a haystack.
Conversation haystacks are everywhere, ever-flowing and ever-changing. Even a small business has hundreds of interactions with customers a day. Big companies have millions. These mounds, left unexamined by most companies, are strewn all over, waiting to be turned into business opportunities.
Conversations: a haystack of hidden business value
Imagine a conversation between a call center agent and your customer, John. John agrees to let you transcribe the interaction to improve service. You activate a speech-to-text translation tool and get a mound of text.
It goes like this:
It’s easy for a human to spot what’s going on here. John is older, unhappy, and doesn’t get the help he needs. The agent told him he needed to make a voice call, and the customer said they would call tomorrow.
The problem is that most companies sample and check less than one percent of their conversations with customers because they have too many to evaluate.
Could we do better?
Unleashing the power of AI to reveal buried opportunities and risks in conversations
AI can help automate the discovery of customers like John and flag them for human intervention. AI can check millions of conversations every day, asking questions like, “Why did this customer contact customer service?” AI responds in the blink of an eye:
“The customer was trying to get their home phone hooked up but had been unsuccessful despite the agent providing all the necessary information.”
The power of AI is prediction. A textual summary could be generated, and a “health score” assigned to help humans prioritize where to intervene first.
And if we prompt AI to evaluate, "Is this customer vulnerable?” we get an even more actionable prediction:
“Yes, the customer is 73 years old and cannot function quickly. This suggests they may be vulnerable or require additional assistance due to age.”
This is a probable instance of “customer churn,” the rate at which a business loses and keeps customers. A Bain study showed that reducing churn by 5% can increase profits by 25% to 95%. So, finding and stopping customer churn is essential for all businesses.
Call centers are just one of many conversation haystacks
Call centers are just one example of conversational analytics at work. Talk is everywhere: between customers and salespeople, chatbot exchanges, Q&A during webinars, on social media, and more.
Digital versions of these conversation haystacks can help find insight needles like,
How clients react to offers by evaluating what they say,
The root cause of product problems,
Finding vulnerable customers based on their tone,
Correct unintended promises or expectations by staff,
Explore sales interactions for win and loss signals,
Spot upsell opportunities in unconventional data sources like support calls,
Analyze these in multiple languages with language translation technology.
The endless possibilities are why many investors believe conversation analytics is among the most valuable new markets and fertile ground for innovation. Talkmap itself has seven patents and many more pending in this exciting field of research.
New technologies make conversational intelligence possible
Why doesn’t every company use AI to analyze conversations? Because until now, the technology hasn’t been advanced enough, simple enough, or cost-effective enough to accomplish this needle-in-a-conversation-haystack trick. But that’s changing quickly.
For a few years, I’ve had a front-row seat watching Talkmap solve this problem, and this example comes Conversational Researcher and Designer David Attwater’s awesome webinar series, What’s so Hard about Conversational Analytics…?
Next, David will explain why implementing conversational analytics isn’t as simple as shoving conversation haystacks into a general-purpose LLM. Register here or subscribe below to learn more.
Examples from the webinar What’s so Hard about Conversational Analytics…? from conversational intelligence company, Talkmap.
Full disclosure: I’m on Talkmap’s board.