Sequoia Capital called it a firestorm. Gartner says it rocketed to the top of their Hype Cycle. ChatGPT reached 100 million monthly active users (MAU) in just six weeks, feeding into the Generative Artificial Intelligence (Gen AI) frenzy. But for most businesses and consumers, Gen AI hasn’t lived up to the hype.
As Sequoia Capital points out, Gen AI is having a value problem. For example, AI-first mobile apps have a 42% median retention rate after the first month. Even worse, the median DAU/MAU ratio for AI-first mobile apps is 14%. That’s a long way from some of the best-performing apps, which come in at 60-65%, or top-rated WhatsApp at 85%.
While both consumers and businesses are excited about the potential of AI, they’re not finding what they need for practical use. Maybe it’s because leading with technology is never going to be as meaningful to businesses as solving a problem for them.
That’s why Chattermill avoided a “tech-first” approach with our proprietary technology, Lyra AI. Long before the Gen AI hype started, we tackled problems organizations encounter when trying to find meaning in their customer experience (CX) data.
Lyra is a powerful blend of AI methodologies allowing companies to analyze customer feedback to answer business questions. It powers the Chattermill customer feedback analysis platform that allows CX and business leaders to discover what customers are really thinking.
Being problem-led instead of tech-led in product design and development is the answer to the first question we often get asked, “How are you different?”
Now that you know the primary driver behind Lyra, let’s take a look at other frequently asked questions.
Lyra capabilities
What does Lyra do?
Lyra leverages AI and natural language processing (NLP) to reshape customer feedback from surveys, reviews, support tickets, social media, chatbots, and custom data into actionable insights. Lyra responds to queries in human language to give users quick, accurate insights into customer feedback, at scale.
Lyra continually strives for high accuracy and precision in analyzing customer feedback.
Who uses Lyra?
Lyra is embedded in the Chattermill customer feedback analysis platform and is not a stand-alone product. Leaders across any organization benefit from Lyra because they can make informed decisions based on insights into what customers are thinking and feeling.
Can Lyra handle open-ended text?
Yes, Lyra is excellent at analyzing written text left in places like reviews, chats, surveys, and email from both internal and external data sources. It determines a score for text-based reviews that it harmonizes with typical 5-star reviews to give depth and meaning to how customers are feeling about your brand.
How does Lyra help me make sense of my customer feedback?
Lyra quickly and reliably analyzes your data to help you identify trends and emerging problems and get to the bottom of customer pain points faster. You can monitor risks in nearly real-time with anomaly detection and identify the “why” behind customer reviews through AI-generated summaries. Lyra highlights what’s working well in your current CX ecosystem and where you need to focus your attention for improvements.
Lyra powers Chattermill’s Insight Assistant to give you on-the-spot analysis for any query. You can also visualize your data in tables, charts, and graphs so you can monitor your key areas of interest at a glance.
How Lyra works
What makes Chattermill’s Lyra stand out from other AI technology for CX?
In a nutshell, Lyra has a unique approach to deriving actionable, accurate, and contextually rich insights from customer feedback. It integrates sophisticated techniques and technologies, building on nearly a decade of product development using best-in-class methodologies for each feature.
For a detailed look at how Lyra is the only model for CX that anables both strategic and tactical levels, head over to our article, The Strategic and Tactical Divide in CX: How We Built Lyra AI to Bridge the Gap.
What’s unique about Lyra’s algorithm?
Lyra has distinctive capabilities not found in alternative approaches. These include:
Nuanced sentiment accuracy – Lyra can identify multiple themes in a single comment and assign sentiment to each one. This ensures high accuracy and relevancy for each piece of text. Lyra provides a more precise and targeted understanding of sentiment at a granular level and even higher topic accuracy.
Reviews are full of shades of grey that most CX AI technology can’t distinguish. Reviews often have both positive and negative observations, but many technologies can only assign a single sentiment to a text review like this:
“While our food was good value for money, the service we received was slow. It upset my dinner companion but I liked that we weren’t rushed.”
Lyra can distinguish good, bad, and even neutral sentiment in a single review. That level of granularity gives CX professionals a much more accurate view of what’s happening in their business.
Excellent recall – Lyra consistently returns the same analysis regardless of how many times you query the data.
Mutually exclusive topics – Lyra ensures that similar themes are not double-counted, by providing a clearer and more accurate understanding of customer feedback.
Collectively exhaustive topics – Lyra ensures that no significant theme or issue is overlooked. It captures and represents all relevant CX topics in customer feedback, to provide a holistic view of your customer experience.
Robustness over time – The Lyra model is highly scalable. It can handle large volumes of text data and keep pace with the changing nature of language and customer behavior. This means Chattermill can be applied to various applications at scale without increasing workload or resource requirements.
Descriptive insight analysis – Lyra generates concise summaries of large datasets. The information provided is described in detail to make it highly actionable.
In-depth analysis of text – Lyra efficiently captures the essence of longer and context-rich phrases, providing comprehensive insights without the need for manual review of individual comments.
Customization – Lyra can be trained to improve performance for your business. You can manually add and manage custom themes or topics so you, and Lrya, can stay on top of emerging trends. Customization can be performed by account users at any time.
Trustworthiness
Is Lyra 100% accurate?
No. No AI technology or machine learning system can anticipate with 100% accuracy the infinite ways people use language to express their likes and dislikes. Someone will always invent a new way to misspell a word or come up with new slang that Lyra won’t know about. But, Lyra is a quick learner!
Lyra outperforms human annotators who operate at about an 80% accuracy rate. It handles big data at scale and has excellent recall, which means you won’t get a different answer every time you make the same query. Chattermill’s focus is on precision and consistency in feedback analysis and we do it better than any other technology in the CX space.
How do I know the data is about my company and not general industry or sector information?
Lyra AI does not work as a black box. While AI technologies can be complex, Chattermill feels strongly that customers should be able to easily drill down into their own customer feedback. You can slice and dice your data using filters. You can identify trends and view your data using charts and graphs to help you quickly assess the state of your own customer experience.
What if there’s no data for a particular time period or topic?
Lyra explicitly tells you when there’s no customer feedback meeting your criteria.
What if my data is mis-tagged or mis-labeled?
Feedback can occasionally be mis-tagged or assigned a wrong label, something that happens when dealing with vast quantities of data. Chattermill customers have the ability to make changes and customize data, which helps Lyra learn.
Privacy, security, and compliance
How does Lyra protect the privacy of people who leave reviews?
Privacy is a critical concern for all Chattermill customers. Chattermill has a strong Personal Identifiable Information (PII) removal stage up front and excludes all identifying details from their product. The information we use has been anonymized and doesn’t contain personal data like email addresses, phone numbers or physical addresses in it.
What kind of compliance regulations do you follow?
It’s a long list because we have developed Lyra to be compliant with international standards, including:
- GDPR
- Market Research Society Code of Conduct
- Data Protection Act.
Powering Lyra with your feedback data
How is a new customer set up?
Once we have the data for a new customer, the first thing we do is unearth key themes. For this task, Lyra uses semi-supervised learning, incorporating thematic analysis and neural embeddings to understand the key groups or clusters people are talking about when they leave reviews.
Next, Lyra takes those clusters of text and determines labels and assigns theme names. That’s all done automatically. At the same time, Lyra is extracting phrases from customer feedback. It groups those phrases in a contextually smart way. This is where Chattermill’s proprietary technology excels in creating true CX Intelligence.
To learn more about how to get started with Lyra, take a look at our article, Quick-start guide for Lyra AI.
Future developments for Lyra
What’s the next feature we can expect for Lyra?
CoPilot is currently in beta testing and is expected to be released later this year. It means Lyra will provide one-to-one insights about CX Intelligence. It leverages Gen AI so CX professionals can ask a question in human language and get everything they need from their customer feedback data in seconds. It’s the next evolution of Insight Assistant.
What’s on the horizon for new developments in the longer term?
Our aim is to let customers work with more independence. They’re eager to do much more themselves and have more flexibility with the product, whether it’s theme management or defining topics or anything else they can control. We’re building autonomy into the product while ensuring Lyra AI maintains a high quality of accuracy, precision, and recall.
