Personalization has been heralded as the key to success for businesses in every industry. Studies show that organizations outperforming their competition attribute 40% of the additional revenue to their personalization efforts.
But delivering personalized messaging at scale across channels is hard to do using only human creative talent. To address the challenge, companies are turning to Generative AI to automate the creation of various forms of content and personalize them for granular audiences. The branch of generative AI focused on creating text is known as natural language generation (NLG), which enables computers to generate human-like language from structured and unstructured data.
In this post, we’ll define what natural language generation is and how it works, and we’ll share some important use cases for NLG in business.
What Is Natural Language Generation (NLG)?
NLG is a sub-capability of natural language processing (NLP), the branch of AI concerned with ingesting unstructured language data—such as written text or voice text from an IVR—and converting it into structured data that a computer can interpret and use. NLP also includes natural language understanding (NLU), the sub-capability that interprets human language to identify what a collection of words and phrases says, while overcoming the challenges of slang, mispronunciation, and syntax. NLG then takes all the inputs from NLP and NLU and produces new language in written or spoken form that is relevant to the context and sounds the way a human would say it—think of how Siri dictates driving directions when you use the GPS on your iPhone, or how Amtrak’s digital assistant “Julie” helps you make train reservations.
How does Natural Language Generation work?
When used in conjunction with NLP and NLU, a natural language generator creates natural, context-appropriate, and helpful language that can be used in marketing communications, in service responses to a customer question or a user request, or anywhere else a business leverages language. And because it’s created by a machine, NLG makes personalized digital engagement possible at exponentially greater scale than would be possible with human creatives—that’s a critical part of the value it delivers.
There are four steps to NLG. They are:
- Data pre-processing
In this step, the AI cleans and normalizes the language data. It also filters the data based on its final purpose (e.g., answering a user query, creating a report, etc.).
- Data understanding
Once the data is prepared, the NLG system interprets it to understand what content to generate in response and how to structure it. Using AI techniques such as machine learning, the NLG system “learns” the relationships between input data and target outputs.
- Content planning
The system decides what content to include in the final output and what order to present it in. This is where an NLG system can get creative and — like a human writer — come up with different ways to convey the required information.
- Natural language generation
In the final step, the NLG system generates natural-sounding text and presents the final output in the right format, whether it’s a chatbot response, an email, or a report.
The business impact of natural language generation
There are many use cases that illustrate how NLG creates a positive business impact. Here are a few examples that demonstrate its value:
Enables hyper-personalization at scale: NLG algorithms analyze data sources and generate text tailored to individual customers. Chatbots that intelligently respond to queries, voice assistants that order your groceries, and conversational AI that can maintain two-way conversations are examples of hyper-personalization.
Increases conversion rates: The right language motivates consumers to click, buy, and subscribe. As noted above, Michaels and Orange used NLG to improve conversion rates in their email campaigns
Drives customer engagement: NLG technology proactively engages with customers and provides them with the information they need when they need it. It’s a critical tool for customer service teams because it helps resolve issues quickly and/or directs customers to self-service materials and information (translating to fewer calls and less interaction from live agents).
Improves customer loyalty and retention: Personalized, relevant content helps build trust and loyalty between a brand and its customers. It improves customer retention by delivering meaningful information at the individual customer level. Upselling, cross-selling, and repeat business become much easier when you have a loyal, engaged customer base.
Reduces operational costs: Automation frees up time and resources across all departments. NLG does this by reducing manual intervention in marketing, customer service, and employee interactions. This helps improve team efficiency, allowing your staff to focus on being more creative and responsive, while improving the accuracy and appropriateness of a given response.
Which departments are using Natural Language Generation?
Businesses use NLG in four key areas to create personalized one-to-one customer interactions and to improve customer experiences: marketing, customer experience, customer service, and digital adoption.
How NLG powers marketing
Automated content generation can drive breakthroughs at every customer touchpoint along the buyer journey, from initial awareness to post-purchase engagement.
Examples of NLG in marketing include chatbots that generate natural-sounding responses to customer questions, automated email marketing campaigns, and dynamic website content. NLG helps increase marketing-led revenue, deliver more qualified leads, and reduce acquisition costs.
How NLG improves customer experience
NLG can help businesses scale personalized engagement while reducing time on manual tasks. This includes everything from responding to customer service requests to making product recommendations.
Given that NLG can help businesses respond to customers across all channels (human and digital), it reduces friction, creating more satisfying customer experiences.
Another bonus: NLG data analysis helps you understand why customers choose to do business with you. Using AI, it analyzes your customer engagement data to predict which ideas, concepts, and messages will resonate with your customers.
NLG delivers better customer service
Automated natural language generation can help customer service teams resolve issues faster and provide better service. For example, NLG can reduce customer support calls by generating natural-sounding responses to customer questions or requests.
NLG alleviates the burden on customer service teams by deflecting issues to digital self-service channels, where customers can resolve common issues on their own. It can also route customer service requests to the right team or agent by understanding the customer’s intent and equipping human agents with the information they need to quickly resolve customer issues.
How NLG facilitates digital adoption
NLG helps customers migrate from analog to digital channels by delivering experiences that are natural, convenient, and easy to use. Automated content generation makes it possible to personalize the digital experience for each customer — without human intervention.
Customers are increasingly demanding digital experiences that are personalized, convenient, and easy to use. By using NLG to generate content, businesses can ensure each interaction is straightforward and personable.
NLG also empowers businesses to use engaging, consistent language across all touchpoints. This speeds up digital adoption by helping customers find the information they need via their chosen channel and device (e.g., SMS, social media, email, or your website).
NLG applications by industry
Businesses across a wide range of industries use NLG to create personalized one-to-one customer interactions and improve customer experience. Here are some industry-specific examples, including a number that use Persado, the Motivation AI platform that leverages language to motivate customers to engage and act:
In banking and finance, NLG converts many types of financial data into human-friendly content, including financial reports, regulatory filings, executive summaries, and suspicious activity reports. It can help onboard customers by teaching them how to effectively use financial products through data-guided language.
BNP Paribas Securities uses NLG to create an executive summary of their (100+ page) Management Information Services book, which includes information about assets under custody, settlement, corporate actions, and income. Chase used NLG to create more effective messaging for its digital mortgage acquisition campaign, increasing mortgage applications by 82%.
In retail and wholesale, NLG generates product descriptions for online shopping and e-commerce. It personalizes customer communication via email, web, chatbot, and SMS.
Michaels used NLG to create more meaningful communication with its customers across email, SMS, and social media. Using NLG technology, Michaels expanded email personalization from 20% to 95% of email campaigns, leading to a 25% improvement in CTR. They also improved engagement and loyalty with SMS campaigns, increasing CTR to 41%.
In insurance, NLG can provide customer-specific underwriting and actionable claims data. It can also streamline and improve customer acquisition by better engaging prospects and motivating them to sign up for a policy.
An insurance industry leader used NLG to produce more engaging language on their quote pages. By pinpointing the most effective and powerful messaging, they increased high-quality leads and ultimately realized a 17% increase in policies.
NLG is a powerful tool that healthcare companies can use to improve patient outcomes — for example, by providing patients with natural language summaries of their conditions that are easy to understand. NLG can also help with provider education and training by generating natural language content from data sources such as EHRs.
Health insurers can meaningfully engage with policyholders by using NLG to generate natural language content like renewal letters, benefit summaries, and educational materials.
Humana helped members thwart the flu by using AI-generated NLG to optimize the language used in its flu-reminder email campaigns. By changing a few words in the subject of the flu-reminder emails — and adding an emotional element — Humana’s open rate for flu emails went from 20% to 31%.
In the telecommunications industry, service providers can use NLG to create usage-based billing experiences for their customers. This type of billing is subscription-based, so it’s important to keep customers updated on their usage to prevent them from incurring overage charges. NLG is also a powerful tool for upselling and cross-selling to existing customers.
Orange, a French telecom provider with over 11 million internet customers, used NLG to identify the best semantic choices for motivating their target audience to open, read, and click on emails. NLG helped improve Orange’s email open rate by over 40% and read rate by as much as 300%. It also increased product orders and helped Orange better understand customer behavior.
Trip planning, booking, pre-trip communications, and post-trip follow-up are areas where NLG can improve the travel experience. NLG gives travel companies the ability to create compelling, personalized communications and campaigns across channels and devices.
It also streamlines the booking process, making it easy for customers to compare dates, rates, and other important criteria while encouraging add-ons and upgrades.
4 examples of NLG models
Markov: The simplest of the four NLG models, Markov chains use a probabilistic approach to predict the next word or phrase in a sequence. They are often used for text generation tasks like chatbots and natural language understanding.
Recurrent neural network (RNN): RNN processes sequences of data, such as text or speech. RNNs attempt to closely replicate the processes of the human brain. They’re useful for NLU tasks like language translation, pattern identification, and object recognition in images.
Long short-term memory (LSTM): A type of RNN used for deep learning, LSTM focuses on remembering long-term dependencies in a sequence. LSTM helps computers understand the context needed to process sequences of data, which is useful in machine translation, summarizing text, and answering questions.
Transformer: A neural network technology, transformers can process inputs of varying length and interpret all the components in combination (rather than sequentially). Examples of Transformer models include:
- Generative pre-trained transformer (GPT), a model that business intelligence systems use to generate reports and related content.
- Bidirectional encoder representations from transformers (BERT), which Google created to recognize speech. It learns human language by analyzing the semantic relationship between words and their meaning.
- XLNet, which is trained using data sets that identify patterns and relationships in natural language. It’s a natural language processing technique that works for tasks such as classifying text and answering questions.,.
NLG can support every business function
NLG is a powerful technology that can transform written and spoken data into compelling narratives. It has near-limitless potential to support a wide range of business functions aimed at improving and refining customer experiences.
Persado’s motivation AI platform enables hyper-personalized messaging across all channels. We employ advanced machine learning and deep learning to generate emotionally-charged messaging that motivates people to engage. Our customers include top brands in industries that include financial services, healthcare, retail, telecom, and insurance.
NLG has the power to improve customer experiences, increase conversions, and drive growth for your business. Explore the impact of language AI with Persado by reaching out to our solutions experts today.