Article • 1 min read
A beginner’s guide to generative AI for business
Generative AI can help streamline workflows, improve CX, and enhance agent performance. Learn how to leverage these AI-powered tools in our guide.
Por Hannah Wren, Staff Writer
Última atualização em October 22, 2024
What seems like a long time ago, in a galaxy far, far away, humans existed without the internet. In just a few short decades, the internet went from existing in science fiction media to a common tool with over 5 billion users worldwide. Recent technological breakthroughs have introduced generative AI to the masses, putting it on a faster track to popularity than the World Wide Web.
According to the Zendesk Customer Experience Trends Report 2023, 65 percent of business leaders believe the AI they use is becoming more natural and human-like—and it’s only going to get better. As more businesses begin implementing generative AI to improve customer service and enhance the employee experience, it’s important to understand how to harness the power of this novel AI technology to its full potential.
Our comprehensive guide breaks down the types, applications, examples, benefits, and challenges of generative AI, so you can maximize your overall customer experience now and in the future.
Table of contents:
- What is generative AI?
- How does generative AI work?
- Who uses generative AI?
- Traditional AI vs. generative AI: What’s the difference?
- Types of generative AI models
- What are some popular generative AI applications?
- Benefits of generative AI
- Challenges with generative AI
- How to use generative AI for customer service
- Frequently asked questions
What is generative AI?
Generative artificial intelligence (GenAI) is an AI-powered technology that uses extensive libraries of information to generate new things, like stories, pictures, videos, music, and software code.
For a more technical generative AI definition, Forrester describes it as “a set of technologies and techniques that leverage a very large corpus of data, including large language models (LLMs) like GPT-3, to generate new content.”
How does generative AI work?
Generative AI uses machine learning (ML) algorithms to analyze large data sets. That means you can feed artificial intelligence a bunch of existing information on a topic, so it can learn and find patterns and structures. Based on what it learns from this data, generative AI can create new, original content.
For example, you can ask ChatGPT—the popular text-to-text generative AI chatbot by OpenAI—to write a sonnet about how Luke Skywalker feels about Darth Vader being his father in Shakespearean iambic pentameter. ChatGPT will quickly gather relevant information on the internet about the Star Wars saga and William Shakespeare’s sonnets and instantly produce new content from what you’ve requested.
And yes, we actually asked ChatGPT this question. Here are the results:
“Upon the starry canvas of my fate,
A revelation dark did cast its shade,
In shadows deep, a truth I did berate,
My father, once a Jedi, now betrayed.
In crimson mask, he sought to rule the skies,
With heavy breath, he walked a twisted path,
Yet in his heart, a shred of good beliefs,
A conflict fierce, an inner war’s cruel math.
O father, kin by blood, but not by deed,
Thy presence haunts my dreams, my waking days,
A legacy of darkness, dark indeed,
But hope persists, a light in Sith’s malaise.
Though Vader’s grip on me is strong and tight,
My love, my faith, may yet redeem the knight.”
Who uses generative AI?
Businesses of all sizes—startups, small businesses, mid-sized businesses, and enterprises—use generative AI in different ways. Businesses may use it to streamline and enhance customer support, sales, marketing, IT, development, HR, and training teams. Some examples of generative AI use cases include:
- Enhancing the existing abilities of customer support agents with AI-powered assistance
- Analyzing large amounts of data for more accurate lead scoring and sales forecasting for sales teams
- Personalizing marketing communications
- Optimizing data center operations for IT departments
- Generating code for software developers
- Creating and updating internal content and documents for human relations (HR) departments
- Streamlining onboarding and agent training
These generative AI examples are just the tip of the iceberg. As generative AI becomes more mainstream, businesses will find more and better ways to implement the technology.
Traditional AI vs. generative AI: What’s the difference?
| Traditional AI | Generative AI | |
|---|---|---|
| Objective | Task-specific and rule-based | Content generation |
| Learning | Uses predefined programming | Identifies patterns from large datasets |
| Output | Task-specific | New content or data samples |
The difference between traditional AI and generative AI is that traditional AI uses machine learning, predefined rules, and programmed logic to perform specific tasks, whereas generative AI learns from large datasets to create human-like content. For example:
- Traditional AI can make ticketing systems more efficient by identifying the customer sentiment, intent, and language of service requests, automatically routing them to the right agent based on predetermined criteria (such as expertise, capabilities, and availability).
Generative AI boosts agent productivity by providing intelligent writing tools, allowing teams to address requests more efficiently and provide consistent support.
Businesses can use both traditional and generative AI to analyze data. While traditional AI can make educated predictions based on the data, generative AI can create new data based on the provided datasets. Generative AI can also adapt to context and produce unique, creative content.
Generative AI vs. machine learning
The difference between machine learning and generative AI is that machine learning isn’t limited to generative tasks. Both types of AI learn from patterns found in large datasets and interactions, but machine learning makes predictions or classifications and doesn’t generate new content.
Types of generative AI models
Generative AI has various use cases, meaning there are many different types of generative models. Here are some of the most common types of generative AI models.
Generative adversarial networks
Generative adversarial networks (GANs) work by training two different learning computers (called neural networks) on the same datasets to generate increasingly more realistic content over time.
The two networks, called the “generator” and the “discriminator,” compete against one another, pushing each other to continuously create better content. Once the GAN receives the same information, the generator creates a data sample (like an image or text) based on the training data. The discriminator then analyzes what the generator created and determines if it’s real or generated data.
GANs are like two players competing in a game. Let’s use Star Wars droids R2-D2 and C-3PO as the competitors.
The game consists of R2-D2 (the generator) creating images of Ewoks, the Millennium Falcon, and other things from the Star Wars universe. C-3PO (the discriminator) examines these images and decides if they look real or fake, just like a Jedi inspecting a lightsaber to see if it’s genuine.
As they keep playing the game, R2-D2 gets better at making the images more realistic, based on C3PO’s feedback.
Transformers
Transformer-based generative AI models are neural networks that use deep learning architecture (algorithms to find patterns in large amounts of data) to predict new text based on sequential data. Transformers can learn context and “transform” one type of input into a different type of output to generate human-like text and answer questions.
Think about the auto-suggest feature on messaging apps. Say Han Solo wants to send Princess Leia a text message. As he starts to type, generative AI predicts the next word in his typing sequence and offers macros (suggested text) for him to quickly select so he doesn’t have to type out every word.
For example, Han might type, “May the” and generative AI might suggest, “force be with you.”
Variational autoencoders
Variational autoencoders (VAEs) are generative models that encode input data, simplify and optimize the data points, and store them in a hidden storage area called a latent space. When prompted, it pulls the data from the latent space and reconstructs the data to resemble its original form. VAEs often create generative AI images and text.
Imagine Yoda, a powerful Jedi master who can use the Force to transform images into scrolls of encrypted text, instantly transports them to a locked chest on the remote planet of Dagobah, and then transforms the scrolls back into the original image on demand.
Say you give Yoda a picture of Chewbacca. Yoda can turn it into a scroll and keep it secure in his chest on Dagobah. A few days later, you ask Yoda for the picture. He once again channels the Force to access the scroll and return it to its original form.
Flow-based models
Flow-based models take complex data distributions and transform them into simple distributions. This type of model is typically used for image generation.
Say young Anakin Skywalker has a set of building blocks and every block is a different color. If Anakin wants to arrange the blocks to create a pattern, he can move the blocks in any position, but he must ensure that he always has the same number of blocks in the pattern. A flow-based model enables Anakin to create new patterns or refine existing ones while ensuring that the Force—or number of blocks—is always in balance.
Recurrent neural networks
Recurrent neural networks (RNNs) are used to process and generate sequential data. Training an RNN on data sequences generates new sequences that resemble learned data. RNNs predict what comes next in a sequence based on what’s occurred in previous sequences. RNNs are the generative AI model for Siri and Google Voice search.
Imagine Princess Leia and Wicket the Ewok are playing catch with a ball in the forest of Endor. Each time Leia throws the ball, Wicket catches it effortlessly. Wicket catches the ball consistently because he’s learned to anticipate the ball’s path and predict where it will land based on all the previous throws (sequences).
See what’s new in generative AI
Check out the highlights from The Next Big Zendesk AI Drop. Our global event details new generative AI capabilities and how they impact customer experience, employee experience, and data security.
What are some popular generative AI applications?
As we continue to learn more and understand the benefits of advanced AI for customer service, new generative AI applications are surfacing. Like the Skywalker lineage, these popular generative AI apps are the bluebloods of artificial intelligence software.
Benefits of generative AI
Generative AI offers numerous advantages, especially for customer service teams. Here are a few of the most common benefits.
Enhanced customer experience
With generative AI, your customer support teams can deliver an enhanced customer experience. Manage high volumes of requests during peak times with instant, automated answers to customer inquiries via generative replies, messaging tools, and chatbot software.
Generative AI allows for more natural, personalized conversations with accurate information. This results in a better customer experience, higher customer satisfaction (CSAT) scores, and customer loyalty. Generative AI also provides multilingual support, recognizing and adapting to languages for 24/7 global customer service.
Improved agent productivity and efficiency
Streamline workflows and make agents’ jobs easier with generative AI tools. Generative AI can handle simple tasks so agents can focus on more complex issues. Here are a few ways to leverage generative AI to boost agent productivity and efficiency:
- Ticket summaries: Generate a quick summary of ticket content so agents can understand the issue and respond faster.
- Advanced bots: Deflect tickets with bots that provide data-driven suggestions for instant, conversational support.
- Content creation: Automate and streamline the process of creating content so content owners don’t have to.
Zendesk, for example, offers generative AI in the unified, omnichannel Agent Workspace. In collaboration with OpenAI, Zendesk harnesses the power of generative AI to boost agent productivity by helping support teams create knowledge base content at scale. Generative AI can also summarize long tickets for agents and transform a brief reply to a customer’s request into a fully fleshed-out response in seconds.
Reduced support costs
AI in the workplace lets your customer support team do more with less. Generative AI helps save time and costs by deflecting tickets, streamlining workflows, and automating repetitive tasks. This means ticket queues are manageable and agents are free to focus on more complex issues, all while helping the same amount of, or more, customers.
Generative AI can also help management teams gather more meaningful insights into what types of customer issues and questions may need automation. GenAI can provide quick answers about which automation gaps exist and which would be the most beneficial to agents and business operations.
For example, it can flag if a high percentage of customers are reaching out about resetting their password or tracking their orders, so support teams can deflect these types of queries to a bot. Admins can then build these automations sooner rather than later, saving businesses time and money.
Challenges with generative AI
Generative AI can offer many benefits and help businesses navigate challenging times. But with all new technology, there may be some unexpected twists and turns. Here are a few things to consider when implementing generative AI.
Biased, outdated, or unreliable information
Generative AI systems create content based on data it’s been trained on, which could include biased, outdated, or unreliable data. It’s important to vet and validate data sources to confirm your generative AI application is pulling reliable information. Create processes and guidelines that allow you to track and remove biased data from your datasets, and monitor and review content outputs regularly to ensure information is factual and unbiased.
For example, Zendesk only makes AI available to customers after it passes rigorous quality checks. Each AI prediction or suggestion must exceed a confidence scoring threshold before being used to build automated processes.
Generative AI hallucinations
Generative AI applications are trained to provide the most reliable outputs to user commands. However, generative AI tools can sometimes produce blatantly wrong information or inaccurate results called “hallucinations.”
A hallucination is when the generative AI application provides false or irrelevant information unrelated to the dataset from which it was trained. Simply put, that means the AI model generated new content based on facts but added its own creative interpretation, resulting in distorted information. These instances do not occur often but could deliver misinformation or insensitive content.
Resource use
The reason LLM-powered bots are so impressively human-like is because the datasets that feed large language models are (as the name suggests) massive. This means the upkeep of a generative AI solution is resource intensive and poses engineering challenges.
Brands looking to implement a gen AI bot for their support might choose to host their own LLM, but the running costs for this can rack up very quickly. As well as the expense, many cloud providers won’t be able to offer the storage space these models need to run smoothly. This can cause problems with latency, meaning it takes the model a long time to process information, and lead to delayed response times.
To avoid these issues, companies might choose to rely on an open-source model, like OpenAI’s GPT-4. This option might seem like the easiest solution, but it comes with its own challenges. Ever tried logging into ChatGPT to find it was at capacity or down for the day? It’s a pretty frustrating experience. As a result, using an open-source third-party API is a risky move in customer service, where reliability is key. On top of this, while choosing an open-source LLM might seem like the most cost-effective option, the cost of single API requests can quickly add up.
The key is to choose AI with the right training data for your use case. For example, Zendesk AI is trained on the world’s largest CX dataset and is designed specifically for customer service.
How to use generative AI for customer service
Using AI for customer service makes it easy for your support team to create an exceptional customer experience with more human-like interactions. Here are a few ways to use generative AI for customer service.
Scale self-service
The opportunities to elevate your self-service resources are practically endless with generative AI. Here are just a couple of ways to use generative AI to scale self-service:
Streamline and accelerate knowledge base content by automating help center article creation.
Inspire creativity for help center content teams with suggestions and recommendations.
Make customer interactions with bots more natural and conversational by using your knowledge base to craft their replies.
With Zendesk AI, for instance, you can adapt the tone of your help center articles to make them more friendly or formal. This ensures that the content resonates with your audience and maintains a cohesive tone across your knowledge base. You can also deploy bots to offer self-service options in areas where customers commonly ask for help.
Optimize bot performance
Generative replies use information from an existing knowledge base, so you don’t need to develop custom answers. This greatly accelerates and optimizes bot-building time, and it enhances the customer experience by improving the accuracy of responses. Furthermore, adding an LLM layer to automated chat conversations enables your bot to greet customers in a friendly way and send natural-sounding replies.
Additionally, pre-trained bots use intent suggestions. This feature highlights the common questions customers are asking so admins can build answers for those intents, improving the bot’s overall performance. It also results in significant time savings and helps teams scale their bots with ease. You can even create a persona for your bots, giving them a consistent voice that reflects your brand personality.
Supercharge human agent abilities
A great way to see value with generative AI is using this technology to structure, summarize, and auto-populate tickets. Not only does this help your support team resolve customer requests faster, but it means your human agents can focus on the rewarding tasks that require their empathy and strategic thinking. LLMs can also predict categories and even analyze message sentiment. This allows agents to send tailored responses, depending on whether a customer is satisfied, seething, or somewhere in between.
Generative AI can also proactively suggest replies to customer requests that agents can then edit or tailor.
Ease agent onboarding and training
The same features that enhance the agent experience can also accelerate onboarding and training for new hires. Generated ticket summaries provide new team members with the most relevant information in the conversation, lessening their learning time.
New agents can get help with response phrasing, too. Say a new agent still needs to learn the company’s return policy and wants help replying to a customer with the appropriate details. The agent can type a few words, and generative AI can predict the rest of the sentence, filling in the blanks with the right information. Agents can also highlight their responses and adjust the tone of the entire message.
With these generative AI tools, businesses reduce training time and get support agents up to speed more quickly.
Frequently asked questions
The future of generative AI
With all the buzz around generative AI, it’s easy to buy into the excitement. However, it’s critical to have a game plan so you can maximize the benefits of generative AI now and in the future.
Our guide to advanced AI for customer service can help you learn how to harness the power of AI. Implementing generative AI now can put you in the driver’s seat to take flight on an exciting journey. We’ll be the Chewbacca to your Han Solo. Join us on the Millennium Falcon, and let’s soar into hyperspace.
