In Part 1, we met Sasha, head of customer service for a midsize software company. Sasha and her team were struggling to meet the demands of their customers, spending too much time on routine questions and not enough on more complex ones. I explored a GenAI chatbot as a potential solution for her problem.
In Part 2, I’m going to dive deeper into using a GenAI chatbot as part of a customer service strategy. I’ll look at biased responses, testing, costs, and timeframes. I help you ask whether the effort is worth it and share the first steps in a GenAI project.
Let’s get started.
Getting Unbiased Responses
Bias exists in every piece of writing. As a result, every chatbot has biases. It’s up to us humans to work to minimize those biases.
The first step is to recognize our own biases and how they affect AI. You can start with something like the Implicit Association Test to learn about yourself and Lenovo’s new AI PCs for Business Online course, which covers biases in modules 2 and 5.
Once you have a better idea of what to look for, thoroughly review all the information you’ll use to train your bot. With your audience in mind, look at for a variety of viewpoints and experiences. For example, does the information presume all customers have a strong, steady internet connection and a blazingly fast computer? Does it rely on having a mouse and keyboard? Is your shipping information focused on your home country while your customers are global?
Also watch for assumptions that your customers have the specialized knowledge that you have. Once we learn something, we can forget that we had to learn it or how hard it was to learn (a.k.a., the curse of knowledge). With my authentication app snafu, I didn’t know what the tech team did: that to launch the app with my old credentials on a new phone, I’d need my old device handy.
For Best Results, Test, Test, and Test Some More
You’ll want to test your chatbot extensively before launch. For the best results, have testers who know what a good response should look like and those who don’t. Those who know what to expect can evaluate the bot’s responses for accuracy and safety. Those who don’t know what to expect can identify responses that don’t make sense. They should try to follow through with the bot’s suggestions. They should try to break it.
Better to break it before launch than after, with real customers.
Don’t let testing stop with launch, however. Regularly test your bot and analyze performance results. Is it still providing accurate information? Are customers happy with responses? If something’s wrong, you need to address it immediately. Every customer interaction is at risk until the bot is fixed.
How Much Will It Cost?
Speaking of purchasing features, how much will this project cost? Like all technology, it’s going to depend on how much customization you want. McKinsey & Co. breaks down the total cost of ownership for three groups of buyers: the takers (buyers of out-of-the-box products), the shapers (buyers of products with some customizations), and the makers (buyers of highly customized products). At the low end, you might spend $2.5 million to get your bot up and running and $0.6 million annually to keep it running. Exact numbers will depend on the features you choose and the number of users daily (see Table 1).
Table 1. Estimated Cost to Build and Run a GenAI Chatbot
Source: McKinsey & Co., 2024, https://ceros.mckinsey.com/genai-cost-interactive-desktop/p/1
How Long Will It Take?
Once you decide on a GenAI chatbot for your site, how long it takes to implement will again depend on the complexities. The more features your chatbot has, the older your site is, or the more information you need to gather and train the bot on, the longer it will take.
This isn’t a project to rush, either. Whether it takes a few weeks or up to a year, once that bot is embedded in your system, you’re fully invested. Take the time to follow best practices, and you’ll be rewarded with happier customers and a strong ROI.
Is It Worth It?
A GenAI chatbot is a significant investment, even at the Taker level. Is it worth it? The answer will be different for every company. Explore the risks for your company before diving in. Here are some of the most common risks to research:
· Legal liability. Laws around AI are still developing. Know your risks now and stay on top of changes.
· Dirty data. You’ll need to train the AI on your information. If it’s not in good shape, you won’t get good responses. Review and edit it.
· AI hallucinations. Your bot needs to give accurate information. Test its responses and test them again. Try to break it.
· Who your customers are. Some customers are happy to use a bot to find an answer and solve a problem. Others want a human, no matter what. If most of your customers are in the latter group, your bot might be a waste of time.
· What customers’ problems are. If your customer service agents primarily handle complicated issues, rarely handling easily resolved problems, a bot won’t be helpful to you.
· Need for 24/7 availability. If your customers don’t need customer service around the clock, why pay for offering it?
· Legacy systems. If your existing systems won’t play nice with your desired bot, is it worth the cost of making them work together?
Create a Steering Committee
You’ve decided that a GenAI chatbot offers potential for your company. It’s time to put together a steering committee and put the idea through the grinder.
Often it’s customer service suggesting a GenAI chatbot. If you’re not in customer service, though, be sure to involve them from the start. They have the necessary insights into customers’ issues and how well the team is able to address those issues. They can also help define success metrics because they will feel the most direct impact from adding the bot.
You will also need input from several department leaders in your company. Table 2 lists who to invite and a sampling of responsibilities in this project.
Table 2. Steering Committee for a GenAI Customer Service Chatbot
Questions the Committee Should Explore
With the steering committee in place, you can now explore whether a GenAI chatbot is appropriate for your company. Together you should ask questions like:
· Can the chatbot provide a satisfactory experience compared to human agents?
· How might our customers react to interacting with AI instead of humans?
· How much can we customize the chatbot to align with our brand voice?
· What languages can the chatbot support?
· How do we ensure the chatbot doesn’t provide incorrect information?
· What happens if the chatbot malfunctions or goes offline?
· How much ongoing effort is required to keep the chatbot effective?
· How will this affect our current customer service team? Will it lead to job losses?
· Can the system handle sudden spikes in customer inquiries?
· How much human oversight will the bot need?
· How long will it take to see a return on our investment?
· What other metrics will tell us how well the chatbot is performing?
· What are the security risks we must address?
· Are there any potential biases in the AI that could affect customer interactions?
· How does this solution compare to what our competitors are using?
Ask questions that relate specifically to your business or industry as well. But don’t debate forever. Build a sense of whether this is worth doing and then make a decision.
Meet Customer Needs with Humans and Bots
Adding a GenAI chatbot to your website can help your customers serve themselves and free your customer service agents to deal with the complicated issues. You’ll be able to offer 24/7 customer service and high-touch customer service on a midsize company’s budget. Following best practices, working to minimize bias, and comprehensive testing can help you get there.
By doing your homework—gaining a thorough understanding of your customers’ needs and how a GenAI chatbot will affect other departments—you’ll be well on your way to harnessing the new potential in GenAI and helping your company grow.
Erin Brenner is the owner of Right Touch Editing, a boutique editorial agency that specializes in helping small and midsize businesses to be more engaging with their audiences, more persuasive in their marketing, and clearer and more precise in their communications.
Erin is also the author of The Chicago Guide for Freelance Editors: How to Take Care of Your Business, Your Clients, and Yourself from Start-Up to Sustainability, Marketing Yourself Guide (with Sarah Hulse), Copyediting’s Grammar Tune-Up Workbook, and 1001 Words for Success: Synonyms, Antonyms & Homonyms. She is an Advanced Professional Member of the Chartered Institute for Editing and Proofreading and a Full Member of ACES. Follow her on LinkedIn and Bluesky.