How and Why to Add a Chatbot For Packaging and Processing

In this episode, we chat with Jessica Lawshe, a seasoned expert in leveraging chatbot technology for enhanced customer experiences and business efficiency. From defining chatbots to exploring their practical applications in industries like manufacturing, trade shows, and beyond, Jessica breaks down how businesses can use these tools to improve operations, engage customers, and even create new opportunities.

Sean Riley: Hi, and welcome back to another episode of unPACKed with PMMI. I'm your host, Sean Riley. In this episode of unPACKed, we chat with Jessica Lachey, a seasoned expert in leveraging chatbot technology for enhanced customer experiences and business efficiency. From defining chatbots, to exploring their practical applications in industries like manufacturing, trade shows and beyond, Jessica breaks down how businesses can use these tools to improve operations, engage customers, and even create new opportunities.

Let's have a listen.

So with all the fancy introductions out of the way, welcome to the podcast, Jessica.

Jessica Lawshe: Thank you, Sean, for having me.

Sean Riley: Oh, the pleasure is all ours. So today, we're talking chatbots. And it would make sense for our listeners if we started off by defining what we are talking about. So I guess to kick things off, what is an AI chatbot?

Jessica Lawshe: Yeah, just to give a brief overview, an AI chatbot, it's a virtual assistant that's available 24/7 to help people find information, or solve problems. Think about visiting a website late at night, no one available to answer your questions, an AI chatbot kind of steps in to fill that gap. It uses artificial intelligence to understand what you're asking, and then responds in a conversational way that feels natural.

Sean Riley: Okay. I've dealt with them. Sometimes you'll get them first. Before they give you a regular person, you deal with the AI chatbot first.

Jessica Lawshe: Yeah, there's two main types, so the rule-based and AI-driven. So rule-based bots stick to a script. They respond to specific keywords or commands, like following a flow chart.

Sean Riley: Okay.

Jessica Lawshe: They're simple, but not very flexible. With the AI-driven chatbots on the other hand, they're much smarter. They can understand the meaning behind a question, even if it's phrased differently, and then get better over time, because they can learn from interactions, and that's what makes them so much more conversational and versatile.

Sean Riley: How specifically could I use a chatbot as an OEM, or a machinery builder?

Jessica Lawshe: Being able to provide instant responses to common inquiries about machinery, or even specifications, or troubleshooting. There's also spare parts lookup that you could train it on, assisting customers in identifying and ordering replacement parts, that would save a lot of time. Being able to generate price quotes based on what the customer is asking, would reduce the time to close those deals. Pre-qualification of leads, so they could ask targeted questions to assess the needs and budget of prospective buyers, and then passing only qualified leads to the sales team.

And then you have operator training, so deliver interactive tutorials, or answer questions from operators learning to use the machinery. But my personal experience has been creating chatbots for the PACK EXPO trade shows. They're great for enhancing the attendee experience, because we don't typically put on-site information on our website, like check hours and locations, or where to go to print your badge. Adding this information into the chatbot allows attendees to get those instant answers. I also train a chatbot to recommend a relevant category of exhibitors, because some people attending the trade show don't know what kind of machine they need, or they have trouble navigating the complete list of exhibitors. And because I've trained the chatbot on packaging and processing industry terminology, and then do extensive fine-tuning, it's able to analyze the user's question, first of all, and then understand the type of machine, and then recommend the category, along with the link.

I've also found that exhibitors were using the chatbot, so not just attendees. What's interesting about a chatbot, they tend to be used more customer-facing. There is another way that they can be used, they can also be used as an internal tool. So say you have a sales team, and they're on the phone, but they can't locate the answer to the question, you can have a chatbot trained on all of your data, and then allow the salespeople to instantly get answers while they're on the phone. So it can be used either way.

Sean Riley: Wow, that's really cool. Okay, so with all these different types and examples that you've given, how do I decide which platform is going to work best for me?

Jessica Lawshe: Yeah, it all starts with defining your goals. What do you want the chatbot to do? Is it there to answer basic questions, provide customer support, or maybe drive sales? Once you've decided that, the next step is looking for a platform that fits your needs by signing up for a trial. You'll also want to think about integration. Does the platform work with the tools you're already using, like your CRM, or your e-commerce system? You can also set up API calls to external data. This is a way to reference real-time data based on certain questions. For example, if an attendee asks if a certain company is exhibiting, you can have an API call reference the exhibitor list that lives in a third-party platform. Lastly, I have to say is, don't forget about analytics. Having built-in analytics is important, because it lets you track how the chatbot is performing, and you get insight into the user behavior.

Sean Riley: So now that I've chosen a platform, how do I go about building the chatbot?

Jessica Lawshe: Figuring out what the chatbot needs to know. So this could mean having it scrape your websites, uploading documents such as reports and presentations, even infographics. You can even connect it to resources like Google Drive. The key here is to make sure that the data that you're feeding it is accurate, well-structured, and easy to process.

Sean Riley: I know data is super key these days, I hear data talked about all the time. So how should the data be structured?

Jessica Lawshe: Yeah, when you have a site scrape your website, the data is scattered all over the place. So when it comes to structuring, keep it simple and clear. For example, make sure your website has logical headings so that the chatbot can find the information easily. And when uploading documents, there is a big difference between a Word document and an Excel document. The chatbot will analyze a Word document to understand the overall context, making it harder to pinpoint specific data, whereas with an Excel file, with structured data in rows and columns, it allows the chatbot to search based on specific fields. So it really helps when the chatbot references those specific links and being accurate. But if you want your chatbot to understand industry terminology, which we've had to do with our industry, and packaging and processing, you could create a spreadsheet with terms and their definitions. Or if your website descriptions are short, you can expand on them in a spreadsheet. So the more organized your data is, essentially the better the chatbot will perform.

Sean Riley: How do I make sure that the chatbot answers the way that I want the chatbot to answer?

Jessica Lawshe: Now, that is where it comes down to the model instructions. So basically, the model instructions are the chatbot's rule book. You want to instruct the chatbot how to behave, what to do if certain topics are asked, or even what not to do. One thing I like to include is guidance for when the chatbot isn't confident about an answer, I tell it to give a structured generic response instead of guessing, which really helps avoid those "hallucinations," where it tries to make something up.

Sean Riley: What if what I'm getting is right, like the chatbot is giving the correct answers, but I don't like the flow of it, or I want it to be worded differently. Is that something I can adjust?

Jessica Lawshe: Yeah. In most chatbot platforms, that's the process of what is called fine-tuning, and it's actually one of the most important parts of the process. So during testing, even if the chatbot answers are technically correct, they might come across as too formal, too casual, or just not aligned with the tone that you want for your brand. For instance, if someone were to ask something simple, "What time does your store close?" it might answer with, "The business operating hours conclude at 9:00 P.M." Now, while that's correct—

Sean Riley: Right.

Jessica Lawshe: It might sound overly formal for a retail brand. So fine-tuning lets you revise that. When you make these adjustments, the chatbot uses your updated phrasing as a reference for similar questions in the future. And it's worth noting that fine-tuning is a continuous effort; you will want to occasionally monitor the chat log and tweak responses.

Sean Riley: From a data perspective, are there ethical considerations that businesses have to keep in mind when deploying a chatbot?

Jessica Lawshe: One big thing is definitely monitoring the chatbot's performance. It's one thing to just feed it all this data and push it live, but you'll definitely want to monitor it. And if it gives an irrelevant or inappropriate answer, that's a sign that you'll need to adjust the instructions. And it's also important to be transparent about the data collection.

Sean Riley: The other thing that pops up when talking about anything that has to do with the internet or anything online—I think of cybersecurity. So is there anything that we can do to ensure the chatbot is secure? Is that even something that happens, where chatbots get hacked, or people have malicious intent?

Jessica Lawshe: Absolutely. There's been some crazy stories I've actually read online about that. So security is absolutely a big concern, and one of the best ways to protect your chatbot is, first, limiting its access to only the data that it needs. That way, even if something does go wrong, the exposure is minimal. But you'll also want to include security measures in your model instruction. These types of safeguards keep the chatbot secure and reliable.

Sean Riley: So I have time for one more question, and I'm thinking about how there usually is a person on the other end of these little conversations, and with the chatbot, there isn't anymore. Are these taking people's jobs? Are these replacing customer service people? How do we address that situation?

Jessica Lawshe: Yeah, it's definitely a common concern that chatbots might replace jobs, especially in customer service or sales. But the truth is, chatbots and human support work best when they complement each other. Chatbots are great for handling the repetitive, low-complexity tasks, but by taking care of these routine tasks, chatbots actually free up employees to focus on the more complex, creative, or high-value work that requires that human touch. So the real magic happens with this smooth handover. If a chatbot comes across a question it can't confidently answer, or it notices user frustration, you can set it up so that it passes that conversation over to a human agent. So it's not replacing the person, but it's taking care of basically those frequently asked questions, and then the person on the other end, if you choose to have one, is there if it's needed. And what's exciting is that chatbots are also creating new opportunities. So companies now need people like data analysts, and AI specialists to manage, and optimize, and grow these chatbot systems. So instead of replacing jobs, chatbots are changing the nature of work, and creating space for new specialized roles.

Sean Riley: I love that. Okay. I can't thank you enough for taking time out of your day to explain and walk this through, because I didn't understand any of the intricacies of chatbots, and now I think I could at least explain to other people what chatbots are. So thank you, Jessica, for taking time out of your day to come on the podcast with us.

Jessica Lawshe: Absolutely. Thank you again for having me.

‍