Looking for a chatbot to help your business? Not sure whether you should build one in-house or use an outside service or software package? Designing and implementing a chatbot isn’t rocket surgery but requires some deep thought and planning. Plus, solid testing and training if you want to do it right.
First, you’ll want to figure out the who, what, when, where, and how of your chatbot.
Who is your chatbot for? Map out a flowchart with all the areas you want to serve. Those could be divisions within your entire company or just in your area. Many Marketing bots eventually morph into Customer Service and/or Sales and vice versa. Hence, it’s vital that you determine what expertise you want your chatbot to have and what the priority is. Chatbots can handle lots of requests at once, and language, tone, and navigation all matter, so it’s essential to design it with purpose.
What do you want your bot to do? Do you want it to help place orders? Point people to the right places on your site? Prioritize leads and ensure they get to the right person/place for proper follow-up? All the above? It’s important to stay focused when you’re doing this. Otherwise, your chatbot – and your users – will run amok.
You also need to know what you want the process to be if the chatbot doesn’t know the answer. You’ll further define this when developing your conversation flows, but it’s important to run through the scenarios before developing a chatbot. I’ve seen several marketing chatbots scrapped because there wasn’t a way/place to hand off the leads or customer service requests properly. Sometimes there weren’t enough people. Sometimes there wasn’t enough information/documentation. Sometimes the companies just weren’t ready to deal with any of it.
When do you want to do it? What’s your timeline? Do you want to get everything ready at once by a specific date? Do you want to build it in stages? Is there urgency? The timing of all this is much more important than one might think because you need to factor in adequate testing time. Training and testing are critical to your success – especially if your bot will use any Artificial Intelligence/Machine Learning.
Where do you want your chatbot to be located? Do you want it on every page? If you’re using it on entry pages, do you want it delayed by user activity? Do you need it in different places for different channels? (Hint: chatbot location is a great thing to test.)
How do you want the chatbot to handle your users? Do you want a rules-based menu bot where you guide the user through your process with a simple yes/no/multiple-choice questions? Or do you want something that gives the user more flexibility and seems “more human?” Does your chatbot need to be multi-lingual?
How do you want to brand it? Do you want to give your bot a name? What kind of branding and language rules does it need to have? What type of personality? If you’re doing a lot of work with Voice Search and Voice devices, this may require a deep discussion of your overall brand and your overall Voice strategy.
How are you handling the fine print? What legal considerations does the team need to know about in advance? What about privacy and security? This is critical if you use your chatbot in your cart, checkout, lead forms, and such.
How are you defining success? Planning for how you will measure success and what metrics you will need is critical.
DATA IS KEY
While you’re figuring out your game plan, you’ll also want to look at your available data and how you will use it. There are several kinds of data you need to take into consideration when preparing to build or design a chatbot.
Are you using product data to feed your system? If so, where is it coming from? Does it need to be massaged for your chatbot? Do you need any supplemental data to support the chatbot? (This is common, especially in AI chatbots.) When was the last time it went through a hygiene process?
Library/Knowledge Resource Data
What additional informational resources does your chatbot need? Is the information prioritized? Does anything need to be edited or rewritten for the bot? (Chatbots need easily chunked/parsed copy.) If you are sending a user links to specific pages, are your links ready/friendly? Do they have tracking enabled?
One of the biggest mistakes I see folks make is that they dump every single document their company has ever had into the bot and let it sort things out. This is a bad idea for a rules-based bot, but it’s a flaming dumpster fire of a plan for an AI-enabled bot. Yes, giving your chatbots a lot of information is excellent, but please be strategic about it. One-sheets from the 1990s are probably not as helpful as one might think.
How are you going to add new people to your list? What’s the flow of actions going to be? Blast and trigger emails, SMS, calls from a sales or customer service rep, etc.
How do you want to match up your existing customers and prospect file(s) with your bot? When a frequent customer visits your site, what does your bot need to recognize them? A phone number? A customer ID? Security information? Nothing at all? There are legal considerations in this area, so it’s important to get the right team members involved. I’m not a lawyer, and I don’t play one on TikTok, but I’ve seen enough companies mess this part up to know that you should carve out a chunk of time for legal, security, privacy, and accessibility reviews.
You’re going to want to build Conversation Flows for each of your user journeys. People often think creating Flows only applies to rule-based chatbots, but it’s a good exercise to do for all chatbots. Remember, you need to train AI chatbots, too. When you’re building a Conversation Flow, you’ll want to map out what happens when. If I come to your site asking how quickly I can get a rowing machine delivered to my house, I would get a very different flow than if I asked how to put together the rowing machine I just bought. Knowing what kinds of conversations you will have before you start building/utilizing your chatbot will help you properly define the navigation, organize the items underneath it, tighten the choices, and so on. It’s a beneficial exercise.
Some marketers design their Conversation Flows and follow-up campaigns separately. I prefer to do them all at once. This helps ensure that the data flows to the right place and that your users get the proper trigger emails, texts, etc.
This may sound like a ton of work, but it’s actually easy and kind of fun. Just focus on your top journeys and build out the ideal experience. Keep reminding yourself: What’s the user’s goal for this visit? How can I meet their needs most efficiently?
After you’ve built out a handful of flows (or decision trees, if you prefer), the rest should come together quickly. Please note: flows can get very complicated if you don’t ask the right questions at the right times. Design your flows and review them step-by-step, cutting and consolidating as necessary. Then, do the same thing when you’re testing. (If you’re using an outside vendor, they usually have this process down to a science.)
You’ll also want to create your intros and outros (greetings for the beginning, thank yous/goodbyes for the end) and any brand rules/legalese you may need to include.
Are you looking for a chatbot, or do you already have one? Have a tip you’d like to share or a question you’d like to ask? Tweet @amyafrica or write email@example.com.
This article is Part 2 in the Chatbot series. Read Part 1 and Part 3 here.