5 Questions to Answer When Planning a Chatbot for Your Company

Mika Tähtinen
June 15, 2017

Let me take a few minutes of your time by sharing a story of a bad customer experience I recently had.

A couple of days ago, I saw an interesting sponsored Facebook video ad from a company manufacturing motorized window blinds installed inside the window frames. Now that summer has started, I've been having trouble sleeping long because of the morning sun. Great marketing, this product is just what I would need and it was shown to me when I need it the most.

To find more information, I browsed the company's website and Facebook page. Everything seemed great, I was ready to order the blinds. Just one thing was missing – I couldn't find the pricing information anywhere. I contacted them using the channel that was the most natural in this case: I left a comment on the Facebook post.

This happened about a week ago and I still haven't received a response. After reading the comments others had left, I noticed that the company was requesting them to call or send an email to inquire pricing information. After this, I just gave up. Opening email and typing up a structured message were more of an effort I had the time or energy for. Calling them was out of the question since it was both late at night and in general, I'd rather get the details in writing so I can think it over.

Great marketing led to poor sales. The whole idea of building chatbots to improve customer experience is based on cases like mine above. Don't wait for the customers to come to you, you need to be where your customers are.

Currently, your customers are on social media and spending more and more time using messaging apps. Now is the time to build a chatbot to make sure you bring value to your customers whenever they need it.

Building a chatbot, as any other software project, requires some planning to ensure a good basis for the implementation. Here are the questions you should answer yourself when starting a bot project.

1. Why do you want a chatbot?

You've read it here and everywhere else: you should have a chatbot. But what do you want to achieve? What do you want to give to your customers? What do you want them to give you? There are no general answers to this question. You need to answer them based on the line of business you're in.

Maybe your product line is complicated and you need a chatbot to give your customers answers to common questions, like a smart FAQ. Or you are selling something online and want to give your customers a possibility to order, track, and even return products using a bot. You might have a busy customer service that spends time with asking the customer's details for each call or chat session when a bot could ask this information beforehand.

So, make a list of objectives for both your company and the customer that is chatting with the bot. What should the bot be able to do? How will the bot fit into your business model and help you grow?

2. How does your chatbot sound like?

You don't need to be a tech expert, but when you start planning a chatbot, spend some time trying out some existing chatbots, reading some blog posts, and in general get to know what chatbots can and cannot do.

When you are testing bots, pay attention to how they sound and feel like. Does the bot have a fun personality or is it strictly business? Is it proactive or only reacting? Is it pushy, offering you more information, selling you things? Think about your own company. What would your chatbot sound and act like to reflect your brand?

3. Who will be using your chatbot?

If you don't yet know who your customers are, do some research. You don't want to go building a chatbot, that only knows how to do things your customers aren't interested in using. Figure out your target group for using the bot. Do they fit the personality of the bot you are planning to implement?

Remember that you may also have internal user groups for the bot. Maybe your sales personnel can also utilize the information that the bot provides. Will you need your customer service to be able to step into a chat with the bot, if the customer requests? If so, you need to make sure they are trained for these situations.

The bot's artificial intelligence (AI) is something that is built around a natural language understanding system, such as Microsoft's LUIS. As the bot is gathering discussions with users, it needs to be taught what the users are trying to say or do. This work is mostly done at the start of a bot project but may also require some effort later. Who in your personnel has the skills and time to do this, or should you outsource it?

When you have a bot, you will need to start marketing it to your customers to get them to use it. To accomplish this, you may need to implement some extra features, plainly for marketing purposes to get people to try it out. So, create a marketing plan, at least on a high level. Where will you market it, to whom will you advertise it, and what will be your approach?

4. What does your chatbot know how to do?

Now you should have enough information to start planning the actual capabilities of the bot. Develop use cases for the users and try estimating how complex the interactions will be. Probably the most common and simple use case for a bot is to reply something when a user says "Hello" or "Bye", but it is still one use case you need to implement.

Things get more complicated when you need multiple pieces of information from the user, like filling out an order form with all the required details or having a conversation which has multiple paths to progress. To make things easier, draw a discussion tree showing all the paths a complex discussion may lead to.

After your happy with your plans, start thinking about cases when things go bad. For example when the user mistypes their address and wants to correct it, but the bot is already asking for their phone number. Or when in the middle of filling up an order the user has a question about delivery times.

What should the bot do, when it doesn't know what to do? Instead of just saying "Sorry, I don't understand.", you could show the user a list of things they can do, or offer an option to start over. It's OK when the bot doesn't know how to do something, but it would be nice if it at least gave some hints on what it can do.

5. Where are your data sources?

On the technical level, chatbots are a network of integrations between multiple systems. They need to connect to messaging platforms, such as Skype and Facebook Messenger, to language understanding systems, and most likely to your data sources, such as ERP, CRM or content management systems.

Connecting to messaging platforms is usually the easiest part when using ready-made frameworks, like Microsoft's Bot Framework, which provides the connection to many popular apps, but there are some differences between apps on data that they provide of the users. You should list the most important messaging channels for your customers and start with those.

Define which systems will provide and store data for the bot, which data will be necessary and what are the interfaces used to fetch and save it. The more documentation you can gather about the interfaces and data formats, the faster it will be to implement these integrations.

Also, consider whether you need to authenticate your users and if your answer is yes, where from will you get the data for this? The messaging platforms will provide you with some detail about, for example, the user that is logged into Skype and chatting with your bot. But if you want them to be able to access data on their account in your systems, you need to find a way of making sure it is the same user.

And remember, whenever building integrations, you need to have enough time to test that everything works seamlessly together.

Commercial chatbots are still a new thing and taking their shape as we speak. Start building a bot that suits your company and customers, and find new ways to change the business.

http://www.youredi.com/blog/5-questi...r-your-company