Chatbots and intelligent messaging solutions can help you deliver smarter customer service with a better customer experience at a very attractive cost per contact. To achieve the kind of results that get you promoted, it's important to consider a customer service solution in its entirety. Overlook one key element of the solution stack, and it'll never achieve its natural flow.
We use the technology stack in this post to talk about goal and transaction oriented Chatbots both internally and with our customers. You can use it as a checklist to make sure you're not overlooking a critical part of a Chatbot solution for customer service.
The Chatbot Technology Stack
In all the Chatbot excitement we often see enterprises overlook critical parts of a customer service solution like easy transfer to an agent or an easy admin interface. Use this 9-layer framework as a checklist to help you think about the questions you're not asking.
Customer Input/Output
The way the customer interacts, with the system using voice, chat, messaging, app or web. As an example, I interact with my friends using the chat app on my mobile phone. I can type as input or use TypeTalk as a voice to text input. This also includes the user interface and mode the bot uses to respond to the customers.
Channels and Devices
The devices and channels used. This includes messaging platform (eg, Facebook Messenger), online chat, SMS, email, website and hardware (Alexa, Google Home). If the bot can use multiple channels, then omni-channel interactions are managed here.
Comprehension and Conversation
This converts the input from the customer into language the computer/bot understands. And it converts bot language into human language (voice, text etc). The conversation is managed here with the ability to ask a question, queue several messages at once, and track when an interaction has ended. One goal is to manage loose probabilistic conversation (the way you and I talk) to structured services the bot understands. Today, the capabilities of this layer are improving rapidly using machine learning and other AI technologies.
Once the Chatbot understands what the customer says, it can choose or generate a response, based on the current input and the context of the conversation.
Workflow
Workflow is where the task, question, or Job to be Done gets performed and managed. Lots of power resides in here and AI technologies are driving new levels of engagement.
Two types of interactions are managed here:
- Push (like an outbound call) – here the bot will push out help or information. For example, the Chatbot can provide a reminder in a conversation to make sure that someone has not forgotten about items in their online shopping basket (if they haven’t checked out). Or it could send out a survey at the end of an interaction.
- Pull (like an inbound call)- this is where the customer initiates the workflow. For example, the customer can ask "What’s my balance?"
There are three types of responses that Workflow can generate:
- Static responses - the bot selects from a predetermined list of responses. The static response could also be a template in which a dynamic response is inserted. “Your flight departs in” XX “hours” is an example.
- Dynamic responses- the bot use resources, such as a knowledge base, to get a list of potential responses, and then score them to choose the right response. This is particularly appropriate if you chatbot acts mainly like an FAQ system.
- Generated response – the bot generates a response on the fly without using canned responses. This technology will take some time to mature, and requires a very large data set and deep learning algorithms.
Data and Services
The bot accesses knowledge and services to do tasks in this layer. APIs connect to CRM systems, services, Knowledge Management, payment platforms, authentication, ecommerce platforms or whatever is needed to complete a task.
Connect to an Agent
The bot also needs to provide seamless access to a customer service agent if and when it's needed. For example, I start an interaction to order new running shoes with a bot. I get all the info I need but have a question that the bot cannot answer. The bot asks if I want to talk to an agent and sends the interaction history and context to an agent. Designed properly this is a friction-free, branded experience.
Also, remember that routing, queuing and agent desktop needs to be included in the solution when an agent is involved.
Don't let it be a surprise: Bots need to transition smoothly to agents.
Administration
One of the most challenging parts about designing a Chatbot is to make the conversation flow as naturally and efficiently as possible. Administration allows you to easily build and manage conversations and workflows. So far, most Chatbot platforms are not designed to be administered by the contact center staff.
Analytics
Three types of analytics can be used for the Chatbot (not considering the voice interface).
- Descriptive- Answers the questions: What happened? What to do? This is typically BI and historical reporting that summarize what happened.
- Prescriptive - Answers the questions: What could happen? What is likely to happen? Text mining can be applied to the text interactions to determine patterns, infer data (like sentiment), and make suggestions.
- Predictive - Answers the questions: What should we do? What should be done? What can we do to make a certain thing happen (like get customers to self-serve for a claim status check)? What is best, What is right? We can use algorithms to forecast possible outcomes and provide next-best-actions.
When analytics are applied to Chatbots, the interactions with the customer can become more useful, quicker and easier.
Compliance and Security
Technologies used by enterprises must meet security and compliance standards. I won't go into detail here, but compliance always needs to addressed.
More on Intelligent Assistance
Intelligent Assistance and Chatbots are going to change the way customer service is provided. We hope this post provides some insight on the basics of Chatbots so it's not so confusing.
Here are a few additional resources on the Chatbot tech stack you might like:
- Chatbots 101 (Forrester, Dec 2016)- you can download from us here (limited time)
- Executive Q&A: Boost Your Chatbot IQ (Forrester, Dec 2016)
- This Chatbot platform feature comparision matrix provide more tech details.
- This article is an excellent read: Designing the chatbot stack...
- Dan Miller and the Opus ResearchTeam rock Intelligent Assistance
Reach out to use if you want to talk more. You can also pilot our messaging app and Chatbot to get hands-on experience in your contact center. Learn more about our Intelligent Messaging pilot here.
