More and more companies are investing in Artificial Intelligence (AI) to provide and develop their customer service, and for good reasons. As the lives of consumers grow busier and human society more fast-paced, most people would prefer to make an enquiry quickly and easily online. For example, in between work and other commitments. This is a welcomed alternative to spending minutes (or even hours) on hold over the phone or actually visiting a company employee in-house.
To back this up with data, approximately 89% of companies compete primarily on the basis of customer experience (this is up from only 36% in 2010). But while 80% of companies believe that they deliver positive experiences, only 8% of customers agree that they do. Chatbots can be a powerful means of beating competition when it comes to winning and keeping customers based on customer experience.
Although this isn’t to say they’re a necessity to all companies, customer expectations are changing in ways that can be helped with chatbots in most businesses. Here are some examples of questions to ask and get answers for before you invest in chatbots for your business.
Are you focused on:
- Automating tasks to reduce cost?
- Improving customer experience end to end?
- Generating more sales?
- Taking pressure off your phone line customer service team?
- Providing candidate interaction early on in the recruitment process?
- Taking pressure off your HR and recruitment team?
- Finding out more about your consumers including FAQs and flaws in the current sales system?
- Providing a means of customer service at any time of day?
We’ve outlined how e-commerce companies can benefit from chatbots and how chatbots can help businesses with staff retention and recruitment before. Today, we’re taking a closer look at what it takes to build a successful bot strategy, whatever industry you’re in.
Seven steps to a successful bot strategy
1. Define your objectives
Before you invest in chatbots for your business, get a clear picture of what you need from them and whether or not this can be delivered through chatbots. Your company objectives may include:
- Generation of sales
- Improved customer service
- Drive traffic to the website
- Build a CRM channel
- Consumer research
Customer/user objectives may include:
- Novel experience
- Easier access to information (without the need for a call or face-to-face visit)
- Quick help / support (at any time)
Define both sets of objectives clearly and look for how these could be achieved with chatbots.
2. Define your domain
Reflect on what topics your chatbot will need to know about, and how deep it’s knowledge needs to go. We refer to this understanding as the ‘bot’s domain’ and it can be split into breadth (understanding of connections across your business) and depth (understanding in a particular area of your business).
Take a ‘room booking’ bot as an example. Are you dealing with booking any kind of accommodation or hotel rooms only? Are you booking accommodation anywhere in the world or just in London? This is breadth of understanding. Is your room booking bot going to try to be aware of every hotel room in London? Or just the rooms our client is selling? This is depth of understanding.
3. Define your metrics
It’s all very well investing time and money into strategies aiming to improve your business. But if you don’t identify key performance indicators or any means of measuring their success, it’s impossible to learn from your experiences. Implement a measurement program that can accurately measure the amount of time your customer is interacting with and actually being helped by your agents. This way, you can assess the speed and accuracy with which the bot identifies customer intent and goes about executing a task. It’s wise to use more than one metric to do this.
For example, handle time may be useful as an indicator but not a clear measure of effectiveness when it comes to chatbots. Average handle time is the average time it takes to handle a call or transaction from start to finish – from call initiation all the way through to any related tasks an agent must perform post-phone call, including any hold time and call talking time. Customer satisfaction may involve handoffs between bots and agents to resolve an issue with minimal effort and time from the customer. When human agents handle calls or text chats and handle time goes up, that could indicate a problem (and tax capacity at the contact center). But when a bot’s handle time goes up, that could indicate the customer needed more help and wanted to go at a more deliberate pace. So bear this in mind when monitoring the success of your strategy.
4. Blend your AI
Artificial intelligence and whether or not robots should be trusted and replace humans entirely has been a topic of conversation in many aspects of society. Films have been made about it, such is our fascination with the concept (iRobot, Transcendence and Big Hero 6, to name a few!). And more recently, as the applications of AI are coming into fruition, the flaws of both human and robot reasoning are becoming apparent. For example, through the use of AI to build driverless cars and design robot doctors. While a robot won’t tire or be affected by human conditions such as stress, fatigue and distraction, it can’t interpret grey areas in a way that a human can, it can only think in black and white.
So, where the use of a robot removes the risk of ‘human error’, there are some decisions and judgement calls that require a human brain. In less drastic scenarios, such as customer interaction, the best strategies allow bots to deal with basic queries and begin the customer interaction. Human services will be on hand to step in should it become apparent that the query is too complex or has not been programmed for yet.
The best strategy will adopt the approach of blended AI. This is when bot technology helps companies execute tasks faster by automating time-consuming or mundane tasks, but is achieved with human supervision and training. Companies shouldn’t invest in automation capabilities aiming to replace humans completely, it’s all about bot and human collaboration.
It goes without saying that there are security measures you need to be aware of when building chatbots. Your bots have the potential to store, access and process information on your behalf, as well as handle sensitive consumer data, so security is paramount. Think regulated industries and activities, consumer laws, data protection and infringement of third party rights.
In September 2018, California became the first state to pass a cybersecurity bill that oversees connected devices. This law requires that bots are clearly identified so everyone knows if they’re talking to a human or virtual agent. Make sure your company is up to speed on the legal side of things and that this is built into your bot strategy.
6. Think ahead
As we all know, technology and customer expectations are evolving constantly. And while this is an exciting and interesting world to live in, any kind of upgrade in technology can be short-lived. What’s state of the art at any one time, won’t be for long.
Bear this in mind when building your bot strategy and think in terms of platforms, channels and devices. Opt for underlying technology that facilitates easy integration with new channels as technologies develop. This way, your initial strategy can be the foundation of a longer-term strategy. You can build up and out from your initial strategy, rather than having to switch channels or programmes and start again. Preparing for the future could save you time and money. The best bot strategy will facilitate agile integration into new platforms, channels and technologies as they arise.
Once you’ve scoped your bot strategy, you’ll begin the more technical steps of mapping out the conversations your bot will be having with customers.
7. Building conversational flow
You’ll map out potential bot conversations according to three pillars of bot conversation planning: conversations trees, keyword responses and data sources.
- Conversation tree – a map of all user journeys and content outputs through the conversation. Think flow charts, mapping each step of a user chatbot interaction from welcome message to final outputs. There’s no chatbot-specific design software at the moment, but there are various mind mapping programmes available such as draw.io.
- Keyword response – the definition of how the bot will respond to keyword inputs. You’ll list keywords and phrases you need your bot to respond to, and how it should respond.
- Data sources –databases, files and repositories containing information and content that will be delivered through the bot. During the design process, outline in as much detail as possible:
- Where your data is coming from (bot-specific databases, existing owned databases, 3rd party APIs, etc.)
- The format each of these delivers in.
- The queries underlying each dynamic component
- Relationships between the different datasets
Rails are preset conversation paths that move sequentially through stages, with clear signposting and buttons for the available options.
Content 1a→ Option A
Content 1b → Option A or B
Content 1c → Option A, B or C
You’ll need to create a chatbot that communicates with your rails database. For example, you’ll want responses to ‘small talk’ keywords and phrases such as ‘hello’ (and its many variants), ‘help’, ‘I want to speak to a human’, ‘how are you?’ and any of your FAQs.
Natural language processing
Nowadays, chatbots are capable of more than predetermined, linear conversations. Natural language processing (NLP) is a technological process that allows chatbots to identify the message’s intent and determine the appropriate response. It involves applying machine learning to a phrase in order to extract meaning from ‘free text’.
There are a number of different ways in which the NLP function can be built and the best options will vary depending on steps 1-6. The necessity of NLP is highly dependent on what you want it to accomplish but generally speaking, integrating NLP means adding a more human touch. Through NLP, a bot analyses a user’s text input, determines their intent, and then delivers the relevant content or action in response. NLP platforms to be coded and configured into your bot include IBM Watson and API.ai. Most chatbots will need both rails and NLP built in.
AI is facilitating a profound transformation when brands and consumers have chatbots transacting on their behalf. Chatbots are proving to be a valuable investment in more and more business sectors and industries but how far is the use of chatbots going to extend? Only time will tell but for now, their value appears to be growing. Interested in learning more? Have a browse of our resources from more information, tips and guides.