Est. reading time: 8 minutes
Author: Steph Locke
Chatbots are computer programmes that provide a (primarily) text-based interface to help people access information or perform tasks. Chatbot conversations can help you improve customer service whilst reducing demand on staff. They can process text using natural language processing and build a conversational workflow to support a intuitive interaction with people.
Chatbots are computer programmes that provide a text-based interface to help people access information or perform tasks. Chatbot conversations can help you improve customer service whilst reducing demand on staff.
Bots are the new apps. People-to-people conversations, people-to-digital assistants, people-to-bots, and even digital assistants-to-bots. That’s the world you’re going to get to see in the years to come. Satya Nadella, CEO of Microsoft
Chatbots are great for providing a natural language interface for accessing information. You can use existing knowledgebases as the basis for surfacing answers to common queries. This cuts down time spent getting information for customers, and reduces load on staff who would normally have to respond to these basic information requests.
Aberdeen City Council1 recently released a chatbot named ‘AB-1’ to answer internal and customer queries.
AB-1 allows us to divert really important resources onto frontline services that require that face to face contact. We’ve managed to release approximately six people’s worth of work and we’ve recovered our costs, which is absolutely fantastic in the first year of delivery. Andy MacDonald, Director of Customer Services
You can develop your business processes into conversational workflows to help people perform tasks. This can be wide ranging; from looking up records to do with their accounts, through to engaging in new services. There are many processes that can be turned into an effective conversational workflow. This typically helps people perform activities more inclusively and conveniently or helps reduce grunt work for employees.
Brisa2, an automotive company, developed a bot to help find company data and perform tasks like password resets, helping to free up the IT team for other tasks.
This […] saved us 400 working hours İsa Kedikoğlu, Software Development Specialist
A chatbot doesn’t have to do just one thing. You can build up a chatbot that routes people to the right business process or information component seamlessly when they say things like “book an engineer” and “what are your SLAs”.
There are some critical pieces of functionality that chatbots have or can use that are key to what you can achieve with chatbots.
Language understanding (LU)
One reason bots are really useful is their Natural Language Processing (NLP) – a subset of Artificial Intelligence (AI). NLP can help translate text we type into a set of meaningful instructions for our processes or knowledge bases. It can account for typos, different ways of typing and more.
There are LU services that can be used to build your model for making relevant translations from plain text to instructions.
Language generation (LG)
Another big area of NLP that’s key to bots is how the bot responds back to people after prompts. This enables you to go from a single fixed response to something that can feel more “human-like” for comfort.
You can use predefined templates in simple processes and answers from knowledgebases. More complex bots can surface information dynamically. These bots construct sentences that sound natural using complex LG based on NLP techniques.
Speech to text and vice versa
We can make use of bots in hands-free situations and more inclusive by supporting speech interactions. This area of AI can take audio and turn it into text, perform translations of different spoken languages into a single language text string, and turn text into human-like speech.
A bot can’t solve all the problems a person might encounter. Keeping people stuck in the bot without a means to talk to a human is a recipe for frustration. As well as providing contact information, you can also perform a hand-off to a human who is using a live chat system.
A bot doesn’t just have to deal in text. You can integrate videos, images, information cards and more into your bot. This enables you to provide a rich experience that can match well with the needs of the people using your bot. This opens your bot up to support ecommerce and browsing experiences.
Bots can be used in many different situations and places. Most frameworks people use for building chatbots come with a range of existing methods that can be integrated into different platforms or channels.
- Internal bots If you make a bot to help your staff access information or perform tasks, then your productivity and communications platform is the ideal place to integrate your bot. You can integrate chatbots with tools like Microsoft Teams and Slack to help staff use your bot seamlessly in their day-to-day.
- B2B bots If you’re using your bot to support other businesses, then your website or application is a key place to integrate it. You can make your bot available for them to implement. Your bot can also be integrated into your email capabilities to provide an automated process or information experience.
- B2C bots If your bot is to help consumers, then you can integrate the bot wherever your consumer is. You can let customers use Alexa, they could tweet your bot, or they can talk to it on social media. Ultimately, the only restriction on where you can integrate your bot is where there is value for you to do so.
You may have heard of user experience (UX) and user interface design but what about conversational design? The way people interact with a chatbot is very different to the way people interact with webpages.
You can’t simply turn a 400-field process into a chatbot process asking one question at a time. In conversation design, you work out what an effective conversation flow would be that works for the person using the bot. This can involve testing with users, producing mock-ups, and analysing the bot analytics.
Conversational design also helps you make sure your brand voice and ideals are reflected into the experience, preventing the bot from being a jarring experience.
Planning a bot project
The first step in planning a bot project is to decide where a chatbot can most help and what goal is. This helps you determine your success measures. If you want your bot to perform several functions, choose the most important to focus on and extend later.
You can then start designing the workflow and designing the conversation you want to happen. This is typically a very business-oriented task involving flowcharts. Next you need to identify the appropriate technologies, compliance needs, and what skills you’ll need to deliver your chatbot. Are you a Microsoft shop? Do you use open source only? Do you need to use technologies and store data only in the UK? Do you need to use a service provider to build your first bot and train your team? Once you know what you need to build and roughly how, you’ll usually kick off the project at a prototype level. This gives you something you can trial and get early indicators of impact. It will also help you uncover any areas where the knowledgebase might have gaps, or edge cases to the processes.
You can then start iterating and scaling your bot across more channels and capabilities taking a demand-driven approach from the data generated by the bot. Monitoring your bot is key. Keep track of the types of questions and answers going in and out, and check whether it is helping people achieve the desired goal. Validating that the bot is working for people is important, and part of continuous improvement.
The ROI of a chatbot manifests typically as cost savings and revenue gains. Success measures also depend on the application of your bot. If you deployed a bot to assist with customer service, cost savings come from reduced load on staff should be measurable. Faster and more convenient handling of problems can improve customer satisfaction and reduce churn, increasing profitability.
LEGO reported a 71% reduction in cost per conversion (compared to email) when using a campaign bot.3
If your bot includes processes or tasks, you can look at process completion rates, data quality, user satisfaction, and more.