A Guide for Enterprise to Get Started with AI Bots
Get Started with AI Bots for your Business
Despite the hype, when you want to get started with AI bots (also known as digital AI assistants) there are many things to consider. When adopting new technology, some businesses lead the charge while others wait and see how things unfold before jumping on the bandwagon. Conversational AI technology has certainly followed this pattern.
Some early adopters experienced unexpected bot failure, such as the highly advertised failure of Microsoft’s Tay chatbot in 2016. Other early adopters pursued a long and expensive path to developing and deploying their bot. Bank of America, for example, reportedly spent 2 years with an in-house team of 100 and a budget of $30 million to launch its famous Erica voice assistant.
The reporting of epic bot failures has died down, thanks to recognized deployments benefiting from advancements in speech recognition, natural language processing (NLP), and machine learning technologies, as well as more sophisticated approaches to implementing a bot strategy.
Many businesses have embraced the need to deploy digital AI assistants and have rolled up their sleeves with enthusiasm, reaping impressive results that lead to even more bot projects and even better results. We saw a great example of this from one of our insurance clients, the AA Ireland, who decided to focus on a sales bot as their initial project and saw an almost immediate and double-digit increase in conversion rates on insurance quotes.
Each organization’s requirements are unique. From our experience working with all types of industries in different regions, similar challenges and concerns arise when companies embrace a whole new technology like conversational AI.
- Where do you start?
- How do you prioritize AI bot projects?
- How do you roll out a bot?
- How do you avoid early bot failures?
These are just some of the many questions that businesses ask.
No matter what industry you are in or how big your organization is, when it comes to starting out with the first Ai assistant there are some valuable guidelines that can guide you towards success. This blog sets out some guidelines for getting started but is also useful for businesses that are expanding their bot projects to other parts of the organization.
How to Get Started with Enterprise AI Bots
Building and deploying your first conversational bot is like navigating unchartered waters. Most businesses don’t really know what to expect along the way, nor what outcomes to predict. But, how you approach choosing and building your initial solution is critical to its success. It also is key to getting buy-in and budget for further deployments. In this part 1 of the series, I’ll focus on where to begin, whether you’re new to chatbots and just starting out or whether you’re expanding your bot projects to other departments or other use cases.
We also offer a Bot Workshop that helps enterprises brainstorm and identify the best starting points for their chatbots. If you are interested in this workshop please contact us for more details.
1. Choose your Use Cases
The use cases for bot solutions are abundant. Wherever you have a customer-facing or an employee-facing interaction, there is potential to explore how a bot could be deployed to handle the conversation and automate manual tasks and workflows. This step in the process is usually a highly interactive one that can take place in a workshop format, involving the appropriate stakeholders.
Essentially, any business process or workflow that can have a human or conversational interaction is a potential candidate for a chatbot. It may not be the first choice or the ideal use case but it’s better to think outside the box and not confine chatbots to routine customer service interactions.
For example, one of our clients started with a customer onboarding use case, in order to improve the experience and reduce the time and friction involved in getting customers enrolled in their plan. Another started with a sales-driven requirement to drive conversions on insurance quotations. And then we have clients who started with employee-centric bot solutions that give staff easier and faster access to knowledge via a Slack channel. Every company can have a different starting point. It’s all down to the objectives and priorities.
Some industries are particularly interaction intensive, for example, insurance, banking, travel, retail, utilities, and healthcare. Others like business process outsourcing (BPO) and contact centers handle all types of customer requests, including customer service, on behalf of multiple clients. The interactions can span the complete customer lifecycle, from lead generation to acquisition, operations, service, loyalty, and retention.
In today’s competitive labor market, many industries are also turning to chatbots as a means to engage better with their employees throughout the lifecycle from recruitment, onboarding, training, enablement, and other HR-related interactions. Each company has a unique business strategy and objectives that they are striving towards. In brainstorming bot solutions it is important to keep these high-level objectives in mind and to tie each bot use case to them. By doing this you can begin to prioritize a project list.
Now that you have a list of potential chatbot projects, examine each of them in terms of different criteria and requirements. If you’re starting out on your first bot solution, it’s good to aim for a quick win. This means honing in on chatbots that are easy to get off the ground and that are sure to yield some quick and attractive results.
Hence, we always recommend starting off with a pretty easy use case. But how do you decide which project from your list is going to deliver a quick win?
This is where you need different criteria for each potential solution, essentially creating a matrix where you can then more easily prioritize your starting point. So it’s back to brainstorming and drilling in a little bit more to each use case.
2. What is the Primary Goal of the AI Bot?
For each use case, identify the primary objective. Is it onboarding customers faster so that revenue can be generated quicker? Or is it reducing agent handling time or missed calls and live chats? It could be lowering the cost of responding to FAQs or increasing conversion rates on incoming customers. It can even be a goal of being more proactive and reaching out to customers to renew memberships or plans.
3. Who are the Bot Users and what do they Expect?
If the use case is customer-centric, consider what user base or segment you want to initially deploy to and what stage of the lifecycle you’re targeting. If you’re thinking of the acquisition stage of the lifecycle, user expectations may differ to those of a customer service engagement. We often tell the story of a company that wanted to start with a bot to handle complaints. This turned out not to be the best use case to begin with. Complaints, by default, mean unhappy customers. Sometimes it can be better to prioritize projects where users will be friendly and open rather than on the offensive. Updating a customer on their order status, scheduling an appointment, or guiding them through a transaction brings more positive sentiment into play.
If the requirement is for an employee-focused bot, similar considerations apply. Which employee segments will the bot serve? What is the specific employee interaction and what are their needs and expectations for this?
Think about the user group for your initial bot rollout. It’s advisable to target a narrow user base, preferably a friendly one, before rolling out to a broader user group. This allows you to test, gather feedback, and tweak the bot experience before it goes mainstream.
Regarding your user’s preferred language, geographic considerations may guide initial language choices. To begin with, the fewer language options that the chatbot needs to support the better.
4. What Channels should the AI Bot Support?
Having profiled your users, choose the channels of engagement according to where these users will be. To start, you may want to only support a text-based channel rather than text and voice.
If your use case involves simple customer interactions that don’t require access to their account information, public platforms such as Facebook Messenger, WhatsApp, or SMS may be ideal. However, in the case of banking interactions where authentication is key to secure access to information these platforms are less attractive. Channels like the web, mobile app, and email often offer a better authentication and security path in these cases. For example, Bank of America’s chatbot only runs on their mobile app so that they have better control over the security of their customer’s information.
In the case of internal employees, collaboration channels like MS Teams or Slack are often supported by the organization so this may be the preferred channel for employee access.
5. How do you Measure Bot Success?
Whatever the primary goal is, consider how it is measured today. Think about how the bot solution may change this and how you’ll use metrics to measure success. For example, in an onboarding scenario, you may want to measure how many onboarding requests the bot handles and the average time it takes the bot to onboard a new user. But also think of the bot’s impact on the current situation so you get the full picture. By offering around the clock accessibility, the number of dropped calls or chats, average agent handling time, agent productivity, and other metrics may be positively impacted.
After going through this type of brainstorming and prioritizing chatbots according to their goals, a shortlist of bot projects begins to crystallize. However, there are plenty of other considerations before you jump into the build process, for example around automation, data integration, security, human handover, training the bot and more. In the next two blogs in this series, I will address these in more detail.
If you are interested in our chatbot workshop that brings you through a process similar to the one described in this blog, please contact us at firstname.lastname@example.org.