Not all chatbots are created equal. In this post, we’ll cover a step-by-step chatbot demo guide of what to look for, including the key questions to ask, to make sure that you adopt the best AI for your business.
Before we get into the questions you should ask during your chatbot demo, let’s quickly cover why now is a good time to explore chatbots.
Why look at chatbots at all?
Automating customer service, meeting evolving customer preferences and expectations, 24/7 availability and ‘always on’ support are among the notable features and benefits of AI-powered chatbots. By this year (2022), chatbots could help to cut business costs by more than $8 billion per year, this same research also estimating that average cost savings will be within the range of $0.50-$0.70 per interaction. Are such cost savings statistics relevant for your business? If yes, then you want to get a demo added to your schedule in the not-too-distant future.
It is great to hear or read about the benefits of chatbots, but before you adopt the technology, it’s critical to understand the differences between chatbot vendors.
If you’re evaluating chatbots for the first time, you might not know what questions to ask or what to look for. What are the key features that need to be touched on, and that will make your chatbot a success? How many external resources are needed in order to launch?
Without further ado, let’s take a look at the top questions you should ask during all your future chatbot demos.
- Does It Connect to Your Knowledge Base?
- How Do You Train The AI?
- How Do I Create Flows?
- How Do I Measure and Analyze Success?
Armed with a strong ‘Do it yourself (DIY)’ attitude, many users today like the concept of customer self service, and of helping themselves. We see this with online banking, or with travelers choosing to check their bags via an airport kiosk. To meet the demands of the modern customer, companies over the past several years have created robust Knowledge Bases (also referred to as Help Centers) on Zendesk and other leading customer experience platforms. Such resources contain articles covering recurring topics and questions on products and services.
Yet the joining of the best knowledge base software, used in parallel with AI-based customer service chatbots is the ultimate combination, offering the perfect combination. Why? When interacting with customers, the most advanced chatbots can summarize and provide the most relevant information to a customer, even if it hasn’t been explicitly trained on a certain topic.
This is much more advanced than a bot that simply picks up on keywords and provides a link within the chat (here, essentially the bot becomes simply a searchable knowledge base). For the best customer experience, don’t link out to articles from a chatbot, but provide all relevant information within the same interface.
Follow up questions to ask:
- How easy is it to integrate with my knowledge base?
- Can the bot go beyond just linking out articles in a knowledge base, and actually provide a full resolution within the chat interface?
- How quickly can integration with my knowledge base be done?
- What internal resources are required for this integration?
- Can you use our knowledge base to train an AI chatbot?
How will the AI be trained – is it manual or is it done automatically? Netomi’s Clustering Engine, for instance, utilizes unsupervised learning to sift through and cluster your historic tickets, honing in on repeatable scenarios based on semantic affinity that may be automated, that is, phrases belonging to the same semantic category.
Common queries, such as ‘ When will I get my money back,’ and ‘I haven’t received my refund yet’ may be grouped together, under the category of ‘refund status.’ Automatically merging these categories together significantly reduces training time and reduces the effort needed to launch an AI that understands the various ways in which a person might pose a question.
In one example, Comcast found that there are 1,700 ways that a person might inquire about the straightforward issue: ‘I want to pay my bill.’ Without using unsupervised learning, an AI might not be able to initially understand that ‘I’d like to settle my account’ carries the exact same meaning, unless it has been explicitly trained to recognize these subtle differences.
Follow up questions to ask:
- How much effort is required from the company to train an AI?
- Can an AI be trained in multiple languages?
- How much training data is required?
Discover the return on investing in AI for customer service with Netomi’s ROI calculator!
While some customer queries are simple FAQs, others require more in-depth conversations. In order to deliver the most relevant and personalized experiences, you might need to ask a customer for an order number, email address, or preferences for flight times. In these instances, chatbots will need to have more complex workflows built. So in a demo, getting a sense of how these flows are created, how arduous or simple the process is, and what internal skillsets are needed is important.
For instance, we at Netomi recently introduced our Visual Response Builder, enabling customer service teams to create user workflows via a simple drag-and-drop interface. Using this tool, teams can easily configure responses, use conditional branching to unlock unique flows based on a customer’s response, and easily escalate chats to a live agent, when needed.
Follow up questions to ask:
- In a workflow, what can be configured?
- What UX tools are available? (buttons, cards, carousels, etc.)
- How difficult are flows to create, and would one be able to create them on their own?
- Is access to this tool available out-of-the-box?
- Does creating flows require any coding?
- How am I able to add personalization to flows?
Prior to deploying a chatbot, there needs to be some way of gauging performance. Is it working as intended? How are your customers adapting and responding to this new tool?
This is why having analytics in place is key. You need to know what analytics your chatbot provider has available, so you can make sure you can track the KPIs that are most important to your team.
Netomi’s analytics dashboard, for instance, highlights real-time performance data on key customer service metrics, including AI resolution rate, in-depth topic analysis, and engagement metrics. It also offers the ability to track how well the AI is performing, and how this is changing over time. Are there any new categories of queries that can potentially be automated?
When it comes to analytics, here are some follow up questions to ask:
- Can I drill down for different date ranges?
- Can I customize analytics?
- Can I export data?
- In what ways can the chatbot be optimized after launch?
Final Thoughts and Key Takeaways
Not every chatbot platform is the same. Before you adopt a chatbot partner, you need to take a strategic approach by figuring out the effort it will take from your team, the steps that are involved, and all tools, analytics and features available, in addition to the user experience it will provide.
When evaluating a demo, take a hard look at its features, and don’t be afraid to ask questions to get the best live chat software for your site!
Now that we have covered the basics of what to look out for, is it time to see a chatbot demo in action? Connect with us at Netomi, and we will demonstrate how you can easily build workflows and integrate with your back-end systems, and scale your support across every channel.