Providing a differentiating customer experience starts with reimagining how you can bring your customers closer to your products and services. This needs to happen at a time chosen by the customer, on a device that the customer has access to and as quickly as possible. Companies make a ton of investments in their digital assets like a website or a storefront. Turns out that this “treasure hunt” exercise on a complex website can often be frustrating for some, so we add a nice search bar to help customers search for what they need. But a search bar can be intimidating if you don’t know what keywords are required to bring the information you are looking for. This has now evolved into an intelligent search that provides your customer with a search result that is more intuitive and sorts the results based on relevance. This also leaves much to be desired as such intelligent searches behave very differently in different sites, the quality of results is based on the data being queried and the relevance of the keywords used. So what can top this experience while giving a personal touch?

Consider a Natural Language Processing (NLP) based approach that supports a conversational experience. The conversational style of engaging with individuals comes naturally to everyone and bringing that approach to customer service can dramatically change the customer experience. Fortunately, such capabilities are now available baked into Enterprise Software platforms like Salesforce to provide a unique customer experience that can be tailored for any customer in any industry. Einstein Bots is one of such offerings that brings this capability for enterprises with investments in Salesforce to build a unique conversational experience to help customers get closer to the product and services offered by the company. Enterprises can now unleash the data in their enterprise applications to help craft a personalized experience for every customer and design a conversation based on who is coming to the digital storefront – anytime, anywhere and on any device.

So how hard is it to introduce a Chatbot as part of your digital strategy? Turns out that if you have made investments in the Salesforce platform this capability is just a simple extension of the investments made in the platform. Rootstock’s deep and broad ERP offering built natively on the Salesforce platform can now leverage this capability to get customers, suppliers and other stakeholders closer to the products and services in ways that was not possible in the past. Getting sales order status related to a specific order like shipment date, item related information like pricing or availability etc. are some of the many use cases that can be now made available to customers on a 24×7 basis. Similarly, Vendors can also track the receipts, payment details in a conversational experience. By providing self-service capabilities, stakeholders of your business feel empowered to access the information at any time without the need to wait for an agent to be available. While this enhanced experience increases engagement, it also helps companies reduce the cost of support operations. When companies are looking to get the biggest bang for the buck for their investments, this investment provides a quick reduction in costs while increasing customer satisfaction.

So how does one go about building a chatbot? There is plenty of information available today on effective ways to build a chatbot using Einstein Bots. However, effective bot design is important to drive a higher level of engagement and trust. Here are some design considerations for the Chatbot that will help improve the engagement and drive successful adoption of your self-service initiatives.

Humanize

Effective conversational experience needs to mimic human interactions as much as possible. A robotic chat experience can leave the user disenchanted and reduce the stickiness of the conversation. Bringing the experience closer to a natural conversational style should be the ideal goal. This means having a name for the bot, a personality, acknowledging that the bot may not be trained on all aspects – these are some critical considerations. We have plenty of examples with digital assistants like Alexa, Siri etc. to model these human-like experiences and bring that in the context of your enterprise data.

Personalize

The user is engaging with the Chatbot after logging into a customer or vendor portal. Chatbots should be designed to understand the 360 Degree view of the customer or vendor – order history, case history, payment history, previous interactions, previous transactions etc. Such information should be communicated to the user at appropriate points in the time of the conversation, so the end-user truly appreciates that the bot has a good understanding of his/her specific needs. For example, Not having to ask basic questions like “What is your Order Number?” and instead provide prompts like “I found 3 Orders placed recently from you – 20201, 20224, 20300. Which of these do you want some more details or enter your order number?”

Contextualize

Conversations become natural when Chatbots provide continuity fetching the right information specific to the user in a prompt manner. Using information like time, day of the week, location of the user and recent transactions are good ways to kick off the conversation. Making use of responses from previously answered questions and avoiding prompts that repeat the same question becomes equally important to avoid frustrations creep in for the user.

Modularize

Breaking the conversation into smaller parts is important from a design aspect as it has a number of benefits. Intent identification of user questions becomes easier to manage and can help you to guide the user to the right bot responses. It allows more reuse and simplifies the overall bot design. Training for Named Entity Recognition (NER) and Intent identification becomes much easier to manage. This will also give the end-user the ability to move from a conversational experience to a pre-determined hierarchical Menu based experience easily based on the end- user preference.

Analyze

The Chatbot you build initially is never perfect. It requires fine-tuning and training on a regular basis to increase its ability to understand user intent to provide accurate responses. When your chatbot is open for global consumption by users in multiple geographies it introduces an element of variation in spoken nuances. There is a good chance that your Chatbot is going to be confused and may not understand the responses. Analyzing the failures, prompts for questions like “Did that help?” etc. become important to drive a continuous learning process. Einstein Bots gives you the ability to track user engagement, log failures. You can use this information to continuously train your bot on an ongoing basis. With Einstein Analytics, user engagement can also be analyzed. Proper A/B testing for different versions of the Chatbot can also help to figure out the right approach that drives maximum engagement.

At Rootstock, we are excited about the possibilities that come with the introduction of AI-based self-service that can be done with Salesforce Einstein Bots. So, what does the future look like? As the capability evolves in the platform a much more intuitive digital assistant will evolve. One can see that such NLP capabilities when made available on a phone, can truly enable voice-to-text interactions. You could simply ask a question like “What is the shipment date for my order number 56”. Enterprise digital assistants based on true Named Entity Recognition (NER) capabilities will recognize entities (i.e Object – Sales Order) and look for information on the specific order (i.e order number 56). It will be more flexible to understand the Parts of Speech (PoS) and quickly determine the nouns (product names, city, country), verbs(action), metrics (total, average etc.), dimensions (Product, Time, Division etc.). NLP based digital assistants will use that to understand the correct intent and get the right information that the end-user is looking for. One can argue that the speed at which such information can be processed and delivered to the end-user could be faster than a human agent and that can become a true competitive differentiator. Having said that bots are not meant to completely replace human agents in every scenario. There are many scenarios that will require human intervention to resolve complex situations that would be difficult for the bot to provide a resolution. The ability for consumers to engage in purposeful conversations to look for information 24×7 will become a natural extension of existing digital assets that enterprises have invested today.

Contact us to see how we can help you bring self-service capabilities in your Manufacturing, Supply Chain and Finance functions to enhance the overall customer experience.