People are staying home and avoiding grocery stores. Frontline workers are taking on demanding schedules. As everyone does their part during these difficult times, we see a high demand for food delivery kits. Customers need us now more than ever.
In the first few weeks of the shelter-in-place order, we doubled our customer base. Existing customers also started ordering more meals, more frequently, leading to a surge in case volume.
As a small business, we needed a solution that would help us quickly scale to support our customers. Having implemented Einstein Bots in 2018, we were prepared — we continue to use the chatbot today to complement our live customer service team and quickly resolve common issues.
If you’re looking to add a chatbot to better accommodate customers, here are four tips to get you going in the right direction.
Identify low-hanging fruit questions
You have the option to give your customers three or more menu options in the Einstein Bot. Start simple and stick to three. You can always branch out from there.
We ran reports to look at volume by interaction type. We identified low-hanging transactional interactions like fulfillment and delivery: “I’m sorry you were missing X ingredient” or “I’m sorry you had quality issues with Y.”
I also worked closely with a team of dedicated admins, premiere support agents, and high-performing agents. These “super agents” helped to map out all of the menu items and the type of clarifying questions the chatbot would ask. It took a few rough drafts and lots of trial and error, but drawing the decision tree on a whiteboard and talking it out helped us hone in on the right questions.
Before COVID-19, we were already using the chatbot to help customers track their orders or packages, report any issues with delays or damage, and get a credit or refund. That has continued through the surge, enabling us to reduce the volume our agents receive. Now agents can work directly with customers to resolve questions and issues around delivery address changes, menu and subscription options, promotions, and gifts.
Bring in customer data
As you build out a chatbot, it’s essential to bring in customer data.
We tied customer data to the Einstein Bot by using an Apex Class invoked by the chatbot. (This guide explains how to do this with your Einstein Bot, and we used the Einstein Bots Developer Cookbook to get started.)
Using the Invocable Method annotation helps you accomplish a lot. It will be your best friend when it comes to grabbing customer data to better assist your customers.
This was a big step forward in improving the effectiveness of the chatbot. It enabled us to use the chatbot to handle transactional requests, like issue credits and refunds and made the chatbot experience what it is today.
Be transparent that you’re using a chatbot
It’s essential to have your chatbot mirror the same tone, diction, and empathy of your agents. But it also needs to identify itself as a chatbot.
When we first launched it, we got negative feedback from customers who had higher expectations of what the chatbot was capable of. I realized we needed to be more upfront and provide clear instructions. Now it says:
The chatbot also uses terms like “helpful humans” (for our agents) and “I’m sorry” (if there’s an issue) to stay on brand and match the empathy of our agents.
By being more transparent, we set clear expectations for our customers. Our CSAT more than doubled afterward.
Meet customers where they are
SMS has become our agents’ most effective channel, with higher CSAT and resolution rates.
Before COVID-19, if the chatbot wasn’t able to handle a question after hours, it would create a case for the agent to follow up via email. Since COVID-19, we implemented the ability for the customer to request a text back instead. And 75% of customers have been choosing SMS over email.
Here’s how it works. The chatbot determines the time in which the chat is taking place. If it’s outside of business hours, the chatbot lets the customer know agents are currently away and asks if the customer would prefer a follow-up via email or SMS. If they choose SMS, we ask for their preferred mobile number (since the number listed on the account may be an office or home number that can’t receive texts.) From there, the chatbot creates a case noting the customer prefers SMS. When the case is routed to an agent in the morning, the agent starts an SMS conversation with the phone number provided.
This chatbot is just the beginning for us. We’re looking forward to adding more Natural Language Processing (NLP) capabilities for the chatbot to make it even more user friendly. We also plan to try Einstein Prediction Builder to streamline agent efficiency.
Being “heroes” for our customers
We have stayed committed to communicating with customers in a clear, transparent way across all channels and agents in both automated and proactive engagements. The response has been positive. One customer noted:
During this period, our customers have rated automation resolved cases at the same level as agent resolved cases: in the low 90% range. We see this as a huge win — that automation can match an agent, even amidst uncertainty.
Customers have called our warehouse staff who prepare our Sun Baskets every week “heroes” because they are “helping them stay safe inside” and serving frontline workers. As one customer shared earlier this month:
That’s the kind of service we’re committed to. Things are changing every day, and we too will continue to adjust as needs change to be the kind of heroes our customers need.
This article is contributed by meal delivery service Sun Basket, a San Francisco-based Salesforce customer using Service Cloud, Digital Engagement, and Einstein Bots.
