With its Summer 2018 release, Salesforce chatbots will soon become a standard offering of Service Cloud via Einstein Bots.

This is not the first time Salesforce has introduced chatbots as a capability—the company has supported chatbots with platform features, and social bots via plugins with Facebook Messenger or through Live Agent. However, Einstein Bots’ click configuration and easy-to-set-up natural language processing engine make this technology accessible to a broader audience, and represents a major leap forward.

Related: Chatbots Integrating with Consumer-Brand Experiences

As Business Insider previously reported, 80% of businesses hope to use chatbots by 2020. Given the impact AI can have on cutting costs, it’s easy to see why—the publication also estimates that the use of chatbots can result in more than $23 billion in annual industry savings. Einstein Bots are in line with where the customer service industry has been moving for the past couple of years and narrows the gap for brands hoping to leverage this technology.

I’ve had a chance to play around with the capability ahead of the GA release, and want to share some considerations you should keep in mind if your company is planning to implement the new technology.

1. Don’t try to fool your customers.

Tell them they are talking to a chatbot, and make routing them to an agent as simple as possible. Throughout every step of the conversation, getting to an agent should be a single click or utterance away. Not everyone will want to interact with your chatbot initially, and that is ok!

2. Don’t try to do everything.

When first creating a chatbot for your organization, resist the temptation to try to do everything.

When first creating a chatbot for your organization, resist the temptation to try to do everything. Keeping things simple will help with your ability to turn out a quality product and will make it easy for your customers to start interacting with your chatbot. Think of the top questions your customers ask and have your chatbot present a menu of options prompting what it can do. As you expand the capabilities down the road, the menu of items can evolve, and you can introduce more natural language options.

salesforce chatbot

3. Your chatbots work 24 x 7. Your agents might not.

Anytime, day or night, your chatbot is available to answer questions, but that does not mean there is always an agent to take over when customers do not get the answer they want. When designing your chatbot, keep this in mind and plan accordingly. Make sure your chatbot tells customers when someone is available to help in person, and if not, create a case for followup or point the customer to your website for more details.

4. Make it personal.

Just because it’s a chatbot, does not mean standard customer service rules do not apply. Customers will still expect a personal touch. If you are capturing details about the customer during chat initiation, such as name and email, use this information during the interaction. “Hello, David …” is a nice touch, and make sure you do not ask for an email address a second time if you already have it.

Just because it’s a chatbot, does not mean standard customer service rules do not apply.

5. Avoid information overload.

Crate a menu tree to guide your customer to where they want to go. Keep in mind, your customer may be using a mobile phone to chat, and the real estate can be limited. Some rules of thumb: if you have more than five or six menu options, start to group them, and if you start to build out a tree that looks like an IT org chart, think about using natural language processing instead to maintain engagement.

6. Use natural language processing.

A great chatbot is the right combination of guiding your customer to the information they need and making available natural language processing for ease of use. Remember, not every menu option needs natural language assistance, but when it comes in handy, you will need to account for training your chatbot with large datasets.

Fortunately, Salesforce has made it easy to train your chatbot and allows for bulk loading of example utterances. Salesforce also has available lists you can use from their AppExchange. To stay in line with best-practice, you should train your chatbot against a minimum of 150 utterance examples.

salesforce chatbot

Einstein Bots’ ability to incorporate into existing customer support channels, such as Live Agent, make the use of chatbots more prevalent and accessible than ever. Now is the time to be thinking about how your organization can take advantage of this powerful digital tool.

If you have questions about chatbots or want to try integrating them into your customer service practice, reach out and we’d be happy to help.

Keep In Touch

Stay up-to-date on the latest digital trends, DEG news, and upcoming events by subscribing to DEG's newsletter.


Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>