Einstein Bots: Using AI to deliver the Future of Customer Service

Why we’re Hot for Bots & What Happened at Dreamforce’17

Two weeks ago at Dreamforce 2017, Salesforce presented Einstein Bots: this is one of the latest products included in the Einstein portfolio: Salesforce’s AI layer spread across its Customer Success Platform. Einstein Bots is in pilot phase and currently being used by Adidas and Hulu.

Honeymoon or My Dreamforce’17 Experience

Over the next paragraphs I’d like to share my exciting experience at Dreamforce’17 showcasing Einstein Bots in a handful of Theater Sessions and Booths.

Introducing MyEinstein: AI for all Organizations

But first, I’d like to take a moment to explain why Salesforce is making a huge investment in AI. Salesforce’s mission is to democratise AI for every organisation. It doesn’t matter how big or small your company is, Salesforce believes AI will bring tremendous value to business and wants to make its AI powered platform easily available: through point and clicks, declaratively, no code. This approach has led to the introduction of MyEinstein at Dreamforce’17.

If last year Einstein’s available features were accessible for the developer community: this year Salesforce is opening these new and existing features, its platform and APIs through a simple point and click approach. Allowing admins and business users to create AI-Powered Apps in ways unimaginable before.

Want to predict the outcome on any field in your salesforce environment, like potential Customer Attrition for instance? Want to create an App that sees then counts and categorises your products’ placement on the planogram? Want to analyse customer sentiment in order to deliver the most adapted answer in a timely manner?

Well, if before you’d’ve needed a team of developers, data scientists and business gurus to get such a party started: today you need a motivated admin [simply because it is better to work with passionate people] and a mouse. “Point, Click, Predict” is one of the mottos I heard during Dreamforce’17 and I think it’s pretty rad.

A little bit of background before the meat:

Today I want to focus on Einstein Bots and before getting into the nitty gritty I’d like to say that for the past years I’ve been spending a lot of my time on the Customer Service Tech Realm and with Bot technology, which we used to call back then virtual assistant” or “avatars. As well, i’ve been playing with Natural Language Processing technology mainly on the voice channel.

Since 2014, while working at Capgemini, I’ve been putting some of these technologies together and building solutions around these very topics. This was very exciting work as well as an enriching experience. My take from back then is that these projects were long and complex [and expensive] to deliver.

Then, I joined Salesforce this June 2017 and have been getting up to speed with its Einstein Magic and being very much involved with its Customer Service Product Offerings and early adopting Einstein Bots, I had the following feelings: (1) I was/am gonna be able to contribute to this beautiful endeavour [which is great for me although likely irrelevant to you] but most importantly (2) The Moment is just Right for Companies to hop in the Chatbot wagon!

An AI and Chatbot Explosion Happening.

The topic is hotter than hot and written by all these magazines: Business, Tech, Journals, News Papers, Specialised Magazines… you name it. “AI and/or Chatbots to transform Customer Experience”, “AI will do this”, “Chatbots will do that”… and what this has created is a lot expectations from AI and Chatbots, which is great!

But in parallel, it has created a little bit of anxiety or fear of…

Will AI Dominate ? Will Chatbots replace my Customer Service Reps ..?

I do NOT believe so. On the contrary: I believe there is a fantastic opportunity for Businesses! Because you see:

Companies trying to build chatbots for a long time.

We likely all remember Clippy, a great idea, but a very frustrating experience: disconnected from the user’s context. Some of us still remember Anna, the Ikea Virtual Assistant, very polite but kept repeating itself over and over at some point.

The problem with these conversational experiences is that they were disconnected from the Customer, disconnected from the Business, disconnected from meaningful Data.

But I believe there’s a great opportunity here for us because there are essentially:

3 shifts in Market preparing us for this very moment!

  1. The Audience is matured, out there, on mobile, on every channel and [more than] ready to accept these AI/Chatbot technologies.
  2. The Technology itself is matured and Salesforce is bringing all of these within Einstein.
  3. Ecosystem: a mayor appetite from consumers and with companies trying to use these tech and trying get it right.

And this has resulted in a Chatbot explosion. There’s an incredible ecosystem out there. Tons of tactical [and awesome] solutions out there that are delivering specific [and amazing] things!

While this is fantastic news for the community, IMO it is difficult for Companies to find a sweet spot. Some solutions allow building bots very fast, no code, but [perhaps] siloed, on a channel… while others very open but fully code dependent for instance.

When it comes down to Business it turns out to be a booming but yet disconnected landscape. Without [tedious] integrations, it remains disconnected from the Customer, Business Processes, and CRM Data.

Does this mean that we’re doomed to fail again? Not quite! :)

Here’s the secret sauce.

What we believe are the 4 key elements for a successful conversational experience.

  1. You need to know your customer’s context. What your customer is all about. What he/she have purchased. What they are looking for…
  2. You need to have connected all the data: CRM Data, Customer Service Data, Other Back End System Data so that your Bot can properly take action and provide meaningful personalised and contextualised answers…
  3. And if the Bot does not understand because the request is too complex, it needs to be able to seamlessly transfer to a Human Rep who can then take over the conversation holding its prior context and being able to resolve the case intelligently.
  4. Finally: creating, training and deploying Bots needs to be easy. Declaratively, with no code.

This is where Einstein Bots come in, and that’s why I was so excited to contribute to these announcements and presentations at Dreamforce 2017.

Here’s a Demo Snippet:

Amazing Shoppers® has implemented Einstein Bots. Here’s how a customer, Carlos, who recently purchased a few things from them can update his most recent order.

Amazing Shoppers® holds a VIP policy. Any customer purchasing more than $500 is eligible and will get a gift and more. If this conditions is met: the company has decided the Bot should transfer the conversation to a Human-Rep in order to cocoon the customer, confirm the gift and the change of status while gently pushing an up-sell or additional services ;)

Here’s how Amazing Shoppers built its Einstein Bot:

Below is Amazing Shoppers® Chatbot Builder. This is where the company is able to define Scarlette’s [its bot] personality: which dialog should it be able to address? Which data should it collect from the customer and fetch from Salesforce in order to provide a meaningful conversational experience to its customers?

In short, Einstein Bots can respond, ask questions, collect data, run scripts and launch process rules : all for the benefit of a smooth conversational experience.

Mix: Menu Scripting & Natural Language Processing

a) Amazing Shoppers® has defined more than a few dialogs that are NLP powered. The company has initially trained the bot manually, leveraging the intent tab in order to get the bot quickly up and running. They will also be able to train the bot with CRM Data, Customer Service Data, previous Live Chat Transcript Data, newly inputed customer requests… automatically and/or in a regular basis.

b) Einstein Bot are deeply connected to Salesforce, thus providing consistent dialogs surfacing relevant data to the Conversational Interface.

Seamless Hand Off to Customer Service Agents

When in doubt: transfer to an agent! Or, in Amazing Shoppers® case: it is a deliberate Business/Strategic decision, to hand off conversations to agents when customers are eligible for the VIP program.

Agents can then take over where the conversation where it was left off, and because they hold the prior context, they can focus on what matters: efficiency, CSAT, and up sell for instance.

Data and Customer Context on Salesforce:

And remember, all of this Data and its context used by the Bot is all in SFDC. It is then you, that is Amazing Shoppers® in this case, who decides which data you to want surface.

In a nutshell:

Coming from the Virtual Assistant/Avatar/NLP world that we call today Chatbots, what excites me the most about Einstein Bots presented at Dreamforce 2017 is the ability to create deeply connected bots with CRM Data, Customer Service Data, Previous Live Chat Transcript Data, Case Data, Knowledge Base Data, and any Back End System Data brought up to Salesforce. Because it is native to the Customer Success Platform: creating, training, and deploying Einstein Bots is made easy and declaratively which is fantastic.

What this does or will do is allow organizations to deliver connected service at scale, handling recurrent, known, repetitive, low to medium added value type of requests automatically. Because, let’s not fool ourselves: companies will need Humans (at least for now :)… for the critical, the hot and juicy leads, the high added value type of requests… we’ll need the human-touch for sure!

Last but not least and my personal favourite is the deco-relation between the chatbot builder and the channels it will serve. This is key and will allow companies to “build once and publish everywhere”. A translation layer will handle the differences between channels, ie: differences you can find between an SMS, Web or Messenger channels. Really cool.

So, let’s use our valuable Human Ressources for what is the most valuable indeed. For the rest: have bot, deeply connected to your business data, handle [some of] it!

Thank you. If you found this article interesting, please 👏 below, I’d really appreciate it. You can also follow me here on Medium, and/or on Twitter @lecharleslozano