Get ready for Einstein bots in 5 steps

Make your company care about AI and Automation

9 min readDec 13, 2019

The preamble contains answers to common questions about Einstein Bots. If you’re up to speed, skip down to “How do I get started with Einstein Bots?”

What is an Einstein Bot?

Einstein bots are, quite simply, a chatbot. For those unfamiliar, chatbots simulate conversations with humans through a variety of different mediums, but most commonly through the web by means of a chat window — or — through applications such as Facebook Messenger, WeChat or SMS. Chatbots are configured to automate simple tasks such as answering FAQs or complex tasks such as qualifying & routing an inquiry to specialized agents.

Gartner and Forrester predict 85% of businesses will use a chatbot by the year 2020. True success of a chatbot lies in its ability to provide a contextual conversational experience.While there are many chatbot providers in the ecosystem, Salesforce Einstein bots have distinct advantage of natively connecting to your Salesforce CRM (customer relationship management software). This means your Einstein bot can search & perform any action within your CRM as your support agents do today. This product differentiator is rare and extremely powerful for providing a tailored and personalized conversational experience.

What benefits can I expect of Einstein Bots?

Business units leveraging Einstein Bots benefit from scalability (processing unlimited volume of concurrent conversation requests), availability (24/7, no days off) and affordability (cents on the dollar) of Einstein bots.

For example, sales business units deploying bots see an uptick in lead conversion due to 24/7 availability of your bot to capture leads anytime and anywhere on your website. Support business units deploying bots see a downtick in support costs as simple, repetitive questions once escalated to a human are now entirely answered and deflected by Einstein bots.

These top line and bottom line benefits are just the beginning of the impact Einstein bots bring to your company.

Fun Facts about Einstein Bots

Einstein Bots are free for Salesforce chat users (formerly LiveAgent Chat). Every 1 license of Chat lands you 25 free bot sessions per month. Navigate to Setup > Einstein Bots & flip the toggle to turn on Bots for your org.
Early adopters of Einstein Bots have seen an average of 60% deflection and 20% reduction in total ticket handle time. This can translate to millions of dollars in savings depending on your ticket volume & ticket cost.
Einstein bots are natively available in the following languages: English, French, Italian, German, Portuguese and Spanish. If you have your own NLP provider, you can integrate these provider with Einstein bots seamlessly. Additional languages become natively available in the coming years.
Early adopters of Einstein Bots have seen an increase in CSAT by 15% when bots & agents work together.
Over the past year the bot builder has become more declarative than ever, significantly reducing setup time from 6 months to less than 3 months.
With a click of a button, you can bring bots to the following channels: WebChat (GA), Mobile Chat (GA), SMS (GA), Facebook Messenger (Pilot), WhatsApp (Pilot) and WeChat (Pilot). How amazing is that??

What pre-requisites do I need for Einstein Bots?

Setting up Einstein bots can be significantly easier if the following questions are a resounding “Yes”:

Do you currently have a digital channel your customers use to interact with your business (i.e. SMS, FBM, Web-chat, WhatsApp…)?
Do you have transcripts of conversations between your support agents & your customers, on your digital channels, for the past 30 days (or more)?

If you’ve answered “No” to any of these, setting up an Einstein Bot may be more difficult, therefore I strongly advise ensuring both requirements are setup within your environment prior to building Einstein bots. You also have the option of setting up requisite #1 concurrently with your Einstein bot, but you may run into delays when building out topics for your bot to handle if requisite #2 is not available at your company.

How do I get started with Einstein bots?

Building a bot is akin to building a fleet of support agents — there are teams, learning material and programs built before hiring and training the very first class of support agents. Similarly, there are 5 milestones (highlighted in teal) which, if achieved, springboard your bot building experience. From there, you’ll be equipped to begin building your very first rules-based bot and then, eventually, be ready to add natural language processing.

Mile marker # 1 — Identify skills your bot will handle

Historical transcripts of your customers interacting with support agents are critical for this mile marker. Skimming transcripts exposes repetitive tasks agents are performing on each support ticket. There are two types of tasks to double-click into: (1) questions agents ask your customers & (2) questions your customers ask agents.

Probing questions agents ask your customers include inquiries to identify who the customer is (name, email, phone…), determine the issue they are inquiring about and to collect business specific information (order #, device type, invoice #, document #, device #…). These types of questions are best handled entirely by your Salesforce Einstein bot who will collect answers to these questions to pre-qualify the ticket.
Factual questions your customers ask agents may include: “I forgot my password & need help to reset” or “What is your return policy?” where your support agents are tasked to recall specific pieces of information. These questions are also ideal for Einstein bots to handle & deflect from your agents entirely.

If historical transcripts are not at your disposal, determine skills your bot will handle through reporting on historically closed tickets. Generally, within a CRM, agents tag every ticket they solve / close with type of issue category or resolution category. A report which summarizes record count by issue category will be your first step towards determining which issue categories are simple enough for your bot to handle.

Mile marker #2 — Identify skills your bot will not handle

Historical transcripts further demonstrate types of questions your bot should probably not handle on its own and would benefit from a human-in-the-loop (i.e. escalation). For example, divergent questions requiring abstract thinking to arrive at the correct resolution such as:

“I’m unsure what size I should purchase for this jacket. I’m usually a size 2 at most retail stores but I’m not sure how your sizes compare”

or affective questions requiring proper management of feelings or attitude such as

“The individual who delivered my order was extremely rude!”

always require Einstein bots and agents to work together for proper handling and care. The best way to think about Einstein bots & agents working together is during an escalation. The examples of divergent and affective questions above are great examples where Einstein bots can quickly identify intent & can dynamically escalate to a specialized human agent, skilled in resolving the unique issue, on the first try. In contrast, lack of Einstein bots require your customer to endure a daisy-chain of manual escalations, going from Tier 1 → Tier 2 → Tier 3 → Specialized agent. Context may be lost in each escalation step forcing your customers to become frustrated in repeating themselves. Einstein bots streamline your escalation process for questions benefitting humans-in-the-loop.

Mile marker #3 — Determine availability of your bot

Your customers can encounter an issue or question with your product or service at anytime of the day. Einstein bots make the biggest splash when used 24/7 at a company.

One consideration when deploying a 24/7 bot is ensuring a fallback protocol when no agents are available for an escalation. An example of a fallback action is to use the Einstein bots to pre-qualify the request, and, once qualified, provide your Einstein bot with permissions to create a support ticket in your CRM. This ticket can be escalated to an out of business hours queue to be actioned by agents when they become available during regular business hours.

Mile marker #4 — Calculate ROI

We’ve identified three major areas your Einstein bot will impact: qualification, deflection & escalation to specialized queues. To your business, these terms align to reduce agent handle time (qualify), reduce ticket volume (deflect) and reduce total time to resolution (specialized escalation). These Northstar business metrics can project savings associated with your Einstein Bot. Here’s a step-by-step ROI projection calculation:

Choose a support channel for your bot to impact. We’ll use webchat in this example.
Choose the availability of your bot. We’ll use 24/7.
Calculate today’s all-in cost to handle a single support ticket. We’ll use $5.50 per ticket.
Calculate today’s average time (minutes) to resolution (solved date/time minus opened date/time) for tickets solved from the support channel you selected. We’ll use 25 minutes.
Determine today’s daily volume of tickets for your support channel. We’ll use 1,000 daily tickets for webchat.

Now, its simple math. If your bot reduces average agent handle time & time to resolution by 5 minutes (20% reduction on average), you amount a total of $1,100 in daily savings:

$5.50 per ticket ÷ 25 minutes = $0.22 per minute
$0.22 per minute x 5 minutes = $1.10 per ticket savings
$1.10 per ticket savings x 1,000 daily ticket volume = $1,100 saved everyday

If your bot, skilled on deflecting cases, is able to resolve 62% of your tickets entirely on its own without escalating to an agent, you’ve amount a total of $3,300 in daily savings — or $1.2 million per year!

1,000 daily ticket volume x 0.62 = 620
$5.50 per ticket x 620 = $3,410 saved everyday
$3,410 x 365 = $1,244,650 saved yearly

With 25 free bot sessions for each chat license in your Salesforce environment, these savings can communicate a compelling ROI story for business stakeholders to fund your Einstein bot project.

Mile marker #5 — Build a team to build your bot

You now know topics your bot handles, topics your bot escalates to a human, channels your bot services, when your bot will be available and lastly, projected ROI savings from your Einstein bot. You’ve practically completed the work of a Product Owner for Einstein bots with this exercise. However, you must remember bots are most akin to human support agents & will require development and continual training. Therefore, it is best to build a team of dedicated resources to maintain & advance your bot. The roles and responsibilities of teammates you should consider are:

Product owner: identifies the business use case for Einstein bots, convinces stakeholders of investment & ensures all team members are working towards predefined problem statements
Administrator: configures dialogs in the Einstein bot builder, integrates Einstein bots with channels (website, SMS, Facebook…), integrates your bot with external systems, versions your bots & frequently manages natural language processing training and development
Content Writer: crafts bot messages to ensure corporate messaging, tone & empathy are conveyed, identifies gaps in conversational design & suggests improvements to the conversation path

Next Steps

Once you’ve tackled these milestones, the next step is to begin building your Einstein Bot. The very first Bot you should build is a rules-based bot. Deploying a rules-based bot is quick, effective and illustrate how your customers interact with your bot. From these interactions, you determine bot dialogs which should be enabled to process free text inputs. This article, along with two additional articles, form a trilogy of instructional material for building an advancing Einstein bots:

About the Author

As a former Salesforce customer, I’ve implemented Salesforce (Sales Cloud, Service Cloud, Pardot), AI tools and telephony systems for 2,000+ users across almost every business unit at a high growth startup — Doordash. Salesforce then hired me as Director of Product to build Einstein Bots with admin-centric features to make bot building intuitive, friendly and powerful. Today, I serve as the CEO of CloudX (www.cloudx.design), a Salesforce implementation firm specializing in Zendesk to Service Cloud migrations, Einstein bot implementations, digital channel development and more. We’ve successfully implemented Einstein bots for companies of all sizes and have done so in as little as 5 weeks. Drop me a line to learn how CloudX can springboard your CRM/AI/Automation strategy!