Contact centres are coming under increasing pressure as their customer inquiry volumes grow. That growth is fuelled by a number of factors, including:
- An expanding customer base
- Releasing new products or services
- Entering new countries and adding new languages
- Launching new channels such as SMS, WhatsApp, Facebook Messenger, Social Posts etc.
- Anticipated seasonal spikes and unexpected emergencies
How do contact centres that use Salesforce Service Cloud handle increasing volumes, whilst maintaining or improving customer satisfaction? How can this be done without a proportional increase in costs? AI is a viable option, as it automates the closure of repetitive inquiries and assists service agents in resolving the remaining cases; this reduces costs and frees up agents to work on more complex, higher-value activities. At Dreamforce 2016, Salesforce launched Einstein, its portfolio of AI app features and platform services, following a series of acquisitions that included companies such as PredictionIO and MetaMind.
I’ve been working in the Salesforce ecosystem for eight years, bringing to market AppExchange products that are built on, or integrate with, Salesforce. For the past four years, my focus has been Salesforce Service Cloud and customer service automation using AI. During this time, the range of Einstein options has grown and matured significantly, which can take time to navigate. If you use Service Cloud, and you are wondering which Einstein features can enhance your customer service, this post will jump-start your Einstein journey by doing some of the initial research heavy-lifting for you.
The post, and the accompanying cheat sheet, will summarise the Einstein features that are available for Service Cloud, outline some typical use cases, describe the licensing models, note any prerequisites and tell you where to find further information.
Note: Community cloud, which plays a vital role in customer support self-service and also benefits from Einstein, is not covered in this post.
Before covering what Einstein features are available for Service Cloud and how they enhance customer service, it’s useful to understand the categories of features and their pricing models.
Feature Categories and Pricing Models
Einstein features for Salesforce Service Cloud fall into two categories:
- App features
- Platform services
And they are made available to Salesforce customers in three ways:
- Bundled with the Service Cloud User feature license
- Included in an add-on SKU
- Usage-based licensing applies
Note: Although Salesforce has done a great job of simplifying their pricing and licensing models, with full transparency on their website, there is some “devil in the detail”. In the interests of simplicity, the cheat sheets do not contain every possible licensing option. Please always contact Salesforce or your preferred consultancy for up to date, accurate pricing.
Einstein app features are standard functionality of the Salesforce Service Cloud app, meaning that they have the same out-of-the-box functionality for all customers. For context, examples of standard Service Cloud features that aren’t from the Einstein portfolio are Omni-Channel and Chat. As with these non-Einstein standard features, a Service Cloud license will bundle some Einstein app features at no additional charge; whereas, add-on SKUs will need to be purchased to add the remaining Einstein app features. Here are two Einstein app features, one bundled, the other licensed via an add-on SKU:
- Service Cloud Enterprise Edition or above includes Einstein Article Recommendations.
- The Service Cloud Einstein SKU adds the Einstein Next Best Action, Einstein Case Classification, Einstein Case Routing and Service Analytics app features to the Service Cloud app.
Einstein platform services are used by developers to extend the standard Salesforce Service Cloud functionality by creating custom apps that incorporate Einstein features. Platform services become available by buying an add-on SKU or by subscribing to usage-based licensing. For example:
- Utilise Einstein Prediction Builder by purchasing the Einstein Predictions SKU or the Einstein Analytics Plus SKU.
- Use Einstein Language intent models or sentiment analysis models by buying multiples of 1 million predictions a month.
Armed with an understanding of the Einstein feature categories and pricing models, we can dive into the 10 Einstein features for Service Cloud.
10 Einstein Features for Service Cloud
As of June 2020, you can add the following generally available (GA) Einstein features to Salesforce Service Cloud to enhance your customer service.
Note: Please refer to the accompanying cheat sheet for licensing and prerequisites.
New features, and enhancements to existing features, are frequently released. For example, the upcoming Summer ’20 release includes a new pilot feature called Einstein Case Wrap-up, that assists service agents with field filling after a Chat conversation has completed.
I’ll now briefly summarise each Einstein feature.
Feature 1: Einstein Article Recommendations
This feature became generally available in the Spring ’20 release and is bundled with Salesforce Service Cloud Enterprise Edition and above.
If you have 100+ English knowledge articles and 1000+ closed cases, you already have everything that you need to build an AI model using a simple, three-step wizard. The model learns what Knowledge articles helped agents to solve cases in the past so it can recommend articles to help resolve new cases.
Service agents access the recommended articles, ranked in order of relevance, using the Knowledge Lightning component in the Lightning Service Console. From there, an agent can attach one of the recommended articles to a case, edit an article or mark an article as “not helpful”. Einstein becomes smarter over time by learning from agent interactions in the Knowledge Lightning component.
Feature 2: Einstein Reply Recommendations
Einstein Reply Recommendations trains a predictive model using past Chat transcripts to identify common responses to customer inquires in the Chat service channel. Once you have built the model, you can select, edit and publish the replies that you want to make available to service agents.
Reply recommendations appear in the Lightning Service Console by adding the Einstein Replies Lightning component. Service agents can select one of the recommended replies as a chat response and optionally edit it beforehand.
Note: At the time of writing this post, this feature is in release preview and will be generally available in the Summer ’20 release in July 2020. Licensing information is, therefore, to be confirmed.
Let’s now move on to how these ten features can enhance your customer service by describing some example use cases.
Feature 3: Einstein Next Best Action
Einstein Next Best Action (NBA) makes recommendations to users based on customer data and business logic. For example, a service agent is resolving a case that is negatively impacting a customer’s satisfaction, and they happen to be coming up for a contract renewal. Based on the customer’s contract length, purchasing history and current CSAT, the recommendation might be to offer a two-year contract rather than one and include a discounted service or product. Recommendations can also be actions, rather than offers; such as suggesting that a proactive email is sent to the customer apologising for the issue.
Interestingly, Einstein NBA doesn’t include any AI out-of-the-box, although you can add Einstein, or third-party AI, models to influence the business logic. We will see this pattern again with Einstein Bots, whereby you can create a transactional Einstein Bot that doesn’t use AI or a conversational Einstein Bot that uses natural language processing (NLP).
Einstein NBA recommendations appear in Lightning Service Console using the Einstein Next Best Action Lightning component. The service agent can select an action for a recommendation, such as “Yes, send the offer”, and Flow will automate the action.
Surfacing Einstein NBA to service agents keeps them in lock-step with the rest of the business when it comes to offers and actions, ensuring that the customer experience is consistent across their marketing, sales and service journey.
Feature 4: Einstein Case Classification
Einstein Case Classification uses historical cases to train its AI model. It learns how service agents have set lookup, checkbox and picklist field values on cases in the past so that the model can recommend field values for new cases. For example, Einstein Case Classification can recommend values for the Type (the type of inquiry) and Reason (the reason for the inquiry) fields based on the contents of the Subject and Description fields.
Recommendations appear in the Lightning Service Console using the Case Details Lightning component. A service agent can select a recommended field value, or Einstein can automatically set a field value if the top recommendation is above a preconfigured confidence threshold.
Recommending and auto-populating field values can not only save time, but it can also improve accuracy and consistency.
Feature 5: Einstein Case Routing
Einstein Case Routing runs case assignment rules and Attribute Setup for Skills-Based Routing rules after Einstein Case Classification has automatically updated a field with a recommended value.
When Einstein Case Classification provides better accuracy and consistency when auto-populating field values, the result is enhanced routing by Einstein Case Routing.
Feature 6: Service Analytics
Although Service Analytics doesn’t have Einstein in its feature name, it’s undoubtedly from the Einstein portfolio. Service Analytics is a type of Einstein Analytics Template; you create analytics apps from templates that display dashboards on desktop and mobile devices. Service managers can view historical and trending KPI data to visualise their contact centre performance. Service agents can view case and customer data to help them make faster and more informed decisions when resolving inquiries.
Service Analytics comes with pre-built dashboards and datasets that can be modified to suit your specific needs. Service Managers access the apps from Analytics Studio, and service agents use sidebar dashboards embedded in Salesforce pages to obtain information about their cases.
Feature 7: Einstein Prediction Builder
With Einstein Prediction Builder, you can build custom prediction models without any coding. For example, you could create a model that predicts the likelihood that a customer will renew their support contract based on their interaction with your contact centre to date, their purchasing history and their attributes such as tenure as a customer, age and location.
Custom predictions assist service agents in providing a more personalised customer experience, and they help contact centres make more informed decisions.
Feature 8: Einstein Bots
Einstein Bots are virtual customer assistants that either resolve a customer inquiry without service agent involvement or assist by gathering up-front information before handing-off to an agent. Einstein Bots integrate with the Chat and Messaging (SMS, Facebook Messenger and WhatApps) channels and use natural language processing (NLP) to provide a more conversational experience for the customer.
By automating the resolution of repetitive customer inquiries, Einstein Bots increase contact centre efficiency and improve KPIs such as first response time, first contact resolution, average handling time and customer satisfaction. Service agents also have a chance to work on more challenging and rewarding cases, which results in higher agent retention. Furthermore, Einstein Bots can be available 24/7/365.
Feature 9: Einstein Vision
Use the Einstein Vision platform service to benefit from pre-built classifiers or create custom models to perform image-recognition tasks, such as counting the number of objects in an image, recognising a product in a picture or reading serial numbers in a photograph.
As they say, a picture paints a thousand words and adding imaging capabilities to customer support, especially in a field service setting, can provide a more productive experience.
Feature 10: Einstein Language
Einstein Language includes Einstein Intent and Einstein Sentiment. Use Einstein Intent to create custom models to understand the meaning in a block of unstructured text. Einstein Sentiment allows you to use pre-trained models, or to create custom models, that classify a body of unstructured text into positive, negative or neutral sentiment. Building Einstein Intent and Sentiment into custom applications enables you to analyse customer inquiries arriving through text service channels, or in social media posts.
Einstein Language allows you to understand which products or services customers are talking about, the meaning of what they are saying and how they feel about what they are stating.
Example Use Cases
Each contact centre faces challenges that are particular to them; hence, there’s an ecosystem of Salesforce partners available with the expertise to advise and implement tailored solutions to solve a business’s specific problems. That said, I thought it might be useful to illustrate how the Einstein features that I outlined above can be deployed individually, or in combination, to address six different problems. Hopefully, they will resonate with those familiar with contact centres, and the solutions will help to put the Einstein features into context.
Problem 1: Growth in Inquiry Volumes
How do we handle growing inquiry volumes across multiple channels, so that we can control costs while maintaining customer satisfaction?
Einstein for Salesforce Service Cloud’s raison d’etre is customer service automation, so it comes as no surprise that all of the Einstein features help contact centres to be more efficient and to absorb increasing inquiry volumes, without a proportional increase in costs. For example:
- Einstein Article Recommendations help agents to find Knowledge articles faster to answer a customer’s question.
- Einstein Next Best Action guides a service agent in what to do next rather than referring to a playbook or making it up as they go along.
- Einstein Case Classification and Einstein Case Routing get cases to the right agents faster.
- Service Analytics helps service managers identify areas for improvement and helps service agents close cases quicker by surfacing up relevant customer data.
- Einstein Prediction Builder, Einstein Vision and Einstein Language allow developers to build custom applications that solve specific problems using AI and so streamline operations.
- Einstein Bots add virtual, rather than real, agents to a contact centre to resolve some cases automatically or assist human agents in solving other cases.
It’s a bit optimistic to expect a contact centre leader to approve the rollout of all the Einstein features simultaneously. Instead, they will want to see a roadmap for a phased implementation where low complexity, high return problems are addressed first, with the delivery of incremental benefits throughout the remaining phases. Later in this post, I will provide an example of a maturity roadmap.
Problem 2: Agent Onboarding, Development and Retention
How do we onboard, develop and retain our service agents so that they are motivated, engaged and want to stay with us?
Attrition in contact centres is a notable cost that includes loss of capacity until a new agent is hired and reaches full productivity, recruitment costs and training costs. A factor that contributes to attrition is agent satisfaction, which can decline if an agent spends most of their time working on repetitive, simple to solve cases and if they struggle to clear their backlog of cases. Automation and agent assistance, covered in the inquiry volume problem, benefit the agent attrition problem in the following ways:
- Agents have more time to work on more challenging and exciting inquiries while burning through their backlog of repetitive inquires faster thanks to Einstein.
- New agents quickly become more productive as the recommendations that Einstein agent assistance makes embodies the knowhow of experienced agents.
Problem 3: Business Alignment
How do we ensure alignment with the rest of the business, so we deliver a consistent experience to our customers?
Key to cross-functional alignment when it comes to consistent customer experience, is everyone having the same view of the customer and everyone knowing how to interact with customers in a way that reinforces a company’s brand.
The Salesforce platforms certainly help to provide a single view of the customer; this unified view is improved further with the addition of the Customer 360 Data Manager.
Implementing an Employee Community, Knowledge and Chatter, all outside the scope of this post, are good examples of how Salesforce products can help keep employees aligned with the company’s DNA through self-service knowledge and collaboration.
A couple of Einstein features specific to Service Cloud that can also help ensure alignment that results in consistent customer experience are:
- Einstein Next Best Action guides service agents in their customer interactions so that all service agents apply the same actions as each other and as other business functions, such as marketing and sales.
- Einstein Bots should be designed to reflect your brand and will provide a dependable customer touchpoint with your marketing, sales and services functions.
Problem 4: Generating Revenue from Customer Service
How do we introduce upselling and cross-selling opportunities so that we contribute to our company’s revenue growth?
Businesses have traditionally classified their customer service function as a cost centre. However, a well-informed service agent is in an ideal position to make a commercial offer to a customer when working on their inquiry. A unified view of a customer’s attributes and their journey with your company to date helps to inform the service agent. Einstein Next Best Action arms the agent with an appropriate offer for the customer’s circumstances. For example, a casual inquiry asking when the next version of a mobile handset will be available can turn into a pre-order and a new contract commitment given the right incentive.
Problem 5: Reputational Brand Damage
How do we stay tuned in to what our customers are saying about our brand on social media so that we can address any negative sentiment?
Consumers readily share their experience of brands on social media, both good and bad. Unfortunately, bad news travels faster than good news as it gets amplified on social platforms with global reach. By monitoring social media, companies can acknowledge posts about their products and services and take remedial action if needed. Utilise the Einstein platform services to create a custom solution to this problem:
- Using Einstein Language, create a custom Einstein Intent model that can understand the meaning of a social post and take advantage of Einstein Sentiment to determine if the poster is being positive, negative or neutral.
- Einstein Vision can analyse any posted pictures to identify the product in question or any photos of serial numbers.
- Einstein Prediction Builder can be employed to predict the poster’s next most likely action to help prioritise and inform a response.
Add the resulting information to a case for a service agent to handle or automate the process using Lightning Flow.
Problem 6: Personalisation
How do we gain a better understanding of a customer’s journey with us so that we can provide them with a personalised and relevant experience?
I’ve already covered the importance of a unified view of a customer. Once you have a single view of a customer’s attributes and their interactions with your company, you can use Einstein Prediction Builder to model and predict a user’s interests, behaviours and possible actions. For example, you can predict the likelihood that a customer will:
- Buy a specific offering
- Use a particular service channel
- Escalate a case
- Share their experience on social media
- Increase their NPS following specific actions
There are likely many more problems that we could address, but hopefully, you get the picture.
With so many potential problems and the range of Einstein features, where do you start? That’s the subject of the next section.
Example Einstein Maturity Roadmap
Although it’s impossible to provide a single “silver bullet” Einstein roadmap, I’ll outline an example to seed some thoughts. It starts with lower complexity features that deliver some immediate value, followed by increasingly sophisticated features that provide incremental value. This approach allows you to reap benefits and in parallel, become comfortable with the technology and develop an understanding of how it can be applied to solve your specific problems. Each feature ought to have a roadmap that increases the range of inquiries that it can automate or for which it can offer agent assistance; however, in the interests of brevity, I’ve omitted these.
Start with the Einstein features bundled with the Service Cloud license, namely Einstein Article Recommendations. Of course, this assumes that you are already using Salesforce Knowledge to help resolve customer inquiries.
If you have rolled out the Chat or Messaging service channels, you have already purchased the Digital Engagement add-on SKU, which includes 25 Einstein Bot conversations per user per month that you might not be using. Implement a menu-driven transactional Einstein Bot to answer the top three or four frequently asked questions (FAQs).
Finally, if you are using the Chat service channel, implement Einstein Reply Recommendations. Note that the licensing model or pricing is not yet known, which might change this recommendation.
Start to leverage Einstein Bot’s AI capabilities to provide a conversational experience for customers. Move beyond FAQs and start to implement actions; for example, rather than supplying a Knowledge article link in response to the question “How do I cancel my order?”, actually cancel the order, which probably means integrating with other systems.
Buy the Service Cloud Einstein add-on SKU, which includes Einstein Next Best Action, Einstein Case Classification, Einstein Case Routing and Service Analytics. Use Service Analytics to a) get a view of your historical and trending contact centre KPIs and b) give service agents a comprehensive picture of customers and cases. Implement Einstein Case Classification and Einstein Case Routing to recommend field values to agents, auto-populate high confidence field values and route those cases. Start developing Einstein Next Best Action strategies and recommendations to guide agents both in terms of actions to perform and offers to make to customers.
At this point, you should be sweating your Einstein app features and your Einstein Bots platform feature. You’ve probably also dabbled with the remaining platform services, namely Einstein Prediction Builder, Einstein Vision and Einstein Language. It’s now time to incorporate Einstein AI in custom apps to solve your specific business problems and to differentiate your customer service from everyone else. You can also build custom prediction models using Einstein Prediction Builder to enhance your Einstein Next Best Action decision making. The Salesforce innovation machine never stops, so there will probably be a bunch of new features and existing feature enhancements to use!
The portfolio of Einstein features that can be used with Service Cloud to enhance customer service has grown substantially over the past few years. This post, and the accompanying cheat sheet, summarises those features and provides example use cases where they help solve common contact centre problems. Hopefully this will help you to get started on a journey to Einstein maturity, boosting your contact centre efficiency and increasing customer and service agent satisfaction. If you have any suggested use cases that you would like to share with me and other readers, please add them as a comment to this post.