Admins / Architects / Artificial Intelligence

Why Do You Need AI features in Your Salesforce Org?

By Lars Malmqvist

AI features have become imperative to the Salesforce platform. We now have Sales Cloud Einstein, Service Cloud Einstein, Marketing Cloud Einstein, Commerce Cloud Einstein, and Einstein across the industry clouds – and that’s just the starting point. If you look at all the AI capabilities across the Salesforce core clouds and additional products such as Tableau and Slack, it is clear that having AI embedded deep in the platform is a key strategic driver for Salesforce’s future growth.

Yet, when I speak to other architects, the real-world adoption of AI features often lags behind the capabilities the platform has to offer, which is a real shame. The truth is that you can adopt robust AI features right now, which will have a significant positive impact for your users.

Assuming that a good example is often the most persuasive way to make a point, let’s have a look at three ways you can leverage AI features today: Personalization, Automation, and Augmentation.

We’ll begin by understanding how intelligent data enables you to overcome some of the key challenges that traditional CRMs face today.

The Value of Intelligent Data

We are living through a data explosion! Let’s take a look at this phenomenon – every day, companies generate two petabytes (that’s two billion gigabytes) of data. That is a big number, but it is dwarfed by the amount of petabytes generated by connected devices, which can generate five petabytes every single day. Both of these numbers are constantly growing.

In 1998, at the start of my career, a client hired us to regularly download a subset of the internet for internal browsing on a CD-ROM so they didn’t have to connect sensitive computers to the internet. That was back in the 90s, and at the time you could fit a good selection of websites on a CD-ROM for this to be viable. However, if you tried to download the internet today on consumer broadband, it would take about 181 million years. Of course, it might have grown even more by then.

CRMs are key to a successful relationship between companies and their customers – they break down customers’ behavior at an ever more granular level across multiple channels and data sources. Salesforce is an advanced CRM that can manage a large range of scenarios smoothly with its standard interface. However, there is a practical limit to how much data you can meaningfully show in a set of related lists while expecting a customer service agent to work effectively.

Customers expect more and better interactions with your company, and taking these interactions into account is important if you want to deliver a great experience. Internally, there is a constant drive for better, faster, cheaper processes for customer service. As a technical person, this can make you feel like you’re ‘caught between a rock and a hard place’, especially when you have to compromise in ways that don’t fully deliver what your users are looking for.

While uncomfortable trade-offs are part of an architect’s job, AI (when combined with a customer-centric platform like Salesforce) provides a way to meet this challenge effectively.

In the following sections, we will look at some of the ways this can be done using features that are available today.

Driving Personalization with AI

Data-driven, individually personalized customer journeys are the ‘holy grail’ for some users. By combining data intelligently and putting it in the context of a consumer, you will be able to show more relevant content that will drive engagement, loyalty, and ultimately more sales. AI is particularly well suited to assist in this task, particularly in the form of machine learning models that crunch large amounts of data about consumers to figure out patterns about their preferences and behaviors.

Marketing Cloud is particularly strong among the Salesforce products. For instance, you may use Einstein Engagement Scoring to assign your subscribers into particular persona groups that reflect their likelihood to engage with your brand. Using Journey Builder, you can personalize the messages your customers will receive based on factors like their level of engagement and current stage in the Buyer’s Journey. Add to this Einstein Engagement Frequency to know how often to send them messages, and Einstein Send-Time Optimization to know what time of day to send them for maximal engagement. Finally, add Einstein Recommendations and Einstein Content Selection to personalize the content and offers shown to your subscribers based on their customer profile information.

In Commerce Cloud, the most heavyweight feature is found in the product recommendation engine. There is a wealth of customizable product recommendations strategies to choose from that take into account both the real-time behavior of the customer and their historical affinity to products and categories. These strategies can be effectively combined with classic options such as “Customers who bought this product also bought.”

In Service Cloud, a great option for personalized service can be created using Einstein Next Best Action. Using an embedded Lightning component, the feature suggests what actions to take in the context of a Salesforce record. The suggestions can use pre-defined strategies that you define in a Flow-builder style interface. However, the real power comes when you use machine learning models to generate options for actions that can consider the personal information of callers to make them particularly relevant. An example could be offering personalized vouchers or a special upgrade at the exact time it is most likely to have effect.

READ MORE: 10 Salesforce Service Cloud Einstein Features & Use Cases

Building Intelligent Process Automation

Automation is often the first factor people consider when looking for ways AI can help achieve their business goals. In some cases, it can lead to unrealistic expectations about the level of automation that is feasible. However, it is clear that you can derive large benefits by automating routine process steps and allowing users to focus on the elements that require genuine human judgment.

In this area, Salesforce provides a lot of tooling, although there are fewer pre-built features than for personalization. We can assume that process variability is too high in most common processes for pre-built automations to be viable at this time. One notable exception is Einstein Case Classification and Routing, which can automate several routine steps in a customer service case, handlingFlow using a pre-built machine learning model, allowing for both partial and full automation of the case classification process.

When I first heard about Einstein Automate, a suite of integration and low-code process definition tools, I wondered about the name. I didn’t see any obvious link to the AI features of the Einstein platform. However, when you consider it in the context of intelligent process automation, the naming becomes obvious. In order to effectively automate process steps, you need AI features combined with tools that orchestrate and integrate processes.

You do this by deploying machine learning models on or off-platform and incorporating them in the kind of automation features offered by Einstein Automate. Salesforce offers two declarative options for building machine learning models.

The first is Einstein Prediction Builder, a simple tool that can be configured by your administrator, but it is also limited in its scope because it only considers data from a single object. If you have lots of repetition in your processes, this may be a quick fix. Einstein Discovery builds on top of the full power of Tableau CRM. Although it has much more power and complexity, it manages to be quite straightforward to configure in many cases. You can also use third-party options via integrations, or even build your own model from scratch on a platform like Amazon Web Services (AWS). The tooling offered by Salesforce will accommodate all of these options.

Augmenting Sales and Service Staff

Just like automation, augmentation focuses on efficiency. However, it is efficiency of a different kind. Instead of trying to replace manual process steps with automated ones, the goal is to provide insight and assistance to human workers, allowing them to focus on their highest value activities and be more proactive

Embedding intelligent analytics that can guide the best course of action within the context of customer data is a fairly obvious way to do this, and Salesforce provides a wealth of pre-built features. For instance, Einstein Lead and Opportunity Scoring give you a quantitative score based on historical data, indicating how likely your Lead is to convert or your Opportunity to be won. This enables your sales team to prioritize the right Leads and Opportunities.

You can create a similar feature yourself for other objects using Einstein Prediction Builder. For general insights, Einstein Account Insights provide prioritized information from news sources about features and updates that may be affecting your account.

You can increase the usefulness by creating your own analytical models in Tableau CRM and deploying them either as insights or actionable predictions with Einstein Discovery. Salesforce is using this pattern heavily in their industry clouds with pre-built Einstein Discovery stories for things like predicting customer churn, individual healthcare needs, or likelihood to attend a given appointment. These are all areas, where the information provided in the prediction can help human workers better target their efforts to achieve maximum results.

Finally, it’s worth considering Einstein Bots. Bots are used to automate routine transactions, and can also be very useful when determining where to route a customer or how to direct experts within your organization to the best starting point when talking to customers. For example, a bot can collect basic household information before redirecting to an insurance advisor, so they have the right information going into the conversation, and can use their time more effectively.


Overall, there are plenty of AI features available on the Salesforce platform today that can deliver fast benefits to your users and customers. Picking a few quick-win use cases is a good way to get started with AI adoption in Salesforce, and I hope I’ve persuaded you to start looking for some to try out. You can design powerful and accurate AI-driven, state-of-the-art solutions that are tailor-made for modern business demands with my book: Architecting AI Solutions on Salesforce.

The Author

Lars Malmqvist

Lars is a Partner at Implement Consulting Group, a management consultancy. He is a Salesforce CTA and the author of Architecting AI Solutions on Salesforce and Salesforce Anti-Patterns.


    March 16, 2022 9:29 am
    Very informative article!

Leave a Reply