Agentforce and Experience Cloud: How to Leverage AI to Improve Customer Service

If you work with Salesforce or are considering Salesforce for your organization, you’ve surely heard about Salesforce’s crown jewel, Agentforce.

While Salesforce has been offering AI and Machine Learning (ML) products in the Einstein family for a few years now, Agentforce allows for the creation of your very own “agentic” AI. The “agents” you can build with Agentforce are capable of using your CRM and external data to perform or accelerate complex tasks that used to require more human intervention.  

Salesforce often showcases AI as tooling that helps internal users speed up tasks, find deeper insight, or remove administrative and research burdens so users can focus their time on higher-value activities. But AI is also incredible for world-class customer care and community management work, most notably within Salesforce’s native Experience Cloud tool.

If you’re an Experience Cloud user already or are thinking about setting up an Experience Cloud implementation, here are some ways you can use Salesforce’s AI tools to curate a killer user experience.

Standard SFDC Personalization and AI tooling

Just because AI is the way forward, it doesn’t mean Salesforce will be scrapping all the personalization tools it has spent years building. In fact, you should start looking to exhaust many of the traditional personalization mechanisms Salesforce offers before you implement any AI into your Experience Cloud site.

Personalization is always going to be easiest on an authenticated Experience Cloud Site as users have to log in. 

For a user to be able to log in to Experience Cloud, a user must have a sympathetic contact record tied to its Experience Cloud user record, which means that your Salesforce CRM has a data infrastructure already associated with your user that it can draw upon to create personalization.

Tailoring your Experience Cloud site should start with segmentation from a place you have already created to collect data. For example, let’s say your organization sells to both resellers who sell on your behalf and consumers who have product questions and service requests, so you decide to create two separate sites curating independent experiences for each category of clientele. 

Let me expand on that. Maybe you have a tiered selling program with your resellers, where resellers have access to different collateral, different co-op marketing, or different records depending on the sales they make on your behalf. Perhaps certain people who work for that partner reseller should be able to see invoices, but others shouldn’t. 

Thinking about the other site, maybe your consumers are only allowed to create service requests based on products they have already purchased from you or need support in a language different from the one your organization uses internally.

All of these personalization examples can be accomplished with the use of standard Salesforce security mechanisms like role hierarchies, record types, permissions, field-level security, and filtering components on field or object-level criteria. 

You don’t need AI to accomplish these fundamental things if you’ve already bought the appropriate licenses from Salesforce.  

While you can build your own custom LWC (Lightning Web Components) and LWR (Lightning Web Runtime) sites to accomplish the segmentation examples suggested above, most of this personalization will be point-and-click if you use the standard Experience Cloud templates designed for these specific use cases. For the two examples above, Partner Central and Customer Account Portal or Customer Service would be templates to consider.

Once you’ve properly set up your standard Salesforce personalization mechanisms and collected enough data, you’re ready to start implementing AI tools. When beginning your AI journey, work in small chunks and monitor your rollout heavily. 

Pick one AI tool or one relatively simple bot and iterate from there. Given the multiple use cases for AI in Customer Service, service workflows are a great place to start. 

Agentforce for Service Sites

The capabilities of Agentforce as they stand today have far and away the strongest customer service use cases, making them an obvious choice to host on an Experience Cloud site where your customers request service. 

While you can host Salesforce agents across any of the Experience Cloud sites and even other websites, customer account portal and customer service will be the lowest-code template options to quickly spin up a site and implement AI tooling. 

Let’s look at some of Salesforce’s AI products and how they can be used in a service-oriented Experience Cloud Site.

Agentforce Service Agent

Agentforce Service Agents are the big kahuna of all Salesforce AI products, and your clients will see them prominently. 

Salesforce Service Agents can be built by your Salesforce administrators and developers to process incoming cases, handle questions while picking up on linguistic subtleties, guide conversations, query complex data, and even triage to live support where necessary. 

Service Agents can act on a customer’s behalf much like a human representative would, taking action on what your end user asks it to while maintaining privacy and sharing settings in Salesforce.

Agents can heavily relieve the burden of internal support workers, while also gathering relevant information to improve and iterate on other support functions. Think about placing an Agentforce Agent on the parts of your Experience Cloud site where people go to collect information – if you have a high volume case-entry page or even a custom page with organizational notifications, an agent can help answer a user question without having to resort to human intervention.

Note: Agentforce agents and Einstein Chatbots are different products that operate at a different cost structure, which we will explore in the upcoming section, Chatbots and OmniChannel Integration. 

Einstein Service Replies and Recommendations

Service reps frequently have to handle the same questions over and over again, and organizations often want to ensure a consistent brand voice, even with generative responses. 

While reply recommendations have been a standard feature in many chat services for several years now, Einstein Service Replies and Recommendations elevate the chat and email experience by reviewing existing knowledge articles and incorporating conversational context into generative AI replies.

Einstein Service Replies and recommendations give a much more personalized feel to your Experience Cloud users as they engage with your organization’s online chat, while also creating a reply recommendation engine in the background from chat transcripts for your organization to review and expand upon for the next customer engagement. 

Einstein Case Classifications, Autofill, and Wrap-Up

Incomplete data is the bane of many users’ existence across Salesforce, especially when data already exists elsewhere in records relating to the one you’re currently working on. 

However, missing data from cases and other support-related handoffs can create significant delays in the client experience. Ensuring that a case record is thoroughly completed is tantamount to an excellent user session in Salesforce Experience Cloud.

Einstein Case Classification applications like Autofill and Wrap-Up fill in missing or derived information for support agents through components in the Service Cloud apps, allowing agents to more efficiently complete cases. 

While Experience Cloud users don’t necessarily see these products in Experience Cloud itself, they can still benefit from better support infrastructure that empowers the people helping them to focus on their needs instead of data entry. 

Agentforce Service Planners 

Service Planner is an incredible tool for organizations with complex service agreements that include things like warranties, SLAs, and complex transactions. 

Service Planners are generative AI that reviews a case and creates a step-by-step guide on how to resolve a case. While the user interface for Service Agents is currently available only in Service Cloud, implementation of it at an organizational level can still improve your Experience Cloud members’ user experience through better overall support. 

Einstein Generated Summaries

Ensuring you provide your end users with the best customer service means follow-up – not just lip service – while they’re on one of your service channels with you. Even incredible service representatives experience challenges when trying to fully annotate a conversation, but Einstein-generated summaries can do this for you. 

The Summarize Record action can summarize voice records, chat records and even messaging records to cleanly document what happened in a support interaction on a case record.

While your end user might not see these summaries unless you enable them to be visible inside your Experience Cloud site, Einstein Work Summaries ensure all aspects of a service encounter are accurately documented within the case and improve the quality of service your Experience Cloud user receives. 

Leveraging AI for Knowledge Base

One of the most underrated aspects of Salesforce is the ability to create an immense catalog of frequently asked questions, help articles, product-how-to’s, and general information in what is known as the knowledge base (KB). 

Salesforce has been adding capabilities to what Salesforce admins lovingly refer to as “The KB” over its lifetime, but its integration into AI is far and away its best application for a better Experience Cloud site.

For organizations looking to empower their users with articles relevant to a particular case or inquiry, check out Einstein Article Recommendations. Einstein Article Recommendations can be configured to surface suggested KB articles relevant to a conversation based on data it gleans from a case record. 

It can listen for terms it finds frequently in an article, review how often an article is voted down for being unhelpful, or even match long chunks of text to the article content itself to provide better suggestions.  

But AI for the knowledge base has gone a step further with Einstein Knowledge Creation – a feature that allows knowledge articles to be generated from a conversational transcript.

Einstein Knowledge Creation eliminates the need to write help articles resulting from a service interaction, allowing generative AI to create a Knowledge Base article by summarizing a service interaction. 

Chatbots and Omni-Channel Integration 

While Agentforce agents are likely the future of the current Salesforce Einstein Chatbot and Omni-Channel setup, Einstein Bots are still classified as a separate product from Agentforce and still have their place in enhancing the Experience Cloud UX. 

Think of Einstein Bots as a simpler version of an agent – while they can leverage similar kinds of Natural Language Processing and Understanding (NLP and NLU), chatbots don’t have the same kind of complex reach an agent would.

Einstein Bots are better for simpler, more rote tasks with repetitive outcomes. Think about your consumers checking in on order status in a chat interface. Surfacing a menu of recent purchases in a chat that can produce tracking numbers is an easy and obvious use case for an Einstein Chatbot. 

Both Einstein Bots and agents can leverage the wide variety of Omni-Channel features and can handle complex routing to queues, skill-based service agents, and transfer to other channels.

Agentforce for Partner Portals and Sales Sites

Even though the bulk of Salesforce’s AI offerings currently come in the form of Service Agents, Salesforce has released a few Sales AI tools that can greatly assist sellers with the Partner Portal or Experience Cloud sites focused on Sales.

An AI Sales agent like Agentforce SDR works well for top-of-funnel sales activities and can assist partner users with activities like lead qualification, lead nurturing, scoring, and routing. Going beyond just assisting your partner sellers in prioritizing the leads they’re receiving, an Agentforce SDR can even be configured to reply to leads, follow up, and connect leads with sales reps. 

However, be advised that some of the UI components from Salesforce’s Agentforce for Sales are only natively available within Sales Cloud.

Final Thoughts

The suggestions for AI implementation offered in this article are barely scratching the surface of what this technology can do, but it’s always best to start small and iterate into complexity after your organization adjusts to the use of AI. Agentforce has a long way to go and will undoubtedly continue to expand well beyond what we know it can do at the time of this article’s publication. 

If you’re an organization that has already begun its AI journey outside of Salesforce, you can also implement “bring your own” large language models into this technology, allowing for an expansive AI experience that transcends different applications.

A final word: Please use AI responsibly, especially in instances that are people-facing. It is still a nascent technology deserving caution and governance, and you owe it to yourself to maintain your understanding while expertise in the field grows. 

Don’t be reckless, don’t use it to do sketchy things, don’t let it run without monitoring mechanisms – and especially – don’t release it to your colleagues or client without sufficient notice and transparency.  

Implement a strong risk analysis framework before releasing AI into the wild, and gather feedback about its accuracy and utility. AI should be designed to augment humans and not harm them.