Agentforce was first introduced in Salesforce 2024 Dreamforce on September 17-19 in San Francisco, California. Within just 3 months since it was first introduced, there have been so many discussions on different platforms regarding this new feature. A lot of Salesforce partners have raced to acquire Salesforce AI Certificate to understand better Agentforce, and a lot of Salesforce clients also showed interest in this new features. This article will summarize the most recent understandings of Agentforce to help anyone who just started their research save time and effort googling around and watching countless of discussions.
There are many right and wrong assumptions about Agentforce because of the lack of information we can get about this feature. As an Integration Architect, I have joined multiple recent Webinars organized by Salesforce discussing this topic and understand much better what Agentforce is and what its capabilities are. Therefore, I will be able to clear some doubts and reconfirm some concepts.
What is Agentforce?
From the perspective of a new Salesforce user
For a new Salesforce user, Agentforce is a proactive, autonomous application that natively integrates with the entire Salesforce Customer 360, enterprise data including Sales, Service, Commerce and Marketing, and uses integrated context to take actions, and recommendations based on the knowledge base, and allowed action list.
In order to integrate with Customer 360, enterprise data, Agentforce grounds prompt with customer data, only returns an answer based on enterprise data. Agentforce is also aware of the user’s permission and only retrieves data and implements actions aligned with the user’s permission. These features ensure data consistency and avoid security breaches.
From the perspective of Salesforce users
On the other hand, for a Salesforce user who has been with Salesforce for a long time, and already has experience with Einstein Copilot, Agentforce is simply a new version of Einstein Copilot with enhanced LLM, added features improving security and integration to expose the agent to end users via omnichannel. However, because what Salesforce showed in the Dreamforce 2024 about Agentforce is not available in the market yet, many people also say that Agentforce is just Einstein Copilot rebranded to align with the AI Agent trend.
Why I say Agentforce is the same as Einstein Copilot? It is because the features and main components of Agentforce were already available with Einstein Copilot including Agent Builder, Topics, and Actions along with Flow Builder, Prompt Builder and Prompt Template. However, there was a shift in focus to emphasise the security of the AI Agent as Salesforce highlighted the Einstein Trust Layer.
Based on what Salesforce presented on Dreamforce 2024, we can expect further features will be GA to differentiate Agentforce from its earlier version, Einstein Copilot. I believe these include the ability to create multiple agents and expose agents to end users via omnichannel.
The Benefits of Agentforce
Boost Agent Productivity and Efficiency:
- AI-Powered Triage: Agentforce leverages AI to intelligently route customer inquiries. Simple issues, such as password resets or basic order tracking, can be automatically resolved by chatbots or AI assistants. This frees up agents to focus on complex, high-value interactions that require human expertise.
- Knowledge Base Integration: Agents have instant access to a comprehensive knowledge base of articles, FAQs, and best practices. This empowers them to quickly find the information they need to resolve customer issues efficiently and accurately.
- Data-Rich Workspace: Agentforce provides agents with a unified view of customer data, including past interactions, purchase history, and preferences. This context allows for personalized and empathetic service, building stronger customer relationships.
- Automation of Repetitive Tasks: Automating tasks like data entry and status updates significantly reduces manual effort, allowing agents to spend more time on meaningful customer interactions.
Enhance Customer Experience:
- Faster Response Times: AI-powered routing and automated solutions ensure customers receive prompt responses to their inquiries, minimizing wait times and frustration.
- Personalized Interactions: By leveraging customer data, agents can deliver personalized service that addresses individual needs and preferences. This builds trust and loyalty.
- Proactive Support: Agentforce can proactively identify and address potential customer issues before they escalate. This proactive approach demonstrates a commitment to customer satisfaction.
- Consistent Service Quality: Agentforce provides a framework for consistent service delivery across all channels, ensuring a positive customer experience regardless of how they choose to interact.
Drive Data-Driven Decision Making:
- Real-time Insights: Agentforce provides real-time dashboards and reports that offer valuable insights into customer behavior, agent performance, and overall service trends.
- Identify Service Bottlenecks: Analyze data to pinpoint areas of inefficiency in the support process and identify opportunities for improvement.
- Measure Key Performance Indicators (KPIs): Track key metrics such as customer satisfaction, resolution times, and agent productivity to measure the effectiveness of support operations.
- Improve Service Strategies: Use data to refine service strategies, optimize resource allocation, and make informed decisions about future investments in customer support.
Reduce Operational Costs:
- Increased Agent Efficiency: By automating tasks and streamlining workflows, agents can handle a higher volume of inquiries with the same level of staffing.
- Improved First Contact Resolution: Faster resolution times and reduced call transfers lead to lower overall support costs.
- Reduced Customer Churn: Improved customer satisfaction and loyalty can lead to increased customer retention and reduced churn rates.
- Streamlined Operations: Agentforce can help streamline and automate many aspects of customer support operations, leading to significant cost savings.
Common Concerns About Agentforce
As I have joined multiple in-person meetings as well as webinars organized by Salesforce, there are few common concerns raised by attendants who are Architects, CEO, Founders from different companies.
Security
- Can users retrieve data from other users when interacting with agent? Salesforce enforces the same permission sets to the user when the user interacts with Agentforce, thus if the user doesn’t have permission to retrieve other users’ data, Agentforce will not allow that action to be done.
- Will the agent provide a hallucinated answer if it can’t find information from the provided data? Salesforce has emphasized the importance of the Einstein Trust Layer in handling data into and from the LLM to ensure no hallucination, no toxicity and zero data retention.
- How can we know if agents use correct actions and provide correct answers to the user in production? In order for LLM to pick up correct Actions, it’s important to provide precise descriptions for the Topics and Actions including what they are about, when to use and what are input and output. The more accurate the description is, the better LLM can precisely pick up actions based on users’ requests.
Data Management
- In order for Agentforce to work effectively, do we require to have all Organisation data migrated into Salesforce? The more data Agentforce can have access to the better accuracy LLM can provide answers to the user. The best way to do this is to make all data available in Salesforce within Data Cloud, so Agentforce can have quick access to Customer 360, Organisation Data. However, it is not compulsory to have all data within Salesforce. Suppose your organisation is using or is looking into Data Cloud. In that case, Salesforce provides a feature called Zero Copy approach, so while being able to avoid physical data movement, you can still get all the benefits of CDP like data harmonization, identity management and analytics. On the other hand, if your organisation is not using Data Cloud, you can always retrieve data from other system via API calls by using Flows. This requires activating a Flow to call external API using APEX or MuleSoft API.
Operation Cost
Operation Cost is always one of the organisation’s main concerns. According to Salesforce, the cost for one conversation with Agentforce is $2. If considering that is the cost of interacting with Agentforce regardless of channel, length of the conversation and the outcome, I do believe it is expensive. However, it is not that simple to claim because we need to consider all of the factors including hiring costs for agents, training cost and whether or not they can handle the tasks with high accuracy and efficiency as Agentforce can achieve. Additionally, it is still unclear how Salesforce would like to charge in different scenarios where users interact with Agentforce.
Therefore, for any one who wants to check if Agentforce fits with their business, they need to do a detailed cost analysis with their specific use case and reach out to Salesforce for the precise pricing details.
Conclusion
Agentforce represents a significant advancement in Salesforce’s suite of tools, offering robust capabilities to enhance agent productivity, improve customer experience, and streamline operations. While it builds on the foundation laid by Einstein Copilot, its enhanced features, such as the Einstein Trust Layer and seamless omnichannel integration, set it apart as a proactive AI solution for modern enterprises.↳
However, as with any emerging technology, understanding its nuances is essential to maximizing its potential. By addressing common concerns such as security, data management, and operational costs, businesses can make informed decisions about adopting Agentforce in their workflows.↳
For organizations exploring this powerful tool, a thoughtful approach that combines in-depth research, strategic planning, and cost-benefit analysis will be key to unlocking its full potential. Agentforce has the promise to redefine customer service—bringing efficiency, personalization, and innovation to the forefront.