At the recent Dreamforce 2024 event, Salesforce CEO Marc Benioff focused on introducing Agentforce, a suite of autonomous AI agents designed to carry out data-driven actions across crucial business functions. Benioff emphasized the collaborative future of humans and AI, stating, “This is about humans and robots driving customer success together,” reinforcing the vision of AI augmenting rather than replacing human capabilities.
That’s a lofty vision, but my attention of course turned immediately to the implications of Agentforce for handling companies’ proprietary data, and especially how it will integrate with and impact ERP and SCM systems. Let’s dig into the impact Agentforce could have on these systems and how it might influence enterprise operations.
What Agentforce Does
Agentforce is part of Salesforce’s “third wave” of AI, in which autonomous AI agents perform tasks without human intervention. The platform also offers low-code development tools, enabling business users to quickly build and customize AI agents that integrate with existing Salesforce workflows and data. At the core of Agentforce is the Atlas Reasoning Engine, which autonomously evaluates and processes data, executes decisions and completes tasks. Atlas can learn from human feedback through reinforcement learning; as users interact with the system and provide input, Atlas can improve to become more effective.
Agentforce is meant to extend Salesforce’s capabilities in a significant way, taking advantage of data flow from various enterprise systems including ERP and SCM to fundamentally change business workflows.
The Devil Is In The Data
In my articles over the past year, I’ve delved into how AI success isn’t just about having data—it’s about having the right data. A well-defined approach to managing data is crucial for enterprises to truly harness AI. Agentforce relies on Salesforce’s Data Cloud, an open platform that unifies data from internal and external sources, including ERP and SCM systems. Data Cloud imports structured and unstructured data from various sources through connectors, import tools, zero-copy integrations, SDKs, APIs and ETL jobs.
Using MuleSoft’s pre-built connectors for enterprise systems such as SAP, Oracle and Microsoft Dynamics, the platform synchronizes data in real time, keeping inventory levels, order statuses and financial data up to date. This type of integration supports automated workflows, reducing manual data entry and minimizing errors. In the bigger picture, having centralized data in the cloud enhances analytics, forecasting and supply chain optimization.
Drawing on Data Cloud allows the AI agents created in Agentforce to make more informed decisions, ensuring that AI-driven processes are accurate and adaptable to current business conditions. This enables Agentforce to deliver relevant responses and perform tasks appropriately across sales, service, marketing and commerce functions. Salesforce says that companies such as OpenTable, Wiley and Wyndham Resorts have already improved their operational efficiency by using Agentforce.
How Agentforce Works With ERP And SCM Systems
Agentforce leverages this integrated data to enhance customer interactions and operational efficiency. For example, when an AI-driven agent needs information about a customer order, Salesforce, through MuleSoft’s connectors, provides direct access to ERP system data. This real-time access ensures that Agentforce AI agents have a comprehensive and current view of customer data, enabling them to respond quickly and accurately to inquiries. As part of this process, the Atlas Reasoning Engine autonomously analyzes this data to make informed decisions, ensuring reliable and accurate results.
One aspect I find particularly valuable is how integrating Agentforce with ERP and SCM systems can remove routine tasks, allowing these systems to work more efficiently. This allows human workers to focus on complex, strategic tasks that impact the business meaningfully. The key to getting the most out of these systems is finding the right balance between automation and human expertise so that technology supports and enhances human decisions rather than replacing them.
Benefits And Challenges For Enterprises Using Agentforce
Let’s take a closer look at the impact of integrating Agentforce with ERP and SCM systems, starting with the benefits.
Automation — AI agents can take over repetitive tasks in areas such as order processing, inventory management, customer service and procurement processes. For example, in a manufacturing setting, Agentforce can monitor raw material levels in real time and automatically reorder supplies when they reach a critical threshold, ensuring that production lines are not disrupted. By integrating with ERP systems, agents can also analyze production data to identify inefficiencies, allowing manufacturers to optimize their operations and reduce waste.
Data integration — A key advantage of integrating data between systems is the enhanced efficiency it creates. In healthcare, for example, linking ERP data with patient management systems can improve scheduling and optimize the use of resources. In the retail sector, integrating CRM data such as customer purchase history and preferences with ERP systems allows for real-time inventory updates. This ensures that popular products are restocked, reducing stockout risk and improving customer satisfaction.
Scalability — Agentforce can handle more work in ERP and SCM systems without requiring additional staff. In logistics and transportation, for example, AI agents can manage increased order volume during peak seasons by processing larger amounts of shipment and delivery data, including tracking numbers, delivery schedules and inventory levels. This enables the system to dynamically handle more shipments and deliveries without overburdening the existing workforce.
Predictive analytics — Agentforce can enhance decision making by offering valuable insights. For instance, Agentforce can predict demand and optimize supply chain operations in the agriculture industry by analyzing ERP and SCM data on crop yields, inventory levels, distribution schedules and external factors such as weather patterns. This integration allows for more accurate planning, helping to improve productivity and resource management and minimizing disruptions.
Integrating AI agents like those in Agentforce with ERP and SCM systems comes with challenges, too, including managing complexity, ensuring security and building trust in—and transparency about—how the AI operates.
Complexity — Enterprises need to integrate AI agents into existing ERP and SCM workflows without causing interruptions. This often requires IT expertise because errors during the process can lead to operational delays or data inconsistencies. If an AI agent is not properly synced with the ERP and SCM systems, it could result in problems such as inaccurate inventory data, leading to production slowdowns or even halted operations.
My advice: Enterprises need to carefully integrate AI agents into ERP and SCM workflows to avoid disruptions. A clear integration plan, agreed upon by both the customer and Salesforce, can help ensure that AI agents sync properly, preventing production delays or other issues.
Security — Agentforce manages sensitive information, making security a key area of focus. Moving on-premises data such as customer records, financial transactions, supplier contracts or patient information into the Data Cloud can introduce security risks for enterprises, especially in tightly regulated industries. Protecting this critical information—whatever its location—deserves close attention from enterprise cybersecurity teams.
Although Salesforce historically has an exemplary record for security, it pays to be cautious. In July 2024, Disney began moving away from Slack after a significant data breach exposed more than a terabyte of internal information, including messages, spreadsheets, PDFs and personal details of employees and customers. The ongoing investigation suggests that both Disney and Salesforce share responsibility for the breach. This event highlights the critical need for effective security practices for cloud-based systems to protect sensitive data.
Trust and transparency — ERP and SCM users need to be able to trust that AI agents handle data securely and accurately. Some of the potential problems on the ERP side have already been suggested; in SCM, any mistakes in AI-driven decisions could disrupt the supply chain and cause financial losses. Transparency in how AI makes decisions is key for building confidence. In this context, it will be helpful for Salesforce to provide clear explanations of how its AI tools carry out tasks such as demand forecasting so that businesses can trust the AI as a reliable partner rather than a “black box.”
Updates To Data Cloud And MuleSoft
Not every announcement at Dreamforce was about Agentforce. I particularly want to mention the new features in Data Cloud. For starters, it now has better support for unstructured data. This is important because it means that all parts of the Salesforce platform can better leverage unstructured files of text, audio, video and more to generate deeper customer insights. Improved search capabilities now factor in customer context, making interactions more relevant. For instance, a customer support rep could quickly find past interactions and specific customer details, leading to faster problem resolution. Real-time data updates ensure timely responses to any changes in customer information. Enhanced security measures, such as encryption and access controls, protect data during transactions, reducing the risk of unauthorized access.
The MuleSoft team within Salesforce also introduced new tools that enable developers and admins to connect structured and unstructured data from CRM and external systems. These enhanced integration features allow AI agents to retrieve and act on data from multiple platforms in real time. For example, an AI agent could automatically pull sales data from an ERP system along with customer interaction history from the CRM, providing a support team with a complete view of a customer’s situation without manual data entry.
What Agentforce Means For Salesforce
Salesforce’s introduction of Agentforce marks a notable advance in how the company integrates AI with enterprise systems. The combination of Data Cloud and the Atlas Reasoning Engine shows Salesforce’s progress in creating a flexible AI platform—one that goes beyond the Einstein platform improvements announced at Dreamforce 2023. However, the effectiveness of Agentforce will depend on finding the right balance between AI and human involvement, particularly in the use of enterprise systems.
My decades of implementing ERP systems immediately suggests the example of maintaining data quality, which in my experience requires at least some degree of human oversight. In this example and others, AI is great at handling routine tasks, and automation can certainly improve efficiency, but human input is still essential for managing the complexities and nuances within ERP and SCM systems. Overreliance on automation—even with the cleverest AI agents—could lead to missed details that only human insight can catch.
Salesforce faces the challenge of ensuring that automation works alongside human interaction rather than trying to replace it. Beyond that, Agentforce competes with established AI solutions from Microsoft, Google, Oracle, AWS and Zoho, which offer their own automation and customer service tools. So although Salesforce has a commanding position in the CRM market, it will have to rise above the noise in a very hectic AI market—and in a season when we’re going to be hearing far, far more about AI agents from many different vendors.
