Innovate Your Business
with Our Insights!

Speaking during Salesforce’s latest earnings call, CEO Marc Benioff addressed a growing concern across enterprise technology leaders: large language models may soon evolve beyond infrastructure tools and become full application platforms, similar to how operating systems once reshaped the software industry. 

Benioff acknowledged that companies such as OpenAI and Anthropic could eventually operate as platforms themselves. However, Salesforce’s strategy is not to compete at the model layer. Instead, the company is focusing on what enterprise organizations still need beyond raw AI intelligence: trusted data context, governance, compliance, workflow execution, reliability, and security. 

According to Salesforce leadership, foundation models generate intelligence, but enterprises create value only when that intelligence completes real business work. Salesforce is positioning its Agentforce platform as the layer that turns AI outputs into operational outcomes across sales, service, and customer experience environments. 

Benioff described foundation models as a new infrastructure layer sitting beneath enterprise applications. While Salesforce integrates models from multiple partners, including OpenAI and Anthropic, the company believes models alone cannot safely run regulated customer operations at scale. Enterprise environments still require systems capable of managing compliance, availability, scalability, and secure automation where humans and AI agents work together. 

Adoption data shared during the call shows that enterprise agent deployment remains in early stages. Salesforce reported nearly 50 percent growth in Agentforce customers during the fourth quarter, reaching approximately 22,000 to 23,000 customers. However, this represents only a portion of Salesforce’s broader ecosystem of more than 150,000 customers worldwide, indicating that agentic AI adoption is still emerging. 

Salesforce executives also emphasized a shift in how AI value should be measured. Rather than tracking token consumption, the company introduced Agentic Work Units (AWUs), a metric focused on completed business actions such as updating records, resolving service cases, or generating operational documents. The goal is to evaluate AI based on productivity outcomes instead of computational usage. 

President and Chief Marketing Officer Patrick Stokes explained that enterprise AI becomes meaningful only when it produces measurable work, not when it simply generates responses. Salesforce now analyzes how efficiently customers convert AI usage into completed tasks, helping organizations optimize adoption and return on investment. 

From a commercialization perspective, Salesforce outlined a three part approach to monetizing enterprise AI. The company is upgrading customers to premium software tiers with embedded AI capabilities, expanding user adoption as productivity improves, and introducing usage based credits known as Flex Credits for customer facing AI agents. 

While acknowledging that foundation models may eventually become dominant platforms, Salesforce is betting that enterprise value will continue to exist at the operational layer where intelligence becomes governed, contextualized, and executable within business systems. 

For enterprise leaders evaluating AI strategy, Salesforce’s message is clear: the competitive advantage will not come from owning models, but from deploying agents that can safely perform real customer work at scale. 

The release introduces several key enhancements aimed at enabling organizations to operationalize outbound service scenarios with greater accuracy and reliability. Among the major additions is Answering Machine Detection (AMD), which allows organizations to automatically identify voicemail or answering machines during outbound calls. When detected, predefined workflows configured through Copilot Studio can trigger actions such as playing customized messages or ending calls, reducing nonproductive agent connections. 

Microsoft has also introduced SIP based call outcome tracking, enabling outbound calls to be automatically categorized using standardized SIP response codes. Outcomes such as Live Answer, Answering Machine, Busy, No Answer, or Invalid Address are captured without additional configuration, providing clearer reporting visibility and enabling service teams to determine appropriate follow up actions. 

Another major capability now generally available is Predictive Dial Mode, designed specifically for service operations. The feature places outbound calls ahead of customer service representative availability by estimating agent readiness using operational metrics such as abandonment rate and average wait time. This approach helps increase connection success rates while improving agent utilization and overall customer experience during high volume service operations. 

Microsoft also outlined upcoming enhancements as proactive engagement capabilities continue to evolve. Conversational SMS support for proactive engagement is currently available in preview, allowing organizations to extend outbound communication beyond voice while maintaining routing, compliance, and outcome tracking standards. Additional improvements to preview dial mode will provide agents with richer customer context before initiating calls, while future native campaign management capabilities will introduce features such as contact chaining, outcome based retries, consent management, configurable calling windows, and file uploads. 

The update, shared by Deepa Bansal, Principal Product Manager for Dynamics 365 Customer Service, marks a significant milestone in Microsoft’s AI led evaluation framework designed to automate and standardize quality assurance across customer interactions. 

Effective February 6, QEA’s conversation evaluation capabilities are now generally available, complementing the previously released case evaluation functionality. The expansion allows organizations to automatically assess live and recorded customer conversations alongside case based evaluations, improving consistency and scalability in service quality oversight. 

Quality Evaluation Agent enables support teams to automate interaction reviews, reduce manual quality monitoring efforts, and maintain standardized evaluation criteria across large volumes of customer engagements. The addition of conversation capabilities broadens QEA’s applicability across contact center and customer service environments where voice and chat interactions play a central role. 

Microsoft also outlined upcoming enhancements planned for the evaluation framework. Future updates will introduce multilingual support, criteria versioning, the ability to flag critical evaluation questions, simulation capabilities, improved adherence to knowledge sources, and additional evaluation intelligence features. 

Announced by Ritu Joshi, Product Manager for Dynamics 365, the update enhances monitoring for batch workloads in Dynamics 365 Finance, enabling IT teams and administrators to move beyond traditional logging toward proactive performance monitoring and operational intelligence. 

As ERP environments evolve into more agentic systems, background processes such as batch jobs, workflow orchestration, and data integrations now play a direct role in delivering financial reporting, supply chain accuracy, and business intelligence outcomes. Microsoft noted that failures or delays in these workloads can create downstream operational disruptions, making advanced observability essential. 

Dynamics 365 ERP applications already integrate with Azure Application Insights for telemetry collection. The latest update expands batch telemetry signals to provide deeper execution level visibility, including batch start and stop events, failure correlation data, throttling indicators, thread availability insights, and queue depth metrics across priority based scheduling queues. 

These telemetry signals are emitted directly into customer owned Azure Application Insights environments, allowing organizations to continue using existing monitoring dashboards, alerting pipelines, and analytics workflows without changing how telemetry data is consumed. 

The enhanced observability enables IT teams to analyze execution behavior using Kusto Query Language, helping organizations investigate performance issues faster, detect workload regressions early, plan infrastructure capacity based on real usage patterns, and align service level expectations with operational performance. 

Microsoft highlighted practical enterprise scenarios demonstrating the impact of expanded telemetry. In one example, a global consumer goods company identified delayed high priority batch jobs caused by queue congestion and thread exhaustion, allowing teams to correct scheduling configurations and restore workload prioritization. In another case, finance teams troubleshooting delayed bank reconciliation processes discovered thread starvation issues that prevented queued jobs from executing, enabling more accurate batch capacity planning. 

According to Microsoft, high fidelity batch telemetry also lays the groundwork for intelligent ERP operations powered by AI agents. These capabilities support automated anomaly detection, correlation between system performance and business impact, and smarter alerting tied to operational outcomes rather than isolated technical events. 

Organizations can enable the new telemetry capabilities by integrating Dynamics 365 ERP environments with Azure Application Insights and activating expanded batch signals through system administration settings. 

The announcement, made by Rupa Mantravadi, Principal Group Program Manager for Dynamics 365 Project Operations, follows the earlier retirement of PSA on the commercial cloud, where support concluded on March 31, 2025. 

Beginning in March 2027, Microsoft will discontinue all support for PSA in GCC environments. The platform will no longer receive feature enhancements, security updates, bug fixes, or technical updates. Microsoft confirmed that any support tickets raised for PSA after this date will be closed with guidance directing customers to upgrade to Dynamics 365 Project Operations. 

Dynamics 365 Project Operations has been available in the U.S. Government cloud since December 2025 and is positioned as PSA’s long term replacement. Microsoft is encouraging government customers to transition early to access expanded project management and financial capabilities. 

Key capabilities available in Project Operations include integration with Planner on Dataverse for advanced scheduling, project budgeting with time phased forecasting, date effective price overrides, quote revision and activation, material usage tracking across projects and tasks, subcontract management, advances and retained based contracts, contract not to exceed controls, task and progress based billing, multi customer contract support, and AI and Copilot powered experiences. 

Microsoft noted that a separate future announcement will address availability timelines for Dynamics 365 Project Operations in GCC High and Department of Defense environments.