Innovate Your Business
with Our Insights!

At the core is live data access inside everyday apps: finance, inventory, and customer records from Business Central surface directly in Excel for two-way editing, in Outlook for quote and invoice actions, and in Teams for collaboration on live records. Power Automate stitches approvals and notifications across systems, while Copilot, Agent Builder, and Copilot Studio add AI-driven reconciliation, order handling, and custom agents that execute routine tasks. 

Microsoft says the shared identity layer via Azure Active Directory simplifies security and access control, and that the combined stack shortens decision cycles by bringing analytics and recommendations to where people already work. A showcased scenario, Sales Order Agent, fulfills inbound email orders against real-time inventory without leaving Outlook; finance teams can run close activities in Excel with synchronized updates back to Business Central. 

Early adopters point to measurable gains. Logistics firm Sage Freight reports a 33 percent faster close and stronger on-time receivables after unifying processes on the Microsoft stack. Microsoft claims more than 45,000 SMBs now use Business Central, with AppSource extensions expanding industry-specific workflows. 

The company frames the move as a strategic shift rather than a feature drop: unify data, automate workflows, and layer AI so teams can scale without multiplying tools. For organizations already standardized on Microsoft 365, the pitch is straightforward—turn familiar apps into a single, AI-ready workspace for finance, sales, service, and operations. 

Until now, the Field Service Mobile app has been primarily offline-first, ensuring workers could continue servicing assets without interruption even when networks failed. While this approach has proven critical in remote or low-connectivity settings, it limited access to real-time updates when signals were available. Online Mode changes that by allowing users to tap into live Dataverse data directly from the application. 

With this capability, technicians can search for records beyond their offline profile, view updates to work orders or assets as they happen, and immediately retrieve schedules or notes without waiting for synchronization. Files saved to the cloud, such as those in SharePoint, also become instantly available. The application allows users to move seamlessly between online and offline modes depending on job requirements or signal availability, ensuring productivity remains consistent regardless of network conditions. 

For IT managers, enabling Online Mode is straightforward, requiring only a configuration update in the Power Apps maker experience. Once activated, the option appears in the app, giving frontline workers the ability to toggle between modes without additional complexity. 

Microsoft recommends rolling out the feature gradually, beginning with technicians in areas where connectivity is strongest, and training teams to understand the differences between operating modes. By combining offline dependability with the advantages of live access, the new mode reduces delays, improves collaboration, and helps ensure technicians never miss critical updates in the field. 

The company has rolled out vertical-focused Agentforce products across retail, field service, human resources, financial services, revenue management, public sector, and most recently manufacturing. Agentforce for Manufacturing, launched in August 2025, integrates with ERP and IoT systems to help companies predict demand, optimize asset management, and reduce downtime. Earlier releases such as Agentforce for Retail and Agentforce for Financial Services introduced pre-built agent skills for customer service, guided shopping, loyalty promotions, and compliance-focused financial tasks. 

Salesforce is also addressing customer concerns about oversight and transparency. In May, the company introduced governance and security features including Backup and Recover, Archive, Data Mask, Shield 2.0, and Security Center 2.0. These tools allow organizations to monitor agent activity, enforce compliance policies, and safeguard data across Salesforce and third-party applications. The Agentforce Testing Center, launched in late 2024, gives customers the ability to audit and validate agent behavior before deployment. 

Integration remains a core differentiator. Agentforce connects with Salesforce products such as Data Cloud, Customer 360, MuleSoft, Tableau, Slack, and Heroku. With the release of Agentforce 3 in June 2025, Salesforce added a Command Center to provide observability and monitoring of agent interactions. It also introduced native support for the Model Context Protocol, allowing Agentforce agents to connect directly to enterprise servers from partners like AWS, Google Cloud, IBM, and Stripe. 

Pricing flexibility has also been a priority. Beyond the initial $2 per conversation model, Salesforce now offers pay-as-you-go Flex Credits, pre-commit contracts for discounted usage, and per-user licenses with bundled Agentforce capabilities. These options are designed to align cost with outcomes, making it easier for businesses to start with small pilots or scale across large deployments. 

Customer adoption has been rapid. Salesforce reports that its own help portal now resolves 83 percent of support queries using Agentforce, reducing human escalations by half. Clients including RBC, Accenture, Goodyear, PepsiCo, and L’Oréal are deploying Agentforce for wealth management, proposal automation, customer support, and personalized shopping experiences. 

With Dreamforce 2025 on the horizon, Salesforce is expected to highlight further use cases, particularly within Marketing Cloud Next. 

The new capabilities build on intent mining, dynamic interviews, and knowledge-driven automation. The agent is now more conversational and configurable, using natural language instructions to guide behavior. Dynamic interviews allow the system to persist information, shift smoothly between topics, and handle interruptions. Playbook-driven workflows combine knowledge articles and connected actions to deliver tailored resolutions, while intent mapping ensures each case is routed to the right logic or even to custom-built AI agents. 

Voice support is a centerpiece of the release. The Customer Intent Agent can now manage calls as well as chats, using natural speech patterns and fast responses that mimic human conversation. Unlike traditional scripted bots, the system dynamically plans each exchange, narrowing down intent and offering solutions without pre-programmed flows. Advanced speech recognition ensures customers can speak naturally and still be understood. 

The release also overhauls intent discovery and management. By leveraging business context, industry profiles, and product hierarchies, the agent now generates fewer but more precise intents. Admins can approve, merge, or segment intents by business line, supported by new analytics dashboards and sample-driven validation. This makes it easier to maintain intent libraries that truly reflect customer needs. 

Learning loops are another highlight. The Customer Intent Agent can continuously improve by extracting new intents and questions from completed conversations. Both semi-autonomous and fully autonomous feedback cycles are supported, enabling organizations to generate new knowledge articles and refine responses based on real-world interactions. 

For organizations aiming to reduce manual effort, cut resolution times, and deliver more consistent customer experiences, the Customer Intent Agent offers a preview of how AI-driven service can operate at scale. 

MAI-Voice-1 is the company’s first expressive speech generation model. Now live in Copilot Daily, Podcasts, and Copilot Labs, it delivers high-fidelity, natural audio across single and multi-speaker scenarios. Capable of generating a full minute of speech in under a second on a single GPU, the model is positioned as one of the fastest speech systems available. Microsoft is showcasing MAI-Voice-1 in Copilot Labs with storytelling and guided meditation demos, highlighting its potential in creative and productivity use cases. 

Alongside its voice model, Microsoft has begun public testing of MAI-1-preview, the company’s first end-to-end foundation model. Trained on approximately 15,000 NVIDIA H100 GPUs using a mixture-of-experts approach, the model is designed to follow instructions and provide reliable responses across everyday scenarios. It is currently accessible via the LMArena evaluation platform and will soon be integrated into select Copilot text-based features. Trusted testers can also apply for early API access to provide feedback. 

Microsoft AI leaders described these releases as the foundation for a broader strategy of orchestrating specialized models tailored to diverse user needs. The company highlighted its next-generation GB200 compute cluster as part of its roadmap for advancing scale and efficiency. 

By developing proprietary models alongside those from partners and the open-source community, Microsoft AI says it aims to deliver flexible, high-performance systems capable of powering billions of daily interactions across products. The announcements reinforce the company’s ambition to shape the future of AI companions with trusted, category-defining platforms.