Mastering Enterprise AI: Salesforce’s Guide to Scaling Success

Scaling Enterprise AI: Overcoming Architectural Oversights

Scaling enterprise AI involves addressing architectural oversights that often hinder pilot projects from progressing to production. This challenge extends beyond just selecting the right models. While it’s easy to create generative AI prototypes, transforming them into reliable business assets requires tackling complex issues related to data engineering and governance.

Franny Hsiao, the EMEA Leader of AI Architects at Salesforce, recently discussed the common obstacles that impede AI initiatives and how organizations can design systems that can withstand real-world challenges ahead of the AI & Big Data Global 2026 event in London.

The Challenge of Scaling Enterprise AI

Many AI initiatives fail due to the controlled environments in which they are developed. Pilot projects often start in isolated settings that provide a false sense of security but struggle to scale when faced with enterprise-level demands.

Hsiao emphasizes that a key oversight preventing AI pilots from scaling is the lack of a production-grade data infrastructure with built-in end-to-end governance from the outset. Starting on “pristine islands” with small, curated datasets and simplified workflows overlooks the complexity of real-world enterprise data integration, normalization, and transformation requirements.

Without addressing the underlying data challenges, attempting to scale these pilots leads to system failures, data gaps, and performance issues like inference latency, rendering the AI systems unreliable. Hsiao suggests that successful companies address this gap by embedding observability and control throughout the entire AI system lifecycle, offering visibility into system effectiveness and user adoption.

Engineering for Perceived Responsiveness

Deploying large reasoning models can create a trade-off between model depth and user patience, as heavy compute results in latency. Salesforce addresses this issue by focusing on perceived responsiveness through Agentforce Streaming, delivering AI-generated responses progressively while the reasoning engine performs intensive computations in the background.

Transparency plays a crucial role in managing user expectations during enterprise AI scaling. By incorporating design elements like progress indicators and optimizing model selection for faster response times, organizations can improve perceived responsiveness and build user trust.

Offline Intelligence at the Edge

For industries with field operations, such as utilities or logistics, reliance on continuous cloud connectivity may not be feasible. Offline functionality becomes crucial for uninterrupted workflow. Hsiao highlights the shift towards on-device intelligence, particularly in field services, where on-device Local Language Models (LLMs) can identify assets, errors, and provide troubleshooting steps offline before synchronizing data once connectivity is restored.

Hsiao anticipates further innovation in edge AI due to benefits like low latency, enhanced privacy, energy efficiency, and cost savings.

High-Stakes Gateways

Autonomous agents require governance to define when human verification is necessary during AI deployment. Salesforce implements a “human-in-the-loop” approach for high-stakes actions, emphasizing accountability and continuous learning rather than dependency. This feedback loop enables agents to learn from human expertise, fostering collaborative intelligence and ensuring transparency through Session Tracing Data Models (STDM).

By capturing step-by-step interactions, organizations can analyze adoption metrics, optimize agent performance, and monitor system health effectively.

Standardizing Agent Communication

As businesses deploy agents from various vendors, standardized communication protocols are essential for collaboration. Salesforce advocates for open-source standards like MCP and A2A for orchestration and OSI to unify semantics, ensuring seamless communication and understanding among agents.

The Future of Enterprise AI Scaling: Agent-Ready Data

The future challenge in enterprise AI scaling will shift from model capability to data accessibility. Organizations must make enterprise data “agent-ready” through context-aware architectures to enable hyper-personalized user experiences. Building robust orchestration and data infrastructure will be crucial for production-grade agentic systems to thrive in the coming years.

As a key sponsor of the AI & Big Data Global event, Salesforce will have experts sharing insights, including Franny Hsiao. For more information, visit Salesforce’s booth at stand #163.

AI News is powered by TechForge Media. Explore upcoming enterprise technology events and webinars here.