Scaling enterprise AI requires overcoming architectural oversights that often stall pilots before production, a challenge that goes far beyond model selection. While generative AI prototypes are easy to spin up, turning them into reliable business assets involves solving the difficult problems of data engineering and governance.<\/p>\n
Ahead of AI & Big Data Global 2026 in London, Franny Hsiao, EMEA Leader of AI Architects at Salesforce, discussed why so many initiatives hit a wall and how organisations can architect systems that actually survive the real world.<\/p>\n
The \u2018pristine island\u2019 problem of scaling enterprise AI<\/p>\n
Most failures stem from the environment in which the AI is built. Pilots frequently begin in controlled settings that create a false sense of security, only to crumble when faced with enterprise scale.<\/p>\n
\u201cThe single most common architectural oversight that prevents AI pilots from scaling is the failure to architect a production-grade data infrastructure with built-in end to end governance from the start,\u201d Hsiao explains.<\/p>\n
\u201cUnderstandably, pilots often start on \u2018pristine islands\u2019 \u2013 using small, curated datasets and simplified workflows. But this ignores the messy reality of enterprise data: the complex integration, normalisation, and transformation required to handle real-world volume and variability.\u201d<\/p>\n
When companies attempt to scale these island-based pilots without addressing the underlying data mess, the systems break. Hsiao warns that \u201cthe resulting data gaps and performance issues like inference latency render the AI systems unusable\u2014and, more importantly, untrustworthy.\u201d<\/p>\n
Hsiao argues that the companies successfully bridging this gap are those that \u201cbake end-to-end observability and guardrails into the entire lifecycle.\u201d This approach provides \u201cvisibility and control into how effective the AI systems are and how users are adopting the new technology.\u201d<\/p>\n
Engineering for perceived responsiveness<\/p>\n
As enterprises deploy large reasoning models \u2013 like the \u2018Atlas Reasoning Engine\u2019 \u2013 they face a trade-off between the depth of the model\u2019s \u201cthinking\u201d and the user\u2019s patience. Heavy compute creates latency.<\/p>\n
Salesforce addresses this by focusing on \u201cperceived responsiveness through Agentforce Streaming,\u201d according to Hsiao.<\/p>\n
\u201cThis allows us to deliver AI-generated responses progressively, even while the reasoning engine performs heavy computation in the background. It\u2019s an incredibly effective approach for reducing perceived latency, which often stalls production AI.\u201d<\/p>\n
Transparency also plays a functional role in managing user expectations when scaling enterprise AI. Hsiao elaborates on using design as a trust mechanism: \u201cBy surfacing progress indicators that show the reasoning steps or the tools being used, as well images like spinners and progress bars to depict loading states, we don\u2019t just keep users engaged; we improve perceived responsiveness and build trust.<\/p>\n
\u201cThis visibility, combined with strategic model selection \u2013 like choosing smaller models for fewer computations, meaning faster response times \u2013 and explicit length constraints, ensures the system feels deliberate and responsive.\u201d<\/p>\n
Offline intelligence at the edge<\/p>\n
For industries with field operations, such as utilities or logistics, reliance on continuous cloud connectivity is a non-starter. \u201cFor many of our enterprise customers, the biggest practical driver is offline functionality,\u201d states Hsiao.<\/p>\n
Hsiao highlights the shift toward on-device intelligence, particularly in field services, where the workflow must continue regardless of signal strength.<\/p>\n
\u201cA technician can photograph a faulty part, error code, or serial number while offline. Then an on-device LLM can then identify the asset or error, and provide guided troubleshooting steps from a cached knowledge base instantly,\u201d explains Hsiao.<\/p>\n
Data synchronisation happens automatically once connectivity returns. \u201cOnce a connection is restored, the system handles the \u2018heavy lifting\u2019 of syncing that data back to the cloud to maintain a single source of truth. This ensures that work gets done, even in the most disconnected environments.\u201d<\/p>\n
Hsiao expects continued innovation in edge AI due to benefits like \u201cultra-low latency, enhanced privacy and data security, energy efficiency, and cost savings.\u201d<\/p>\n
High-stakes gateways<\/p>\n
Autonomous agents are not set-and-forget tools. When scaling enterprise AI deployments, governance requires defining exactly when a human must verify an action. Hsiao describes this not as dependency, but as \u201carchitecting for accountability and continuous learning.\u201d<\/p>\n
Salesforce mandates a \u201chuman-in-the-loop\u201d for specific areas Hsiao calls \u201chigh-stakes gateways\u201d:<\/p>\n
\u201cThis includes specific action categories, including any \u2018CUD\u2019 (Creating, Uploading, or Deleting) actions, as well as verified contact and customer contact actions,\u201d says Hsiao. \u201cWe also default to human confirmation for critical decision-making or any action that could be potentially exploited through prompt manipulation.\u201d<\/p>\n
This structure creates a feedback loop where \u201cagents learn from human expertise,\u201d creating a system of \u201ccollaborative intelligence\u201d rather than unchecked automation.<\/p>\n
Trusting an agent requires seeing its work. Salesforce has built a \u201cSession Tracing Data Model (STDM)\u201d to provide this visibility. It captures \u201cturn-by-turn logs\u201d that offer granular insight into the agent\u2019s logic.<\/p>\n
\u201cThis gives us granular step-by-step visibility that captures every interaction including user questions, planner steps, tool calls, inputs\/outputs, retrieved chunks, responses, timing, and errors,\u201d says Hsiao.<\/p>\n
This data allows organisations to run \u2018Agent Analytics\u2019 for adoption metrics, \u2018Agent Optimisation\u2019 to drill down into performance, and \u2018Health Monitoring\u2019 for uptime and latency tracking.<\/p>\n
\u201cAgentforce observability is the single mission control for all your Agentforce agents for unified visibility, monitoring, and optimisation,\u201d Hsiao summarises.<\/p>\n
Standardising agent communication<\/p>\n
As businesses deploy agents from different vendors, these systems need a shared protocol to collaborate. \u201cFor multi-agent orchestration to work, agents can\u2019t exist in a vacuum; they need common language,\u201d argues Hsiao.<\/p>\n
Hsiao outlines two layers of standardisation: orchestration and meaning. For orchestration, Salesforce is adopting open-source standards like MCP (Model Context Protocol) and A2A (Agent to Agent Protocol).\u201d<\/p>\n
\u201cWe believe open source standards are non-negotiable; they prevent vendor lock-in, enable interoperability, and accelerate innovation.\u201d<\/p>\n
However, communication is useless if the agents interpret data differently. To solve for fragmented data, Salesforce co-founded OSI (Open Semantic Interchange) to unify semantics so an agent in one system \u201ctruly understands the intent of an agent in another.\u201d<\/p>\n
The future enterprise AI scaling bottleneck: agent-ready data<\/p>\n
Looking forward, the challenge will shift from model capability to data accessibility. Many organisations still struggle with legacy, fragmented infrastructure where \u201csearchability and reusability\u201d remain difficult.<\/p>\n
Hsiao predicts the next major hurdle \u2013 and solution \u2013 will be making enterprise data \u201c\u2018agent-ready\u2019 through searchable, context-aware architectures that replace traditional, rigid ETL pipelines.\u201d This shift is necessary to enable \u201chyper-personalised and transformed user experience because agents can always access the right context.\u201d<\/p>\n
\u201cUltimately, the next year isn\u2019t about the race for bigger, newer models; it\u2019s about building the orchestration and data infrastructure that allows production-grade agentic systems to thrive,\u201d Hsiao concludes.<\/p>\n
Salesforce is a key sponsor of this year\u2019s AI & Big Data Global in London and will have a range of speakers, including Franny Hsiao, sharing their insights during the event. Be sure to swing by Salesforce\u2019s booth at stand #163 for more from the company\u2019s experts.<\/p>\n
See also: Databricks: Enterprise AI adoption shifts to agentic systems<\/p>\n
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information.<\/p>\n
AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
\nThe post Franny Hsiao, Salesforce: Scaling enterprise AI appeared first on AI News.<\/p>\n","protected":false},"excerpt":{"rendered":"
Scaling enterprise AI requires overcoming architectural oversights that often stall pilots before production, a challenge that goes far beyond model selection. While generative AI prototypes are easy to spin up, turning them into reliable business assets involves solving the difficult problems of data engineering and governance. Ahead of AI & Big Data Global 2026 in [\u2026]<\/p>\n
The post Franny Hsiao, Salesforce: Scaling enterprise AI<\/a> appeared first on AI News<\/a>.<\/p>\n<\/div>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"site-container-style":"default","site-container-layout":"default","site-sidebar-layout":"default","disable-article-header":"default","disable-site-header":"default","disable-site-footer":"default","disable-content-area-spacing":"default","footnotes":""},"categories":[1,66,71,58,69,72],"tags":[3],"class_list":["post-886","post","type-post","status-publish","format-standard","hentry","category-ai-and-ml","category-ai-and-us","category-ai-business-strategy","category-ai-market-trends","category-features","category-governance-regulation-policy","tag-ai"],"yoast_head":"\n