Xsight Group has been featured as a research case in Harvard Business Review’s September–October 2026 issue, highlighting how the Hong Kong-based technology group uses modular AI agents and orchestration to coordinate complex cross-functional decisions. The case reflects a broader shift in enterprise AI from automating individual tasks toward connecting specialized agents, shared data and human judgment across organizational silos.
Enterprise AI has largely focused on making individual employees or departments more productive. But the harder problem is what happens when a business decision crosses procurement, logistics, finance, legal, sales and other functions at the same time.
That challenge is at the center of a new Harvard Business Review research case featuring Xsight Group, a Hong Kong-headquartered technology company whose approach uses specialized AI agents and an orchestration layer to coordinate complex decisions across organizational boundaries.
The September–October 2026 HBR article, How AI Agents Orchestrate Work Across Silos, argues that many companies have reached the limits of task-level AI because individual agents typically operate within functional boundaries. When their outputs are not coordinated, recommendations can conflict or fail to account for information held elsewhere in the organization.
Xsight’s approach is notable because it starts with the decision itself rather than the AI model.
The company breaks complex decisions into defined modules with explicit inputs, outputs, constraints and objectives. Specialized agents can then handle individual components such as contract analysis, demand forecasting and capacity planning, while an orchestration layer coordinates the results.
That architecture addresses one of the emerging problems in AI agents and autonomous systems: a collection of capable agents does not automatically create an intelligent enterprise.
Gartner’s recent analysis of more than 100 agentic AI deployments found that the largest opportunity is increasingly in specialized, domain-specific agents. The research firm predicts that 80% of tangible ROI from agentic AI will come from specialized agents by 2028.
The Xsight case fits that direction. Rather than asking one general-purpose agent to understand an entire business, its architecture distributes work among agents with narrower responsibilities and uses orchestration to connect their outputs.
The company says connectors validate results against management-defined guardrails and route information between workflows, while a master AI orchestrator structures analyses and incorporates human input.
That human layer is important. HBR’s research describes effective agentic orchestration as a combination of AI analysis, information routing and trade-off identification with human context, tacit knowledge, guardrails and final decision-making.
In other words, the goal is not to create an autonomous executive.
It is to reduce the coordination workload surrounding executive decisions.
This distinction could become increasingly important as organizations deploy larger numbers of AI agents. Gartner predicts that 33% of enterprise software applications will include agentic AI capabilities by 2028, up from less than 1% in 2024. It also expects at least 15% of day-to-day work decisions to be made autonomously through agentic AI by that point.
But scaling agentic AI across departments introduces a new infrastructure problem: interoperability.
Procurement may have one definition of demand, finance another definition of acceptable risk, while logistics operates with different capacity assumptions. An agent can produce an accurate recommendation based on its own data and still contribute to a bad enterprise decision if the assumptions underlying that recommendation differ from those used elsewhere.
Standardized task definitions and shared information models therefore become part of the AI infrastructure.
Xsight’s broader business provides a useful environment for developing that approach. Its UX168 platform operates across international sales channels and a global supply chain network, while its XsOS operating system is designed to standardize and modularize activities spanning product selection, pricing, fulfillment and marketing.
The company says XsOS has helped increase its new-product hit rate from 10% to more than 20%. That is a company-reported performance figure rather than an independently verified benchmark.
Xsight also says it serves more than 80 million consumers globally through more than 30 international sales channels, with operations and R&D across locations including Hong Kong, mainland China, the United States, the Netherlands and Japan.
That operational complexity provides a natural test case for AI automation platforms and enterprise orchestration. Cross-border supply chains contain precisely the kinds of dependencies that individual AI assistants struggle to manage: inventory, demand, pricing, manufacturing, transportation, regulatory requirements and market conditions can change simultaneously.
The HBR case also points toward a broader evolution in enterprise AI architecture.
The first wave concentrated on copilots and task automation. The next phase is increasingly about agents collaborating across applications and business functions. Gartner expects collaborative agents to become more important as enterprises move toward ecosystems in which specialized agents can dynamically work across multiple applications.
That creates opportunities for platforms that can provide the connective tissue between models, business applications, data and governance controls.
It also raises a question about where enterprise AI value ultimately resides. Models remain important, but organizations may increasingly differentiate themselves through proprietary data, workflow design, domain expertise and the orchestration layer that determines how AI capabilities interact.
For Xsight, that is part of a larger strategic transition. The company says it intends to evolve from a technology-enabled trading business into an open vertical digital supply chain platform.
Whether that ambition becomes a broader industry model remains to be seen. But the HBR case highlights a practical lesson for enterprises adopting agentic AI: more agents are not necessarily better.
The greater opportunity may come from decomposing complex decisions, giving specialized agents clearly defined responsibilities, enforcing shared rules and allowing humans to resolve the trade-offs that machines cannot fully contextualize.
That moves agentic AI away from the idea of replacing management and toward something potentially more useful: an AI orchestration layer for enterprise decision-making.
Market Landscape
The enterprise AI market is moving from isolated copilots toward agentic AI orchestration, where specialized agents collaborate across applications and business functions. HBR’s research identifies this as a next frontier for enterprise decision-making, with humans still supplying context, tacit knowledge and final judgment.
Gartner’s research points in a similar direction: specialized agents are expected to produce most tangible agentic AI ROI by 2028, while collaborative agents are expected to increasingly operate across application and data environments.
The competitive opportunity is therefore shifting beyond individual LLM capabilities. Enterprise AI platforms will increasingly need agent orchestration, interoperability, workflow redesign, governance, shared data models and human oversight.
Top Insights
- Xsight’s HBR case highlights the shift from task-level AI assistants toward orchestration systems connecting specialized agents across business functions.
- Modularizing complex decisions can give AI agents clearer responsibilities while creating common inputs, outputs, constraints and objectives.
- Gartner expects specialized domain-specific agents to generate most tangible agentic AI ROI by 2028.
- Human judgment remains essential when enterprise decisions involve conflicting objectives, incomplete information and context that cannot be encoded easily.
- Supply chain operations provide a strong use case for agentic orchestration because decisions span demand, inventory, pricing, manufacturing and logistics.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
