Industry executives and experts share their predictions for 2026. Read them in this 18th annual VMblog.com series exclusive.
By Waqas Ahmed, VP of AI Engineering, OpenText
Enterprise AI is entering a decisive year. After several rounds of pilots, organizations are beginning to weave AI into full scale production. As they do, it is becoming clear that the monolithic architectures used during the early experimentation phase cannot support the needed scale, governance or complexity of modern data center environments. In 2026, enterprises will shift toward more modular, context-aware and controlled AI systems that better match real-world operational needs.
From Monolithic Wrappers to Composable Architectures
The most significant architectural change will be the emergence of agentic systems built from many specialized agents, coordinated through choreography rather than centralized control. Many enterprises initially simplified adoption by placing a single workflow or orchestration layer in front of everything. That approach worked when projects were small and isolated, but it will not hold at enterprise scale. Organizations will instead move toward early forms of an agentic mesh, where agents become the primary execution units and coordination replaces reliance on a single controlling layer.
Model strategy will evolve in parallel. Enterprises will increasingly rely on hybrid deployments that combine open models, private LLMs and cloud-based systems. Large frontier models will remain important for high-level reasoning, but smaller, fine-tuned or locally deployed models will take on high-volume and privacy-sensitive workloads. Rather than replacing large models, this approach allows enterprises to better align models to specific use cases, balancing cost, performance and sovereignty.
Context Becomes the Core of Enterprise Intelligence
As AI becomes more embedded in daily operations, organizations will realize that model capability alone is not enough. Real intelligence comes from context. LLMs do not inherently understand business rules, policies or institutional knowledge, which means enterprises must supply that grounding themselves.
In response, organizations will invest more heavily in context layers such as knowledge graphs, metadata systems, RAG 2.0 pipelines and enterprise memory frameworks. These systems anchor AI outputs in verified data, reduce hallucinations and improve reliability. They also elevate data quality, lineage and semantic consistency from supporting concerns to core priorities across hybrid and distributed environments.
Agentic Systems Demand Trust, Oversight and Observability
Agentic AI will also begin influencing how software is built and maintained. In 2026, enterprises will see early versions of systems that propose fixes, generate patches or optimize workflows, initially in tightly controlled environments where oversight is clear and failure modes are well understood. AI assistants will also enable non-technical employees to create simple automations using natural language, expanding adoption beyond traditional technical teams.
As these systems scale, trust and traceability become essential. Because AI is not fully deterministic, enterprises will invest more in observability, reliability engineering and clearly defined risk thresholds. Leaders will expect AI systems to explain how decisions were made and which contextual inputs influenced outcomes. These expectations will shape the evolution of policy, access controls and governance in AI-native environments.
All of these shifts point to a broader transformation from traditional applications to composable, conversational and agentic enterprise systems. The future of AI is not a single model or interface, but an ecosystem of specialized agents, diverse models, shared context layers and governance designed for complexity. In modern data centers, AI will increasingly resemble the rest of the stack to be distributed, modular and observable.
If 2025 was about experimenting with AI, 2026 will be about engineering it with purpose, and the organizations that recognize this shift will lead with new found enterprise intelligence.
##
ABOUT THE AUTHOR
Waqas Ahmed is Vice President of AI Engineering at OpenText, where he leads the design and evolution of enterprise AI platforms across cloud, SaaS, and hybrid environments. His work focuses on scalable AI architecture, agentic systems, and the integration of large language models with enterprise data, governance, and security. He has held leadership roles at Schneider National and SRA Platinum Solutions and holds a bachelor’s degree in computer science, summa cum laude, from Michigan Technological University.
