Microsoft Research leaders described 2025 as a turning point where AI moved beyond marginal gains into systems that “reason and adapt while collaborating” with people, setting up 2026 as a year of scaled impact across science, work, and software.
In day-to-day work, Microsoft’s chief product officer for AI experiences, Aparna Chennapragada, characterized 2026 as a new era of alliances between people and AI, where agents become “digital coworkers” that handle data crunching, content generation, and personalization while humans steer strategy and creativity. She imagined small teams able to launch global campaigns in days, arguing that organizations that design for people to learn and work with AI will “get the best of both worlds” by elevating the human role.
On the software side, GitHub’s Mario Rodriguez pointed to “repository intelligence” as the next edge: AI that understands not just lines of code but their relationships and history, enabling smarter suggestions, earlier error detection, and automated fixes as development activity hits record volumes.
Agents, Security, System Intelligence
Multiple voices said 2026 will be defined by agentic systems—autonomous agents that collaborate, negotiate, and transact on behalf of people and organizations. These agentic ecosystems will reorganize digital marketplaces and require platforms and protocols to be rethought through an “agent-native lens.” They will also hold context across months, track evolving goals, surface forgotten assumptions, and stabilize complex innovation workflows so teams can move faster without losing intent.
With agents joining the workforce, security becomes a first-order concern. Microsoft Security leader Vasu Jakkal said every agent should carry protections comparable to humans, including clear identity, scoped access, managed data, and defenses against attackers so they do not become “double agents” carrying unchecked risk. As such, security will become ambient, autonomous, and built-in, while defenders also deploy security agents to detect and respond to attacks that use AI.
At the infrastructure level, Azure CTO Mark Russinovich described a shift from simply building more datacenters to orchestrating “superfactories” of AI capacity, where computing is densely packed, dynamically routed, and measured by the quality of intelligence produced rather than sheer size. Complementing this, other researchers point to system intelligence, light-based chips, and adaptive infrastructures that are co-designed with AI to unlock the next 1,000x in performance and efficiency.
Quantum, Discovery, the Next Leap
Looking further out, Microsoft EVP Jason Zander argued that quantum computing is entering a “years, not decades” phase where hybrid quantum–AI–supercomputing will begin solving problems classical systems cannot. He pointed to Microsoft’s Majorana 1, which uses topological qubits to stabilize inherently fragile quantum bits and is designed to catch and correct errors, as a key step toward chips with millions of qubits.
Zander linked this to breakthroughs in materials and medicine, suggesting that quantum advantage will redefine how AI and science work together rather than simply accelerating existing approaches. Meanwhile, Jianfeng Gao, head of the Deep Learning Group at Microsoft Research, placed current AI advances in a broader arc of scientific revolutions, arguing the field is shifting focus from encoding world knowledge in large language models to enabling reasoning through interaction, simulation, and “mentalizing” with environments.
Across these perspectives, Microsoft Research presents 2026 as the year AI systems start to behave less like tools and more like collaborators, colleagues, and companions embedded in infrastructure, workflows, and physical systems. The emphasis on agency, context, security, inclusion, and hybrid quantum-AI architectures suggests the next chapter of AI will be judged by how well it aligns with human goals, governance, and global needs.
What This Means for ERP Insiders
AI-driven autonomy will raise expectations for ERP. As agentic systems mature across research, healthcare, and productivity, ERP leaders should anticipate pressures to embed agents that maintain long-running context, reason over governed data, and act across finance, supply chain, and operations with the same “digital coworker” patterns seen elsewhere. This evolution will shift roadmap discussions from adding isolated AI features to designing ERP as a platform where agents, copilots, and humans co-own outcomes.
Infrastructure and system intelligence advances will reshape ERP scalability and integration patterns. The move toward AI “superfactories” and protocol-level system intelligence implies that high-end ERP workloads will increasingly assume elastic, globally optimized infrastructure as a baseline. For enterprise architects and transformation owners, this elevates considerations like context engineering, protocol-mediated actions, and security-by-design for agents to first-class architectural concerns rather than optional enhancements.
Sectoral breakthroughs will influence expectations for vertical ERP and industry clouds. As multimodal, agentic systems make waves in different environments, ERP vendors and partners will face growing demand for domain-fluent agents that can reason over specialized data and processes while preserving trust and psychological well-being. Product and partner strategists who can translate these research trajectories into governed, industry-specific agent capabilities atop ERP platforms will be better positioned to shape the next generation of enterprise solutions.
