Microsoft Foundry’s Agentic Bet: From Pilots to Production
Microsoft Foundry agents are production-ready AI systems that combine frontier language models, governed runtime infrastructure, and direct integration with everyday work applications to automate complex, multi-step tasks for large organizations. Instead of isolated chatbots or proof-of-concept demos, Foundry concentrates on reliable, observable agents that can be deployed, monitored, and scaled across finance, HR, legal, and other enterprise functions in a single platform. This shift matters because AI only creates value when it shows up in real systems—systems that are reliable, observable, and aligned to business outcomes.
Microsoft is now pushing that vision into general availability by combining frontier models and a production runtime in one place. GPT‑5.6 is now generally available in Microsoft Foundry alongside new hosted agents in Foundry Agent Service and an Asia-Pacific Data Zone that keeps data processing within the region for compliant deployments. More than 100,000 organizations are already building on the platform, and some, including Adobe, Telefónica, and Tata Consultancy Services, are already running agents in production. The message is clear: Foundry is no longer an experiment; it is an opinionated agentic workflow platform aimed squarely at enterprise AI automation.
Frontier Models Built for Work, Not Demos
The new GPT‑5.6 model family underscores that this launch is less about novelty and more about matching the right engine to the right workload. GPT‑5.6 Sol delivers the most advanced reasoning capabilities yet, explicitly tuned for extended reasoning, agentic workflows, and code-focused scenarios that power demanding enterprise workloads. GPT‑5.6 Terra is positioned as the balanced choice for everyday work, providing performance competitive with GPT‑5.5 at a lower cost and making it ideal for scaling intelligent applications across the enterprise. GPT‑5.6 Luna, the fastest and most affordable model in the lineup, targets high-volume, latency-sensitive workloads where speed and unit economics matter most.
Microsoft is also making these models broadly reachable from day one, with GPT‑5.6 available through Global Standard and Global Priority Processing across existing regions and Data Zones. Quote: “GPT‑5.6 Sol delivers the most advanced reasoning capabilities yet, supporting extended reasoning, agentic workflows, and code-focused scenarios, for the most demanding enterprise workloads.” Pricing is clearly tiered: GPT‑5.6 Sol at USD 5.00 (approx. RM23) input and USD 30.00 (approx. RM138) output per million tokens, GPT‑5.6 Terra at USD 2.50 (approx. RM11.50) input and USD 15.00 (approx. RM69) output, and GPT‑5.6 Luna at USD 1.00 (approx. RM4.60) input and USD 6.00 (approx. RM27.60) output.
Long-Running Agents for Finance, HR, and Legal
Where Foundry becomes strategically interesting is in long-running, production AI agents that can handle complex, multi-step workflows. Hosted agents in Foundry Agent Service are now generally available, giving developers one production runtime for agents built with frameworks like Microsoft Agent Framework, GitHub Copilot SDK, LangGraph, OpenClaw, and Hermes. These agents inherit enterprise identity, security, and compliance controls by default, so they can safely access tools, data, and users without stitching together disconnected systems.
For finance, HR, and legal teams, the critical upgrade is resilience. New resilient task support in hosted agents (currently in private preview) makes it easier to build agents that survive failures and resume long-running workloads. The platform keeps the sandbox running while the agent harness handles checkpointing, allowing multi-turn conversations, reasoning loops, and human-in-the-loop approvals to pick up where they left off without custom recovery, retry, and state-management logic. Memory and routines in Foundry Agent Service, now in public preview, add procedural, user, and session memory so agents can carry context across interactions and respond to real-world events. This is exactly what complex finance approvals, HR case handling, or contract workflows need to move from scripted automation to adaptive, agentic workflows.
From AI Labs to Enterprise Desks
Enterprise AI automation has often stalled at the last mile: getting agents in front of the people who actually do the work. Foundry’s answer is to treat distribution as a first-class feature. Production agents in Foundry Agent Service now come with toolboxes and publishing paths to Microsoft 365 Copilot and Microsoft Teams. Publishing to Microsoft Teams and Microsoft 365 Copilot will be generally available next week, meaning the same agents developers build can appear in the applications where hundreds of millions of people already work, with identity, permissions, and policy flowing through automatically.
This is not just a convenience; it turns Foundry into one of the few agentic workflow platforms where build, run, govern, and distribute are all covered. GPT‑5.6 Terra’s positioning as a model “for everyday work” makes sense in this context: it is aimed at scaling intelligent features across regular business tasks rather than isolated pilot projects. For ordinary users, the shift will feel less like adopting a new AI product and more like their existing tools quietly gaining agents that can orchestrate multi-step processes, summarize context, and trigger workflows across systems. The risk for slow-moving enterprises is obvious: in a world where agents are a click away in core productivity tools, delay means handing the advantage to faster competitors who treat Foundry as a production platform, not an experiment.
Compliance as a Competitive Weapon
A major reason many enterprises hesitate on AI is data governance. Foundry tackles this by expanding where and how frontier models can run while staying compliant. The Asia-Pacific Data Zone is now generally available for Microsoft Foundry, giving customers in that region an option to run OpenAI frontier models while keeping data processing anchored locally. With Global, Data Zone, and Regional deployment options in one platform, organizations can align AI adoption to sovereignty, compliance, performance, and scale requirements without fragmenting their development and operations experience.
According to Viva Republica’s Chief Data and AI Officer, Microsoft Foundry’s APAC Data Zone allows them to keep data processing regionally anchored while accessing advanced AI models at scale, giving them confidence to accelerate AI innovation responsibly and pursue their goal of being a leading AI-powered financial platform in Asia. This is the strategic angle: compliance is not a brake; it is a moat. By placing frontier models, production AI agents, and distribution hooks into the same governed platform, Microsoft is betting that the next wave of competitive advantage will come from enterprises that can automate entire workflows—finance, HR, legal, and beyond—without breaking their regulatory posture.
