> ## Content Index
> Fetch the complete content index at: https://www.stirup.si/llms.txt
> Use this file to discover other available public pages before exploring further.

# Designing agent-first platforms: What changes when agents do the work
- URL: https://www.stirup.si/designing-agent-first-platforms-what-changes-when-agents-do-the-work-443300f1/
- Published: 2026-10-12T05:44:59.000Z
- Updated: 2026-10-12T05:44:59.000Z
- Description: The organizations pulling ahead are not simply adding AI to what they already have. They are designing for a different kind of software.
- Author: Satish Das
- Tags: Cloud, Stirup Research

## Agent-first platforms: A shift in software design

For decades, applications have been designed to wait. A user clicks, a request arrives, code runs, and a response is sent back. This model has been refined over time with scaling, enterprise controls, and operational discipline. However, the rise of agent-first platforms is challenging this traditional approach. These platforms are not simply adding AI to existing systems; they are rethinking the very nature of software design.

Agent-first platforms operate on a fundamentally different principle. Instead of deterministic code where every step is predefined, these systems rely on multi-agent applications that reason through problems at runtime. Developers no longer encode the path; instead, they describe the outcome, and agents determine the steps to achieve it. This shift requires a new infrastructure that supports continuous, autonomous action rather than passive response.

## Challenges in agent-first platform design

Running an agent is no longer the hardest part. Connecting a capable model to tools and grounding it in company data can yield useful results quickly. However, the real challenge lies in transitioning from a pilot to a production-ready agent that the business can rely on. Agents must be governed with the same rigor as any other production system, including identity management, runtime monitoring, and compliance with security standards.

Microsoft Foundry is an example of a platform designed to support agent-first development. It provides agents with first-class identity through Entra Agent ID, ensures traceability, and allows governance through the Foundry Control Plane. However, the execution environment for agents is a separate concern. Most agents inherit the runtime environment of the host application, which can lead to security and operational risks.

## Isolation and execution environments

When agents move from answering questions to completing tasks, they must execute code. This requires a runtime environment that is isolated and secure. Azure Container Apps Sandboxes offers a solution by providing fully isolated environments for each agent execution. These environments are created in seconds, run as a controlled identity, and are dissolved after the task is complete. This isolation ensures that agents have access only to the systems they are permitted to interact with, without compromising security or operational integrity.

The use of hardware-isolated microVMs enables strong separation and sub-second startup times. This architecture allows teams to deploy thousands of agents concurrently without increasing risk. The agent remains governed by Foundry, ensuring that all actions are traceable and compliant with enterprise standards.

## Practical implementation and industry examples

The agent-first model is being adopted across industries, with platforms like Microsoft Foundry and Azure Container Apps Sandboxes forming the foundation. For example, KPMG’s Digital Gateway powered by Claude uses Azure to provide a secure, AI-powered workspace for professionals. This platform supports engagement-specific workspaces, ensuring client data remains isolated and secure.

Similarly, Cognite Atlas AI uses Azure OpenAI models in Foundry to deliver traceable answers to complex industrial questions. By leveraging Azure Container Apps Sandboxes, agents can execute arbitrary code and work with files directly, ensuring isolation and security while maintaining the ability to pause and resume tasks without losing context.

These examples illustrate the practical benefits of agent-first platforms. They enable organizations to scale AI-driven operations while maintaining control, security, and compliance. The shift from deterministic code to agent-driven execution is not just a technical evolution—it is a fundamental rethinking of how software is designed and deployed in the modern enterprise.

**Source:** [Azure Updates](https://azure.microsoft.com/en-us/blog/designing-agent-first-platforms-what-changes-when-agents-do-the-work/?ref=stirup.si). Analysis based on publicly available information; no hands-on testing is claimed.