Insights in Foundry analyzes “traces” to determine likely causes of mistakes by AI agents and recommend ways to overcome them.
software development
As it takes on work ranging from coding to running productivity apps to combining existing AI experiences, Copilot emerges as the driver of an expanding range of business activities.
AI isn’t killing software; instead it’s filling gaps between commercial systems. The more products you bought, the more gaps you had.
New functions can support apps that combine Dynamics 365 apps with business context from WorkIQ for complete experiences, Microsoft partners say.
New features enhance positioning of Teams as a platform to support humans, agents working together to accelerate GitHub-powered dev projects.
Agent Optimizer functions ensure Foundry Hosted Agents adhere to user-defined criteria for the best instructions, skills, models, and tool descriptions.
NVIDIA-powered ‘Dev Box’ delivers horsepower to execute massive AI models from local systems on an ‘unmetered’ basis, while reference designs aim to make AI computing ubiquitous.
The bifurcation between AI-native and AI-bolted-on organizations is becoming more apparent as the AI-centric approach’s benefits come into sharper focus.
Microsoft provides three distinct patterns — orchestrator/specialist, MCP-based connected agents, and browser session use — for agent development; discover recommended use cases for each.
Microsoft and AI expert sorts out key pricing questions and licensing considerations for Copilot Studio-built agents, and explains why costs still pose confusion for some customers.
Brent Wodicka, CTO of AIS, explains how ISVs and corporate software developers need to build ‘agent-friendly pathways’ to software platforms to enable autonomous actions.
‘Routines’ are used to configure automations of common processes, then store and invoke them when needed to further accelerate development processes.
Plugin for popular coding platforms injects the ‘skills’ needed to deliver agentic automation without deep knowledge of an organization’s Dataverse environment.
Initial release supports actions across agents, multiple AI models, and interoperability through the two most widely embraced interoperability standards.
Dona Sarkar calls out the “tech bro”-driven AI hype machine and shares a blueprint to push AI into mainstream use cases that will deliver on the tech’s considerable promise.
Framework lets users and developers bring their own models, supports both A2A and MCP for seamless integration of diverse AI tools, Will Hawkins and Rich Hawkins note.
One example workflow is a daily status report for developers that maintain repositories; the coding agent will interact with developers to confirm their specific needs and intent, check validity.
Failure becomes cheaper, and learning happens faster, as teams utilize AI to generate and test variations before committing to engineering.
Making Copilot Studio available within Visual Studio Code is a big step forward, integrating AI agents directly into developers’ daily workflows.
Paul Swider of RealActivity said demos highlighting Microsoft’s work with NVIDIA gave important insight into AI agent and product development opportunities.