Microsoft says these new AI tools will help coders avoid tedious tasks and burnout 

At Microsoft’s Ignite conference on November 18, the company unveiled new AI-powered software features designed to make coders’ lives easier—including a tool to automatically fix security issues as new vulnerabilities are discovered. 

“Over this past year, the nature of being a software engineer has really started to change,” says Amanda Silver, corporate vice president and head of product for apps and agents at Microsoft.  “And our focus has been on tackling the most miserable, soul-draining parts of the job and really transforming them, so that developers can kind of bring joy back to their day-to-day lives.” 

One result of that effort is an AI offering, now in public preview, that combines the runtime application protection of Microsoft Defender for Cloud with GitHub Advanced Security’s protection for source code to spot and help fix a variety of security vulnerabilities. When Microsoft Defender for Cloud detects that an app on the Microsoft Azure cloud system has a security issue, perhaps based on information from a published vulnerability report, that knowledge can be channeled into Microsoft-owned GitHub, to help set up what’s called a security campaign. That’s a GitHub feature designed for a coordinated effort to tackle security holes. Once it’s set up, GitHub’s AI Copilot Autofix tool can automatically suggest code changes to address the issue. 

“The developer doesn’t have to write the code to respond to the issue,” Silver says. “Rather, GitHub Copilot actually issues the pull request, and the developer just has to review and accept it.” 

Even when the problem is caused by a security flaw in third-party code, like an open-source library, Copilot can help in upgrading to a later edition of the library without the bug and help with code changes required for compatibility with the new version.  

The announcement follows the May debut of Azure SRE Agent, another AI tool designed to spot and help analyze certain security issues and other problems, assisting engineers in quickly finding and fixing the causes of incidents affecting cloud systems. It’s one of a number of AI tools recently released by various companies that can flag problems and help engineers comb through the often-voluminous log files generated by applications, operating systems, and other software to understand the root cause, ideally before an issue becomes urgent. 

“Nobody joined the industry because they want to be woken up in the middle of the night because they’re on a live site incident call,” says Silver. 

And for developers building software designed to itself integrate with artificial intelligence to process data or answer user questions, deciding which AI model is best suited for a particular task can be a complex question, particularly when considering factors like speed and cost as well as accuracy. To help address that challenge, Microsoft also on November 18 unveiled what it calls the Model Router in Azure AI Foundry, which automatically dispatches particular AI prompts to an appropriate model in real time as an app runs. Smaller (and cheaper) models can be used when they’ll likely do the trick, while bigger and more costly models can be used for more complex scenarios, with reasoning models invoked for tasks requiring their skills.  

Microsoft has also been working on ways to help businesses upgrade aging code and move older applications to the cloud. It’s another notoriously tedious task that AI programming assistants like GitHub Copilot can help automate.  

Internally, Microsoft reports, units including the Xbox team have used GitHub Copilot to help modernize code, at times dramatically cutting the developer effort required. And a new offering also unveiled November 18, called Managed Instance on Azure App Service, makes it easier for developers to move code to the cloud with fewer tweaks in the first place, by providing closer compatibility with older Microsoft software. 

That, along with AI help in ultimately making further upgrades, should help stave off burnout as developers dodge the tedious tasks of getting aging code to run on today’s cloud systems, Silver says. 

“No engineer really joined the industry to get assigned a months-long refactoring job of doing thankless migration work,” Silver says. “That’s the kind of developer toil that really quietly drains morale, and it burns out great teams and great engineers.” 

View the full article