Agustin Huerta discusses the new Anthropic Code Review feature and the importance of AI governance.
As more and more organizations and professionals use such technology make coding easierthey may also present additional risks, as the speed at which code can now be executed can lead to poor security practices and malicious behavior.
In March, US AI and research firm Anthropic introduced Code Review, a new feature designed to catch and eliminate bugs before they make it into the software codebase. The move Globant’s senior vice president of digital innovation, Agustin Huerta explained that it shows “a change in the evolution of software development work as AI tools increasingly start to own more of the software development life cycle”.
He told SiliconRepublic.com, “It uses multiple special agents to review code for vulnerabilities and bugs, check between each other and prioritize the most relevant issues for reviewers.”
But he noted, while this helps teams better manage large amounts of code, it doesn’t replace human reviewers and raises a few concerns when it comes to long-term security and best performance.
A key coding concern?
“The concern is not that the code can write and update itself, but that organizations may think that less monitoring is needed,” said Huerta, who elaborated, saying that in fact the same principles that control and regulate traditional software development remain equally important when AI agents are involved, if not more so.
“Processes and workflow frameworks that once managed human code must be adapted to manage agents, including workflow integration, human review, data readiness and visualization. Teams need clear visibility into how code is created, reviewed and improved across locations, and defined test points to ensure results.”
He said, although agents can perform several tasks, for example, help, recommend and even carry out instructions within a set of defined guidelines, code quality and risk management should always be the responsibility of people who follow a clear process.
You find that these days, more and more organizations are choosing to outsource tasks, like this one debugging and coding to AI agents, instead of a real worker, which increases the likelihood of a risk, although not the only one Concepts of AI and mistakes sneak past the automated staff.
“The biggest concern is overreliance and unchecked trust in agent autonomy. Overreliance on agent-driven work without proper checks and balances can create blind spots and escalate small problems into bigger problems, such as system outages or security vulnerabilities.
“For example, version control systems and code collections are a way to maintain visibility over human-written code, supported by systematic review processes. If this workflow is automated without adding an additional layer of human oversight, organizations risk compounding errors and introducing major structural problems that are difficult to detect or resolve.”
He finds that, although human involvement is irreplaceable, equally important, throughout the development life cycle, is organizational transparency. “Organizations need visibility into how agents access data, how they think and why tasks are considered complete. This level of visibility is critical to managing the workflow of human agents, identifying areas for growth and maintaining accountability.”
Moreover, when used and directed correctly there are clear and significant benefits.
Business AI
AI agents undoubtedly bring something new to work, better or worse, but there are tangible benefits, such as the ability to increase productivity, reduce difficult work, complex data operations, support engineers in the coding process and identify problems or patterns that are often overlooked by humans.
Huerta said, “By taking over repetitive work that was previously handled by humans, agents allow teams to focus on high-value tasks and activities. These benefits are best seen when AI is used as an enhancement, not a substitute, for human judgment.”
“The most successful models are hybrids of human-agent teams, where the speed and scale of AI is combined with human observation to refine and improve workflows, instead of just automating them.”
A key challenge going forward, he explains, will be in establishing a balance between the adoption and implementation of AI agents and integrating them seamlessly. proper use. He said, as agents become more advanced and skilled, organizations are at risk of losing sight of basic best practices in key areas such as those that manage software development.
“Leaders must continue to prioritize visualization, management and collaboration despite the pressure to prove ROI on AI initiatives.”
Don’t miss out on the information you need to succeed. Sign up for Daily BriefSilicon Republic’s digest of must-know sci-tech news.
