Latest news with #codingagents
Forbes
2 days ago
- Business
- Forbes
AI Coding Agents: Driving The Next Evolution In Software Development
Vikas Mendhe is a solution architect and digital transformation expert specializing in API-driven solutions in financial technology. As artificial intelligence continues to reshape industries, one of the most significant innovations in the software world is the rise of coding agents. They are reshaping how code is written, tested and maintained, marking a new era in software development. What Are Coding Agents? Coding agents are intelligent systems powered by large language models that write, debug and optimize code. They generate APIs, refactor legacy systems, write tests and even build apps with minimal input. Popular tools include GitHub Copilot, Amazon CodeWhisperer and Tabnine. New-generation assistants such as Cursor, Windsurf (recently acquired by OpenAI) and Cline focus on deeper IDE integration, context retention and developer autonomy. Industry Adoption The adoption of coding agents is gaining momentum across sectors. Tech companies are embedding coding assistants into their workflows, startups are exploring autonomous agents like AutoGPT and Devin for rapid prototyping and governments are integrating them cautiously for tasks like data transformation, compliance automation and internal tool development. While accuracy and oversight concerns remain, the shift toward AI-assisted development is well underway. Language-Specific Strengths Of Popular Coding Agents As coding agents continue to evolve, developers often look for tools that best support the languages they work in. • GitHub Copilot thrives in Python, JavaScript and TypeScript, with robust IDE integration. • Amazon CodeWhisperer specializes in Java, Python and JavaScript, featuring AWS-native tools and cloud focus. • Cursor excels in TypeScript and Python, with built-in memory and pair programming. • Tabnine supports Java, Python, C++ and Go with offline capability and customization. • Claude Code optimizes Shell, Python and Bash for terminal-based tasks. • Devin, a Python-based agent, enables complex, multi-step, end-to-end coding automation. Real-World Case Studies Let's just take a look at GitHub Copilot's applications in the real world. ANZ Bank's 2024 trial of GitHub Copilot showed engineers completing tasks 42% faster with improved code quality. Accenture's enterprise study found Copilot users coding 55% faster, with 90% reporting higher fulfillment. And a 2025 ZoomInfo case study involving over 400 developers reported a 33% code acceptance rate and 72% satisfaction. These findings show coding agents reduce repetitive work and free developers for higher-value tasks. Impact On Software Development Coding agents could transform software development from end to end. For developers, they act as smart copilots, automating repetitive tasks and simplifying complex workflows. Businesses gain faster delivery, lower costs and greater agility, turning ideas into prototypes in days instead of weeks. These tools also democratize development: Non-coders can build apps using natural language, and junior developers can produce better code with minimal oversight. Educational studies confirm this potential. AI code completion tools enhance student productivity and engagement while preserving problem-solving and conceptual learning. Programs such as the Stanford Institute for Human-Centered AI are exploring how such tools support computer science education at scale. Behind The Scenes Of Coding Agents Most coding agents are built on transformer-based LLMs such as OpenAI's Codex and GPT-4. Popular tools like GitHub Copilot and Amazon CodeWhisperer operate through IDE plugins, sending prompts to remote model APIs. GPT-4o mini supports a 128K token context window, enabling broader file-level reasoning. Claude 3.7 Sonnet offers 200K tokens for extended reasoning workflows. Gemini 1.5 Pro surpasses both with a 2M token context, ideal for workflows spanning entire codebases. More autonomous agents, such as AutoGPT and Devin, use frameworks like LangChain to chain prompts, memory and shell commands, completing multi-step engineering tasks with minimal human input. Terminal-Based Coding Agents In parallel, new terminal-based coding agents are emerging to support command-line workflows for professional developers. Tools like Claude Code, Codex CLI and Gemini CLI bring AI-powered development directly into the terminal environment, enabling agents to execute commands, write scripts and interact with live file systems, all while preserving developer autonomy. Coding Agents As A Service Despite advances, coding agents can still produce insecure or low-quality code. Safeguards like validation mechanisms and inline linting help, but human oversight remains essential. Rigorous testing, linting and code reviews should be part of every deployment pipeline. Code Quality, Security And The Role Of Supervision Despite advances, coding agents still generate insecure code and lack deep understanding of intent. Recent advancements have introduced better safeguards, validation mechanisms and inline linting. However, ongoing oversight remains essential. This underscores the need for rigorous testing, linting and human code review pipelines before production deployment. Getting Started With AI Coding Agents Before adopting AI coding agents, focus on clear, high-value use cases and choose tools suited to those needs instead of automating everything. Keep humans in the loop by ensuring AI-generated code undergoes rigorous testing, security scans and peer reviews. Research shows nearly half of developers don't fully trust AI output and often spend extra time debugging it. Be mindful of data privacy, intellectual property and licensing rules to avoid compliance issues, and set governance policies to prevent security blind spots and vendor lock-in. To mitigate common pitfalls—such as inaccurate code, scope creep, security risks and hidden costs—start with structured pilot programs that have measurable outcomes. Enterprise case studies show that successful rollouts often begin with controlled experiments, formal risk assessments and well-defined change management plans. Strong guardrails, clear policies and an ongoing review process help organizations capture productivity gains while maintaining quality and security. Conclusion Coding agents are not meant to replace human developers—they are tools that help make their work faster and easier. As more companies start using them, it's important to find the right mix between automation and human control. When used responsibly, coding agents can help teams work more efficiently, come up with new ideas and change the way software is built in the AI era. Everyone, not just developers, should understand what coding agents can and can't do, especially those shaping the future. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Yahoo
01-08-2025
- Business
- Yahoo
Andy Jassy says Amazon has chosen to 'embrace' AI, promising it ‘will make all our teammates' jobs more enjoyable'
Andy Jassy, who sent shockwaves through the jobs market as one of the first major chief executives to say that 'AI will mean fewer jobs,' sounded a different tone on the earnings call accompanying Amazon's earnings on July 31. He reiterated his view that artificial intelligence (AI) will be a transformative force, saying it 'is going to change very substantially the way we work' and emphasizing sweeping impacts already under way. It's changing the way Amazon does coding, finance, all sorts of things, he said: 'really the way we do business process automation, the way we do customer service.' But then he pivoted. Jassy said AI 'will make all our teammates' jobs more enjoyable,' freeing them up from having to do the 'rote' functions that could not previously be automated. Companies have a choice in the AI revolution, he added: they can embrace the change that's happening and help shape the new era, 'or you can wish it away and have it shape you.' He said he has worked to make clear, internally and externally, that Amazon will embrace this moment. 'Much more advanced' While AI's promise and pitfalls have dominated tech headlines for the past two years, Jassy's comments detailed concrete examples of how Amazon is rapidly embedding advanced AI into both its internal workflows and customer-facing services. He highlighted the company's investments in generative AI agents that can assist with—or even independently perform—complex coding tasks. 'Coding agents, having AI do a lot of the coding for us … allows our teammates to start from a much more advanced starting spot,' Jassy explained. This philosophy of combining human creativity with AI-powered efficiency is reshaping other vital departments as well. In research and finance, Jassy described AI tools that can quickly synthesize vast quantities of information or flag anomalies in financial data, freeing up skilled employees for strategic work. Jassy also spotlighted AI's growing influence in Amazon's expansive call center and customer service operations. He pointed to services like AWS Connect—the company's cloud-based call center solution—which now has deep AI integrations for more natural customer interactions and automated issue resolution. Jassy's transformative outlook Jassy has been emphasizing the increasing impact of AI for several months now, for instance suggesting that employees attend AI trainings while promising investors that AI will make them 'very happy' down the road. Amazon had delivered strong earnings earlier on July 31, yet investors sent the stock down roughly 7% in post-market trading with investors concerned about trade headwinds and Amazon's long-term spending plans. Jassy told analysts on the call that, with regard to the impact of tariffs through the first half of the year, 'we haven't yet seen diminishing demand, nor prices meaningfully appreciating.' For this story, Fortune used generative AI to help with an initial draft. An editor verified the accuracy of the information before publishing. This story was originally featured on Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
