gettyAI tools are now common in software development, and teams can generate code faster than ever. But faster output doesn’t automatically mean better results, especially when hidden bugs, security gaps, integration issues and maintainability problems also move rapidly through the pipeline.That’s why DevOps discipline—operations and quality teams working together to build, test and improve software—becomes more important, not less, as AI is integrated into the workflow. Below, members of Forbes Technology Council discuss the DevOps practices organizations need to reinforce to turn AI-driven speed into reliable, secure and sustainable software delivery.Pair AI-Assisted Development With Production GuardrailsOrganizations must strengthen engineering excellence through disciplined review and quality gates. AI can speed code creation, but it can also amplify defects, security gaps and inconsistency. The real advantage comes from pairing AI-assisted development with stronger testing, peer review and production guardrails. - Vikas Mittal, Walmart Global TechStrengthen Testing DisciplineAs AI accelerates code generation, the volume and speed of code changes will increase dramatically. However, AI-generated code can introduce subtle bugs, security vulnerabilities and inconsistencies that may not be immediately visible. Without strong testing discipline, organizations risk deploying faster but with lower quality and higher risk. - Kirti Acharya, Anlage DigitalForbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?Mandate Comprehensive Automated TestingComprehensive automated testing is essential, specifically across multiple levels, including unit, integration and security testing. Monitoring is another critical area; with faster deployments and more code in production, visibility and monitoring are vital. - Rajesh Rangarajan, OnTrac AIContinuously Test Data PipelinesI would stress the continuous testing of data pipelines, not just application code. AI-generated code can write a function in seconds, but if that function is pulling from ungoverned data sources, you’re scaling bad decisions. - Jacqueline DeStefano-Tangorra, DataOpsCombine Automated Checks With Human ReviewCode review and continuous code integrity checks will be essential. Faster code generation means more PRs, more review fatigue and a greater risk of bugs, security gaps and quality issues slipping through. To keep quality high at speed, teams will need layered automated verification plus structured human review to preserve security, reliability and architectural consistency. - Itamar Friedman, Qodo.aiInvest In Agentic Application SecurityThe news is all about what AI might break tomorrow. Anthropic’s Glasswing is a pioneering project to address that. AI models make it faster and cheaper to exploit the known vulnerabilities that exist in production today, and risk will only increase as AI-generated code proliferates. This all makes modern, agentic application security an urgent and essential part of your DevSecOps stack. - Sandeep Johri, CheckmarxPreserve End-User TestingAs AI generates more code and produces updates more quickly, I find that we have to ensure the DevOps process enforces end-user testing of changes to understand the impact on their workflows. Many of the changes introduced may be functionally better, but we can’t forget the impact they can have on human operators. - David Van Ronk, Bridgehead ITProtect Long-Term Maintainability Through ReviewWhen output increases, review matters more. The point is no longer to catch small mistakes but to protect architecture, boundaries and long-term maintainability. In faster development environments, code review is where teams make sure short-term speed does not create long-term instability. - Daniel Gumucio, AssureSoftTrack Code ProvenanceAs AI accelerates code generation, DevOps must strengthen code provenance. Teams need reliable ways to track what’s human‑written, what’s AI‑generated and what’s mixed so they can manage risk, validate behavior and spot vulnerabilities. It’s the same discipline they apply to good data management. - Carl D’Halluin, DatadobiReview Code For IntentCode review is essential—not for syntax, but for intent. AI-generated code can be correct line by line and still be wrong architecturally. The risk isn’t bugs; it’s systems that work today and become unmaintainable tomorrow because no human actually understands what was built. As AI writes more code, the engineering team’s job shifts from writing to judgment, and judgment requires review. - Charles Yeomans, AtombeamReinforce Security, Telemetry And Stakeholder Approval With AI, we must focus on three primary things. First, security—one of the most important aspects; second, telemetry—to ensure we are getting it right; and third, gates—because at each layer, even though we are using AI, we must still have all stakeholders’ approval to cover both direct and indirect dependencies. - Nitin Mukhi, Coforge INC.Double Down On CI/CDA strong pipeline—tests, linting, security checks and deploy gates—catches most issues that arise from AI-generated code. Pair this with lightweight human review for edge cases and QA. Much of the review can be standardized and automated, but human judgment still matters. Over time, even QA becomes increasingly automatable. - John Jeong, CharStrengthen Automated Test PipelinesAutomated testing is essential. AI removes the coding bottleneck, but without strong test pipelines, it just moves the bottleneck downstream to manual review. More code volume means more risk, more edge cases and more room for subtle errors AI won’t catch itself. Strengthening automated testing is how teams actually ship faster without sacrificing quality or security. - Mateusz Mucha, Omni CalculatorDeepen Observability PracticesOrganizations will need to strengthen observability. As AI generates more code, complexity and unpredictability increase. Strong monitoring, logging and tracing ensure teams can quickly detect, understand and fix issues, maintaining reliability, performance and trust in rapidly evolving systems. - Dax Grant, Global TransformValidate Code Before Scaling VelocityWe must strengthen code review and automated testing, not sideline them. AI accelerates output, but not accountability. Without rigorous validation—CI pipelines, tests and peer review—you scale bugs and vulnerabilities just as fast. Speed without control compounds risk; disciplined validation is what turns velocity into reliable delivery. - Faustino Júnior, First.DoctorEnforce CI/CD Testing GatesStrengthen automated testing and validation (CI/CD gates). AI can generate code fast, but its outputs are not always correct or secure. Robust unit, integration and security testing ensures quality, prevents vulnerabilities and maintains trust before code reaches production. - Sanjoy Sarkar, First Citizens BankUse Observability To Govern Rapid ChangeStrengthen observability-driven DevOps. As AI accelerates code generation, hidden defects, drift and security gaps can scale just as fast. Deep telemetry, traceability and feedback loops ensure rapid detection, validation and governance, turning speed into reliable, production-grade outcomes rather than amplified risk. - Pabitra Saikia, Truist BankMake Rollback Rehearsal A Standing PracticeStrengthen rollback rehearsal as a standing practice rather than crisis response alone. As AI accelerates deployment frequency, reversing a bad release cleanly becomes more valuable than deploying quickly. Organizations that practice rollback only when something breaks discover the procedure itself is broken too. Regular drills turn recovery from an emergency skill into a reliable operational muscle. - Jagadish Gokavarapu, Wissen InfotechStrengthen Version Control Through Prompt CaptureAs AI accelerates coding, the real asset becomes the prompt, not just the code. Organizations should capture and manage prompts from business users, analysts, architects and developers alongside code in Git, ensuring transparency, reuse and alignment with intent. - Ann Blakely, Baker Tilly