[{"data":1,"prerenderedAt":649},["ShallowReactive",2],{"/en-us/the-source/authors/emilio-salvador/":3,"footer-en-us":34,"the-source-banner-en-us":341,"the-source-navigation-en-us":353,"the-source-newsletter-en-us":381,"emilio-salvador-articles-list-authors-en-us":392,"emilio-salvador-articles-list-en-us":423,"emilio-salvador-page-categories-en-us":648},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"config":8,"seo":10,"content":12,"type":26,"slug":27,"_id":28,"_type":29,"title":11,"_source":30,"_file":31,"_stem":32,"_extension":33},"/en-us/the-source/authors/emilio-salvador","authors",false,"",{"layout":9},"the-source",{"title":11},"Emilio Salvador",[13,24],{"componentName":14,"type":14,"componentContent":15},"TheSourceAuthorHero",{"config":16,"name":11,"role":19,"bio":20,"headshot":21},{"gitlabHandle":17,"linkedInProfileUrl":18},"esalvadorp","https://www.linkedin.com/in/emiliosp/","Vice President, Strategy and Developer Relations","Emilio Salvador is vice president of strategy and developer relations at GitLab. A technology executive with more than 20 years of experience, Emilio has held roles at Amazon and Microsoft, and most recently led strategy and operations for the Developer Advocacy and Experience team at Google. He holds an MBA from MIT Sloan School of Management.",{"altText":11,"config":22},{"src":23},"https://res.cloudinary.com/about-gitlab-com/image/upload/v1751463402/g0snp3uisjm4tj3pkqcw.jpg",{"componentName":25,"type":25},"TheSourceArticlesList","author","emilio-salvador","content:en-us:the-source:authors:emilio-salvador.yml","yaml","content","en-us/the-source/authors/emilio-salvador.yml","en-us/the-source/authors/emilio-salvador","yml",{"_path":35,"_dir":36,"_draft":6,"_partial":6,"_locale":7,"data":37,"_id":337,"_type":29,"title":338,"_source":30,"_file":339,"_stem":340,"_extension":33},"/shared/en-us/main-footer","en-us",{"text":38,"source":39,"edit":45,"contribute":50,"config":55,"items":60,"minimal":329},"Git is a trademark of Software Freedom Conservancy and our use of 'GitLab' is under license",{"text":40,"config":41},"View page source",{"href":42,"dataGaName":43,"dataGaLocation":44},"https://gitlab.com/gitlab-com/marketing/digital-experience/about-gitlab-com/","page source","footer",{"text":46,"config":47},"Edit this page",{"href":48,"dataGaName":49,"dataGaLocation":44},"https://gitlab.com/gitlab-com/marketing/digital-experience/about-gitlab-com/-/blob/main/content/","web ide",{"text":51,"config":52},"Please contribute",{"href":53,"dataGaName":54,"dataGaLocation":44},"https://gitlab.com/gitlab-com/marketing/digital-experience/about-gitlab-com/-/blob/main/CONTRIBUTING.md/","please contribute",{"twitter":56,"facebook":57,"youtube":58,"linkedin":59},"https://twitter.com/gitlab","https://www.facebook.com/gitlab","https://www.youtube.com/channel/UCnMGQ8QHMAnVIsI3xJrihhg","https://www.linkedin.com/company/gitlab-com",[61,88,160,228,290],{"title":62,"links":63,"subMenu":69},"Platform",[64],{"text":65,"config":66},"DevSecOps platform",{"href":67,"dataGaName":68,"dataGaLocation":44},"/platform/","devsecops platform",[70],{"title":71,"links":72},"Pricing",[73,78,83],{"text":74,"config":75},"View plans",{"href":76,"dataGaName":77,"dataGaLocation":44},"/pricing/","view plans",{"text":79,"config":80},"Why Premium?",{"href":81,"dataGaName":82,"dataGaLocation":44},"/pricing/premium/","why premium",{"text":84,"config":85},"Why Ultimate?",{"href":86,"dataGaName":87,"dataGaLocation":44},"/pricing/ultimate/","why ultimate",{"title":89,"links":90},"Solutions",[91,96,100,105,110,115,120,125,130,135,140,145,150,155],{"text":92,"config":93},"Digital transformation",{"href":94,"dataGaName":95,"dataGaLocation":44},"/topics/digital-transformation/","digital transformation",{"text":97,"config":98},"Application Security Testing",{"href":99,"dataGaName":97,"dataGaLocation":44},"/solutions/application-security-testing/",{"text":101,"config":102},"Automated software delivery",{"href":103,"dataGaName":104,"dataGaLocation":44},"/solutions/delivery-automation/","automated software delivery",{"text":106,"config":107},"Agile development",{"href":108,"dataGaName":109,"dataGaLocation":44},"/solutions/agile-delivery/","agile delivery",{"text":111,"config":112},"Cloud transformation",{"href":113,"dataGaName":114,"dataGaLocation":44},"/topics/cloud-native/","cloud transformation",{"text":116,"config":117},"SCM",{"href":118,"dataGaName":119,"dataGaLocation":44},"/solutions/source-code-management/","source code management",{"text":121,"config":122},"CI/CD",{"href":123,"dataGaName":124,"dataGaLocation":44},"/solutions/continuous-integration/","continuous integration & delivery",{"text":126,"config":127},"Value stream management",{"href":128,"dataGaName":129,"dataGaLocation":44},"/solutions/value-stream-management/","value stream management",{"text":131,"config":132},"GitOps",{"href":133,"dataGaName":134,"dataGaLocation":44},"/solutions/gitops/","gitops",{"text":136,"config":137},"Enterprise",{"href":138,"dataGaName":139,"dataGaLocation":44},"/enterprise/","enterprise",{"text":141,"config":142},"Small business",{"href":143,"dataGaName":144,"dataGaLocation":44},"/small-business/","small business",{"text":146,"config":147},"Public sector",{"href":148,"dataGaName":149,"dataGaLocation":44},"/solutions/public-sector/","public sector",{"text":151,"config":152},"Education",{"href":153,"dataGaName":154,"dataGaLocation":44},"/solutions/education/","education",{"text":156,"config":157},"Financial services",{"href":158,"dataGaName":159,"dataGaLocation":44},"/solutions/finance/","financial services",{"title":161,"links":162},"Resources",[163,168,173,178,183,188,193,198,203,208,213,218,223],{"text":164,"config":165},"Install",{"href":166,"dataGaName":167,"dataGaLocation":44},"/install/","install",{"text":169,"config":170},"Quick start guides",{"href":171,"dataGaName":172,"dataGaLocation":44},"/get-started/","quick setup checklists",{"text":174,"config":175},"Learn",{"href":176,"dataGaName":177,"dataGaLocation":44},"https://university.gitlab.com/","learn",{"text":179,"config":180},"Product documentation",{"href":181,"dataGaName":182,"dataGaLocation":44},"https://docs.gitlab.com/","docs",{"text":184,"config":185},"Blog",{"href":186,"dataGaName":187,"dataGaLocation":44},"/blog/","blog",{"text":189,"config":190},"Customer success stories",{"href":191,"dataGaName":192,"dataGaLocation":44},"/customers/","customer success stories",{"text":194,"config":195},"Remote",{"href":196,"dataGaName":197,"dataGaLocation":44},"https://handbook.gitlab.com/handbook/company/culture/all-remote/","remote",{"text":199,"config":200},"GitLab Services",{"href":201,"dataGaName":202,"dataGaLocation":44},"/services/","services",{"text":204,"config":205},"TeamOps",{"href":206,"dataGaName":207,"dataGaLocation":44},"/teamops/","teamops",{"text":209,"config":210},"Community",{"href":211,"dataGaName":212,"dataGaLocation":44},"/community/","community",{"text":214,"config":215},"Forum",{"href":216,"dataGaName":217,"dataGaLocation":44},"https://forum.gitlab.com/","forum",{"text":219,"config":220},"Events",{"href":221,"dataGaName":222,"dataGaLocation":44},"/events/","events",{"text":224,"config":225},"Partners",{"href":226,"dataGaName":227,"dataGaLocation":44},"/partners/","partners",{"title":229,"links":230},"Company",[231,236,241,246,251,256,261,265,270,275,280,285],{"text":232,"config":233},"About",{"href":234,"dataGaName":235,"dataGaLocation":44},"/company/","company",{"text":237,"config":238},"Jobs",{"href":239,"dataGaName":240,"dataGaLocation":44},"/jobs/","jobs",{"text":242,"config":243},"Leadership",{"href":244,"dataGaName":245,"dataGaLocation":44},"/company/team/e-group/","leadership",{"text":247,"config":248},"Team",{"href":249,"dataGaName":250,"dataGaLocation":44},"/company/team/","team",{"text":252,"config":253},"Handbook",{"href":254,"dataGaName":255,"dataGaLocation":44},"https://handbook.gitlab.com/","handbook",{"text":257,"config":258},"Investor relations",{"href":259,"dataGaName":260,"dataGaLocation":44},"https://ir.gitlab.com/","investor relations",{"text":262,"config":263},"Sustainability",{"href":264,"dataGaName":262,"dataGaLocation":44},"/sustainability/",{"text":266,"config":267},"Diversity, inclusion and belonging (DIB)",{"href":268,"dataGaName":269,"dataGaLocation":44},"/diversity-inclusion-belonging/","Diversity, inclusion and belonging",{"text":271,"config":272},"Trust Center",{"href":273,"dataGaName":274,"dataGaLocation":44},"/security/","trust center",{"text":276,"config":277},"Newsletter",{"href":278,"dataGaName":279,"dataGaLocation":44},"/company/contact/","newsletter",{"text":281,"config":282},"Press",{"href":283,"dataGaName":284,"dataGaLocation":44},"/press/","press",{"text":286,"config":287},"Modern Slavery Transparency Statement",{"href":288,"dataGaName":289,"dataGaLocation":44},"https://handbook.gitlab.com/handbook/legal/modern-slavery-act-transparency-statement/","modern slavery transparency statement",{"title":291,"links":292},"Contact Us",[293,298,303,308,313,318,323],{"text":294,"config":295},"Contact an expert",{"href":296,"dataGaName":297,"dataGaLocation":44},"/sales/","sales",{"text":299,"config":300},"Get help",{"href":301,"dataGaName":302,"dataGaLocation":44},"/support/","get help",{"text":304,"config":305},"Customer portal",{"href":306,"dataGaName":307,"dataGaLocation":44},"https://customers.gitlab.com/customers/sign_in/","customer portal",{"text":309,"config":310},"Status",{"href":311,"dataGaName":312,"dataGaLocation":44},"https://status.gitlab.com/","status",{"text":314,"config":315},"Terms of use",{"href":316,"dataGaName":317,"dataGaLocation":44},"/terms/","terms of use",{"text":319,"config":320},"Privacy statement",{"href":321,"dataGaName":322,"dataGaLocation":44},"/privacy/","privacy statement",{"text":324,"config":325},"Cookie preferences",{"dataGaName":326,"dataGaLocation":44,"id":327,"isOneTrustButton":328},"cookie preferences","ot-sdk-btn",true,{"items":330},[331,333,335],{"text":314,"config":332},{"href":316,"dataGaName":317,"dataGaLocation":44},{"text":319,"config":334},{"href":321,"dataGaName":322,"dataGaLocation":44},{"text":324,"config":336},{"dataGaName":326,"dataGaLocation":44,"id":327,"isOneTrustButton":328},"content:shared:en-us:main-footer.yml","Main Footer","shared/en-us/main-footer.yml","shared/en-us/main-footer",{"_path":342,"_dir":343,"_draft":6,"_partial":6,"_locale":7,"visibility":328,"id":344,"title":345,"button":346,"_id":350,"_type":29,"_source":30,"_file":351,"_stem":352,"_extension":33},"/shared/en-us/the-source/banner/the-economics-of-software-innovation-2025-08-18","banner","The Economics of Software Innovation","The Economics of Software Innovation—AI’s $750 Billion Opportunity",{"config":347,"text":349},{"href":348},"https://about.gitlab.com/software-innovation-report/","Get the research report","content:shared:en-us:the-source:banner:the-economics-of-software-innovation-2025-08-18.yml","shared/en-us/the-source/banner/the-economics-of-software-innovation-2025-08-18.yml","shared/en-us/the-source/banner/the-economics-of-software-innovation-2025-08-18",{"_path":354,"_dir":9,"_draft":6,"_partial":6,"_locale":7,"logo":355,"subscribeLink":360,"navItems":364,"_id":377,"_type":29,"title":378,"_source":30,"_file":379,"_stem":380,"_extension":33},"/shared/en-us/the-source/navigation",{"altText":356,"config":357},"the source logo",{"src":358,"href":359},"https://res.cloudinary.com/about-gitlab-com/image/upload/v1750191004/t7wz1klfb2kxkezksv9t.svg","/the-source/",{"text":361,"config":362},"Subscribe",{"href":363},"#subscribe",[365,369,373],{"text":366,"config":367},"Artificial Intelligence",{"href":368},"/the-source/ai/",{"text":370,"config":371},"Security & Compliance",{"href":372},"/the-source/security/",{"text":374,"config":375},"Platform & Infrastructure",{"href":376},"/the-source/platform/","content:shared:en-us:the-source:navigation.yml","Navigation","shared/en-us/the-source/navigation.yml","shared/en-us/the-source/navigation",{"_path":382,"_dir":9,"_draft":6,"_partial":6,"_locale":7,"title":383,"description":384,"submitMessage":385,"formData":386,"_id":389,"_type":29,"_source":30,"_file":390,"_stem":391,"_extension":33},"/shared/en-us/the-source/newsletter","The Source Newsletter","Stay updated with insights for the future of software development.","You have successfully signed up for The Source’s newsletter.",{"config":387},{"formId":388,"formName":279,"hideRequiredLabel":328},1077,"content:shared:en-us:the-source:newsletter.yml","shared/en-us/the-source/newsletter.yml","shared/en-us/the-source/newsletter",{"amanda-rueda":393,"andre-michael-braun":394,"andrew-haschka":395,"ayoub-fandi":396,"bob-stevens":397,"brian-wald":398,"bryan-ross":399,"chandler-gibbons":400,"dave-steer":401,"ddesanto":402,"derek-debellis":403,"emilio-salvador":11,"erika-feldman":404,"george-kichukov":405,"gitlab":406,"grant-hickman":407,"haim-snir":408,"iganbaruch":409,"jlongo":410,"joel-krooswyk":411,"josh-lemos":412,"julie-griffin":413,"kristina-weis":414,"lee-faus":415,"ncregan":416,"rschulman":417,"sabrina-farmer":418,"sandra-gittlen":419,"sharon-gaudin":420,"stephen-walters":421,"taylor-mccaslin":422},"Amanda Rueda","Andre Michael Braun","Andrew Haschka","Ayoub Fandi","Bob Stevens","Brian Wald","Bryan Ross","Chandler Gibbons","Dave Steer","David DeSanto","Derek DeBellis","Erika Feldman","George Kichukov","GitLab","Grant Hickman","Haim Snir","Itzik Gan Baruch","Joseph Longo","Joel Krooswyk","Josh Lemos","Julie Griffin","Kristina Weis","Lee Faus","Niall Cregan","Robin Schulman","Sabrina Farmer","Sandra Gittlen","Sharon Gaudin","Stephen Walters","Taylor McCaslin",{"allArticles":424,"visibleArticles":647,"showAllBtn":6},[425,467,503,544,579,614],{"_path":426,"_dir":427,"_draft":6,"_partial":6,"_locale":7,"slug":428,"type":429,"category":427,"config":430,"seo":434,"content":439,"_id":464,"_type":29,"title":440,"_source":30,"_file":465,"_stem":466,"_extension":33,"description":437,"date":441,"timeToRead":442,"heroImage":438,"keyTakeaways":443,"articleBody":447,"faq":448},"/en-us/the-source/ai/to-maximize-the-750b-ai-opportunity-human-innovation-is-key","ai","to-maximize-the-750b-ai-opportunity-human-innovation-is-key","article",{"layout":9,"template":431,"featured":328,"articleType":432,"author":27,"gatedAsset":433,"isHighlighted":6,"authorName":11},"TheSourceArticle","Regular","software-innovation-report-2025",{"config":435,"title":436,"ogTitle":436,"description":437,"ogDescription":437,"ogImage":438},{"noIndex":6},"Maximize the $750B AI opportunity with human innovation","Discover how human-AI partnerships can save tens of thousands of dollars per developer annually while boosting innovation and productivity across your team.","https://res.cloudinary.com/about-gitlab-com/image/upload/v1754661325/ntf0xsctetcx7uq1yfpy.png",{"title":440,"description":437,"date":441,"timeToRead":442,"heroImage":438,"keyTakeaways":443,"articleBody":447,"faq":448},"To maximize the $750B AI opportunity, human innovation is key","2025-08-12","5 min read",[444,445,446],"Companies save an average of $28,249 per developer annually through strategic AI investments, creating $750 billion in global value potential.","The majority of executives want 50/50 human-AI partnerships, but teams currently operate at 75% human, 25% AI — revealing untapped efficiency opportunities.","Technical leaders who master AI orchestration and quality governance will drive the next wave of software innovation and competitive advantage.","Imagine that your team is tasked with an enormous challenge: Due to intensifying customer demands, you need to double your feature delivery velocity in 90 days without increasing the team’s size. The budget is frozen, hiring is on hold, but the business timeline is non-negotiable. The answer isn’t working nights and weekends. It’s fundamentally changing how your developers collaborate with AI to architect *solutions* instead of just writing code.\n\nWhile studying classical languages years ago, I discovered that Latin demanded architectural thinking — seeing how complex systems interconnect, understanding cross-layer dependencies, and building logical structures that withstand analysis. Every sentence was an interconnected puzzle where changing one component affected everything else.\n\nThat linguistic training unexpectedly prepared me for today’s software development revolution. AI is fundamentally changing how we build software, and engineering leaders who understand architectural approaches to human-AI collaboration will shape the next era of innovation.\n\nThe financial impact is substantial. [GitLab’s recent C-Suite survey of thousands of executives globally](https://about.gitlab.com/software-innovation-report/) shows that AI-enhanced software innovation delivers exceptional returns: Organizations save an average of $28,249 per developer annually from AI investments. Applied to the world’s 27 million developers, this represents over $750 billion in potential global value, with $149 billion in the U.S. market alone.\n\nHowever, the survey data also reveal a striking disconnect: While 73% of executives believe the ideal human-AI partnership should be 50/50, in reality, humans manage three-quarters of development work, with AI contributing just one-quarter. This gap represents a massive unclaimed competitive advantage.\n\n## Why architectural thinking matters for AI collaboration\n\nThe engineering leaders succeeding in this transformation aren’t simply writing better prompts. They’re cognitive architects who break down business challenges into core principles and design “thought blueprints” that guide AI systems toward precise solutions.\n\nThis architectural mindset mirrors the systematic approach I learned analyzing Latin texts. You need to grasp the underlying structure before manipulating surface elements. With 89% of executives expecting agentic AI to become standard practice within three years, engineering leaders who think systematically about human-AI workflows will be essential.\n\nThe C-suite is taking notice: 91% of executives say software innovation is a core business priority, and 58% have experienced business growth from innovation efforts in the past year. With leaders estimating increases of 44% in revenue and 48% in developer productivity from AI adoption, this isn't a distant future — it’s today’s competitive advantage.\n\n## Strategic moves for engineering teams\n\nThree key strategies can position your engineering organization for optimal 50/50 AI-human collaboration:\n\n**Build AI communication and context management capabilities.** Effective AI collaboration isn’t about crafting perfect prompts. It’s about designing process-oriented thinking that guides AI through complex tasks. Focus on developing frameworks for framing problems, providing appropriate context, and structuring interactions with AI. Create workflows aligned with business goals, breaking down complex problems into manageable components that AI can handle efficiently.\n\n**Cultivate system-level thinking across your team.** As AI becomes more capable of code generation, the value of engineering teams shifts from code creation to strategic architecture and design principles. Invest time in defining system-to-subsystem connections, establishing business logic, and building context-rich environments for AI tools. Position your team as software orchestrators, rather than just code writers, through upfront analysis and planning, and then thoroughly review outputs to prevent technical debt.\n\n**Establish quality and security standards for AI.** With 52% of executives identifying cybersecurity threats as their primary concern around agentic AI adoption, engineering teams that can validate AI reasoning processes, implement adversarial testing, and establish specialized review procedures for AI-generated code will be highly valuable to the business. This represents an evolution from traditional debugging to validating AI reasoning and ensuring business logic alignment. It's not enough to consider code security and quality; you’ll also need to ensure that you have [guardrails in place that control the behavior of AI agents](https://about.gitlab.com/the-source/ai/implementing-effective-guardrails-for-ai-agents/) to prevent introducing risks (such as the unintended deletion or alteration of code) into your software development lifecycle.\n\n## The human advantage in an AI-driven world\n\nThe survey data also reveal the criticality of human input in the age of AI: 99% of executives believe human contributions remain valuable for software development. The value of AI isn’t in replacing engineers. It’s about amplifying human capabilities.\n\nExecutives identified creativity and strategic vision as the top two most valued human inputs. This makes sense, since although AI excels at pattern recognition and code generation, the architectural thinking required to understand system interconnections, anticipate dependencies, and design for long-term stability remains distinctly human.\n\nOrganizations that [optimize human-AI partnerships](https://about.gitlab.com/the-source/ai/three-ways-to-operationalize-ai-for-engineering-teams/) today will define tomorrow’s software landscape. Our survey found that 53% of executives have already implemented regulatory-aligned governance measures for agentic AI, and 52% have developed internal AI policies.\n\n## Capturing the opportunity\n\nTransformation is accelerating rapidly. The vast majority (82%) of executives say they are willing to invest over half their annual IT budget in software innovation, and 90% report their organizations have adopted frameworks linking development activities to key business outcomes.\n\nEngineering leaders who embrace 50/50 human-AI partnerships and think architecturally about human-AI collaboration, while maintaining the creative vision and strategic thinking that only humans provide, will drive this transformation forward.\n\nThe $750 billion opportunity represents an opportunity for engineering teams to do better work, solve larger problems, and create unprecedented value. AI is already transforming software development — the question is whether your engineering organization will be ready to lead that transformation. The future belongs to teams that build effective bridges between human creativity and AI capability.",[449,452,455,458,461],{"header":450,"content":451},"How much money can companies save per developer with AI investments?","Organizations save an average of $28,249 per developer annually from AI investments, according to GitLab's C-Suite survey of thousands of executives globally. This substantial financial impact demonstrates the measurable ROI of strategic AI adoption in software development teams.",{"header":453,"content":454},"What is the ideal human-AI partnership ratio according to executives?","73% of executives believe the ideal human-AI partnership should be 50/50, but current reality shows humans managing three-quarters of development work with AI contributing just one-quarter. This gap represents a massive unclaimed competitive advantage for organizations.",{"header":456,"content":457},"What percentage of executives expect agentic AI to become standard practice?","89% of executives expect agentic AI to become standard practice within three years. Additionally, 91% say software innovation is a core business priority, with leaders estimating 44% revenue increases and 48% developer productivity gains from AI adoption.",{"header":459,"content":460},"What are the top human contributions that remain valuable in AI-driven development?","Executives identified creativity and strategic vision as the top two most valued human inputs, with 99% believing human contributions remain valuable for software development. Architectural thinking for system interconnections, dependencies, and long-term stability design remains distinctly human.",{"header":462,"content":463},"How much are executives willing to invest in software innovation initiatives?","82% of executives are willing to invest over half their annual IT budget in software innovation, and 90% report their organizations have adopted frameworks linking development activities to key business outcomes, demonstrating significant commitment to transformation.","content:en-us:the-source:ai:to-maximize-the-750b-ai-opportunity-human-innovation-is-key:index.yml","en-us/the-source/ai/to-maximize-the-750b-ai-opportunity-human-innovation-is-key/index.yml","en-us/the-source/ai/to-maximize-the-750b-ai-opportunity-human-innovation-is-key/index",{"_path":468,"_dir":427,"_draft":6,"_partial":6,"_locale":7,"config":469,"seo":471,"content":476,"type":429,"slug":499,"category":427,"_id":500,"_type":29,"title":472,"_source":30,"_file":501,"_stem":502,"_extension":33,"date":477,"description":473,"timeToRead":442,"heroImage":474,"keyTakeaways":478,"articleBody":482,"faq":483},"/en-us/the-source/ai/from-vibe-coding-to-agentic-ai-a-roadmap-for-technical-leaders",{"layout":9,"template":431,"articleType":432,"author":27,"featured":328,"gatedAsset":470,"isHighlighted":6,"authorName":11},"source-lp-enterprise-guide-to-agentic-ai",{"title":472,"description":473,"ogImage":474,"config":475},"From vibe coding to agentic AI: A roadmap for technical leaders","Discover how to implement vibe coding and agentic AI in your development process to increase productivity while maintaining code quality and security.","https://res.cloudinary.com/about-gitlab-com/image/upload/v1751463655/i6g9scccza0l35n6i1bf.png",{"ignoreTitleCharLimit":328},{"title":472,"date":477,"description":473,"timeToRead":442,"heroImage":474,"keyTakeaways":478,"articleBody":482,"faq":483},"2025-06-12",[479,480,481],"AI-assisted development is transforming software creation, enabling teams to focus on business logic and user experience rather than syntax details, but requires proper governance to ensure quality.","Organizations should adopt an evolutionary approach to AI implementation — starting with basic assistance, then expanding across the development lifecycle, establishing governance frameworks, and gradually introducing autonomous agents.","The engineering landscape is shifting as AI handles routine coding tasks, creating demand for new specialized roles and requiring developers to focus on strategic thinking, architecture design, and effective AI collaboration.","A new wave of generative artificial intelligence (AI) tools is redefining how we build software and who can participate in the process. At the forefront of this revolution is \"vibe coding\" - using natural language prompts to generate functional code without having to fully understand how the code works.\n\n[According to GitLab research](https://about.gitlab.com/developer-survey/), 78% of teams have already integrated AI-assisted coding tools into software development workflows, and AI is demonstrating measurable efficiency improvements. Vibe coding lowers the barriers to entry for development. However, when software engineers use AI-generated code without critical evaluation or deep comprehension, that might also lead to lower quality and increased security vulnerabilities.\n\nTraditional development approaches rely heavily on specific programming languages and syntax rules. Vibe coding lowers the need to fully comprehend the nuances of every language and development pattern, but it does not eliminate that need. This tension between accessibility and quality reflects a broader transformation in software creation.\n\nAI is fundamentally shifting what development means. Team members can focus on desired outcomes rather than implementation details. Logic, business requirements, and user experience precede syntax correctness and language expertise. Organizations increasingly value professionals who can effectively bridge product vision with technical execution - often without writing traditional code.\n\nWhile vibe coding offers tremendous potential to accelerate development and democratize software creation, it must be implemented thoughtfully with proper governance to ensure that speed doesn't come at the expense of quality and maintainability.\n\n## Agentic AI and vibe coding\nVibe coding is about getting something to appear to work quickly rather than building a robust, efficient, and maintainable solution based on solid knowledge. This is where [agentic AI](https://about.gitlab.com/the-source/ai/agentic-ai-unlocking-developer-potential-at-scale/) can help. Agents can take abstract instructions like \"build a customer database\" and autonomously handle all the technical implementation details, bridging the gap between quick prototypes and properly engineered solutions.\n\nWhile vibe coding primarily focuses on code generation through natural language prompts, agentic AI expands these capabilities into an autonomous development ecosystem. Vibe coding involves a human developer using AI without requiring deep understanding, while agentic AI takes on a more proactive, autonomous role in building software based on a given goal.\n\nThe two approaches complement each other perfectly: vibe coding provides a solid foundation for human-AI interaction through natural language, while agentic systems build upon this foundation to create self-directed development partners that handle complex tasks by making independent decisions and taking action with minimal supervision.\n\nAgentic AI systems enhance vibe coding by integrating deeply into development workflows, conducting sophisticated code reviews, recommending infrastructure optimizations, and adapting to changing requirements. [Industry research from Deloitte](https://www2.deloitte.com/us/en/insights/industry/technology/technology-media-and-telecom-predictions/2025/autonomous-generative-ai-agents-still-under-development.html) indicates that 25% of companies using generative AI will implement agentic AI pilots in 2025, which is expected to double by 2027.\n\nSuccessfully implementing vibe coding and agentic AI together requires careful planning. Organizations must establish robust security protocols, ensure regulatory compliance, and create clear communication channels between AI systems and existing tools. Despite these challenges, the combined power of vibe coding and agentic AI delivers significant benefits in development speed, code quality, and resource optimization.\n\n## Implementation strategy for teams and leadership: An evolutionary approach\nDevelopment teams and technical leaders can follow this evolutionary path to effectively implement vibe coding and agentic AI:\n1. **Begin with AI assistance**: Introduce developers to AI tools that improve productivity for routine tasks. Focus on building familiarity, comfort, and confidence with AI assistance for coding, documentation, and simple problem-solving.\n1. **Expand AI assistance across the software development lifecycle**: Move beyond code generation tools to integrate AI into testing, debugging, code review, and documentation. Identify repetitive, time-intensive workflows where AI can create immediate value with minimal disruption.\n1. **Establish governance frameworks & interoperability standards**: Create clear policies for AI tool usage, including data access permissions, security protocols, and quality standards for AI-generated code. Define protocols for how AI systems will share information and collaborate across platforms, as well as the level of human input required when using AI tools.\n1. **Introduce autonomous AI agents for specific tasks**: Deploy agents to handle self-contained development tasks with a degree of autonomy. These agents take abstract goals like \"optimize this database query\" and handle the implementation details independently while maintaining code quality.\n1. **Scale agent implementation across the organization**: Expand the scope of tasks handled by agents and introduce multiple agents working together on complex projects. Integrate agents deeply into the end-to-end software development lifecycle and redesign team structures to create cross-functional groups combining technical expertise and domain knowledge.\n1. **Continuously improve through feedback and education**: Implement systems to monitor agent performance with clear metrics and correction protocols. Invest in organization-wide AI literacy through training programs for prompt engineering, AI collaboration techniques, and effective system oversight.\n\nThis evolutionary approach ensures technical implementation and organizational leadership progress together in the AI transformation journey, maximizing the benefits of vibe coding while building robust, efficient solutions.\n\n## The changing developer landscape\nThe engineering role is evolving as vibe coding and [agentic AI](https://about.gitlab.com/the-source/ai/emerging-agentic-ai-trends-reshaping-software-development/) handle more of the heavy lifting in software development. Less experienced developers face a steeper learning curve with fewer straightforward tasks available for initial skill-building. Simultaneously, senior engineers must adapt as AI takes over more complex tasks and traditional oversight responsibilities.\n\nBeyond the changing dynamics for existing roles, we’re seeing the emergence of entirely new positions like prompt engineers who guide and refine AI outputs. The most valuable engineering skills have shifted toward architecture design, strategic thinking, and effective AI collaboration.\n\nWhile this disruption creates uncertainty for some traditional roles and compensation models, it also opens doors for those who position themselves at the intersection of human creativity and machine efficiency. The most successful engineers will be those who strategically delegate routine work to AI while applying their uniquely human expertise to innovation and complex problem-solving.\n\nFor technical leaders, the strategic implications are clear: organizations that embrace vibe coding and agentic AI gain decisive competitive advantages through accelerated development cycles, improved code quality, and more efficient resource allocation. However, organizations will need to adopt AI responsibly, with governance frameworks to ensure that efficiency doesn’t come at the expense of security. Those who fail to do so may find themselves multiple innovation cycles behind in an increasingly AI-powered development landscape.",[484,487,490,493,496],{"header":485,"content":486},"What is vibe coding and how does it change the development process?","Vibe coding refers to the use of natural language prompts to generate functional code without needing deep knowledge of programming syntax. It shifts the focus from how code is written to what it achieves, allowing developers to prioritize business logic and user experience.",{"header":488,"content":489},"How does agentic AI complement vibe coding?","While vibe coding enables quick code generation, agentic AI takes it further by autonomously executing tasks based on goals. It bridges the gap between rapid prototyping and robust implementation, acting as a self-directed development partner that enhances productivity and quality.",{"header":491,"content":492},"What are the first steps organizations should take to implement AI in development workflows?","Start with AI-assisted tools that handle routine tasks like documentation, test creation, and debugging. As teams grow more confident, expand AI's role across the development lifecycle and introduce governance policies to manage its use responsibly.",{"header":494,"content":495},"How does the rise of AI change the skills developers need?","AI is reshaping engineering roles. Developers now need stronger architectural thinking, system design, and AI collaboration skills. Routine coding tasks are increasingly delegated to AI, so human expertise will focus more on oversight, creativity, and strategic problem-solving.",{"header":497,"content":498},"What are the risks of using AI without governance in software development?","Without governance, AI-generated code may introduce security vulnerabilities, lack maintainability, or violate compliance standards. Organizations must establish security protocols, clear usage policies, and quality checks to ensure responsible AI adoption.","from-vibe-coding-to-agentic-ai-a-roadmap-for-technical-leaders","content:en-us:the-source:ai:from-vibe-coding-to-agentic-ai-a-roadmap-for-technical-leaders:index.yml","en-us/the-source/ai/from-vibe-coding-to-agentic-ai-a-roadmap-for-technical-leaders/index.yml","en-us/the-source/ai/from-vibe-coding-to-agentic-ai-a-roadmap-for-technical-leaders/index",{"_path":504,"_dir":427,"_draft":6,"_partial":6,"_locale":7,"config":505,"seo":506,"content":510,"type":429,"slug":540,"category":427,"_id":541,"_type":29,"title":507,"_source":30,"_file":542,"_stem":543,"_extension":33,"date":511,"description":508,"timeToRead":512,"heroImage":509,"keyTakeaways":513,"articleBody":517,"faq":518},"/en-us/the-source/ai/agentic-ai-unlocking-developer-potential-at-scale",{"layout":9,"template":431,"articleType":432,"author":27,"featured":328,"gatedAsset":470,"isHighlighted":6,"authorName":11},{"title":507,"description":508,"ogImage":509},"Agentic AI: Unlocking developer potential at scale","Explore how agentic AI is transforming software development, moving beyond code completion to create AI partners that proactively tackle complex tasks.","https://res.cloudinary.com/about-gitlab-com/image/upload/v1751463876/kiw4eb54r8xtzztvbozf.jpg",{"title":507,"date":511,"description":508,"timeToRead":512,"heroImage":509,"keyTakeaways":513,"articleBody":517,"faq":518},"2025-04-08","6 min read",[514,515,516],"AI agents can slash development time from weeks to hours by autonomously handling complex tasks like codebase modernization, while maintaining configurable human oversight for critical systems.","Unlike basic code assistants, AI agents can work with other agents to accomplish different tasks, freeing developers to focus on innovation and high-value problem-solving.","Specialized AI agents powered by different models can excel at specific tasks like security and testing, delivering better results than one-size-fits-all solutions.","AI has already changed how developers work. [GitLab research](https://about.gitlab.com/developer-survey/2024/ai/) found that 39% of DevSecOps professionals reported using AI for software development in 2024, up 16 percent from the previous year. AI-powered code assistants are now common tools that help teams write code faster, understand codebases, and create documentation. But now we’re seeing a big shift: the emergence of AI agents that work as active partners, not just passive helpers.\n\nThis change from reactive assistants to proactive agents is reshaping how developers build software. Agentic AI is making software creation easier for more people, driving a boom in innovation as more builders can create software that reaches billions of users. However, leaders will need to seek out agentic AI solutions with strong security and compliance guardrails to get the most out of this new wave of innovation without introducing unnecessary risk.\n\n## AI agents vs. AI assistants: What’s the difference?\nThe main difference between AI assistants and agents is how they behave. Code assistants are reactive, waiting for developers to ask questions or request tasks. While helpful for faster coding and understanding code, these assistants are passive in the development process.\n\nAI agents act more like team members. They exhibit reasoning, planning, and maintain context over different tasks, coupled with a certain degree of autonomy to make decisions, interact with other agents, and adapt to changing circumstances. With the shift to agents, AI becomes a true partner in building software.\n\nUnlike assistants that just help write code while teams handle everything else, AI agents can actively orchestrate complex processes, from security checks to compliance reviews. For example, a code review agent can automatically check code, find problems, and offer fixes. Where an assistant needs human input at each step, an agent can move between tasks based on project goals. Unlike simple assistants who can't remember past interactions or learn from mistakes, agents can also learn and adapt over time.\n\n## The spectrum of autonomy\nOne of the most powerful aspects of AI agents is their configurability and level of interaction. While some agents can be highly interactive, others can execute complex tasks in the background with limited to no human interaction. Teams can therefore set different levels of human oversight based on the agent’s work and the task’s importance.\n\nFor simple tasks like summarizing code or drafting documentation, teams might let an agent work independently, only notifying a human team member when the task is finished. For critical tasks involving key business logic or sensitive data, teams can set up approval checkpoints or closely monitor the agent’s work.\n\nThis flexibility helps balance the speed of automation with the need for human control. It’s not all-or-nothing - teams can fine-tune the level of autonomy for different types of tasks and stages of the development lifecycle.\n\n## The power of specialization\nToday’s AI code assistants usually use a single large language model. But the future will bring many specialized agents, each powered by different models built for specific tasks.\n\nWe’re beginning to see the emergence of specialized agents for tasks such as:\n- Code modernization (converting codebases to newer language versions)\n- Security vulnerability detection and remediation\n- Test generation and execution\n- Performance optimization\n- Documentation generation\n- Root cause analysis for pipeline failures\n\nEach task works best with a model built specifically for that job. This specialization allows each agent to excel at its particular task rather than trying to be a jack-of-all-trades.\n\nWhat’s emerging is an ecosystem of specialized agents working together, each powered by different language models optimized for specific tasks. This multi-model approach promises to deliver better results than trying to handle all development tasks with a single, general-purpose model.\n\n## The real-world impact of AI agents\nTasks that once took weeks can now be done in hours with AI agents. For example, updating a large Java codebase to a newer version - work that might take a team weeks - can now be handled much faster by agents.\n\nMore importantly, AI agents help developers reach their highest potential. By handling routine tasks, agents free developers to focus on what they do best: solving complex problems and creating new solutions. This isn’t about replacing developers with AI, but boosting their abilities and letting them focus on higher-level thinking, innovation, and the creative work that needs human insight.\n\nWith AI agents, developers can work at a scale never before possible for individuals or teams. This shifts work from reactive, prompt-based tasks to proactive workflows that link all parts of software creation, helping with coding, planning, design, testing, deployment, and maintenance.\n\n## What to consider when adopting AI agents\nTo prepare for rapid growth in software development and code, companies need to plan ahead. Before adding AI agents to your process, focus on these key areas:\n\n1. **Think about how to boost real productivity, not just add new tools and processes for teams to learn**. By adopting [agentic AI workflows as part of a DevSecOps platform](https://about.gitlab.com/blog/gitlab-duo-workflow-enterprise-visibility-and-control-for-agentic-ai/), you can help developers spend more time creating value for customers without contributing to [AI sprawl](https://about.gitlab.com/the-source/ai/overcome-ai-sprawl-with-a-value-stream-management-approach/). The platform’s built-in reports and dashboards will also help you [measure success](https://about.gitlab.com/the-source/ai/4-steps-for-measuring-the-impact-of-ai/) so you know your team is on the right track.\n2. **Seek out solutions that will work for your whole team**. The best AI agents make everyone more efficient, not just a select few developers.\n3. **Prioritize security and compliance**. As more and more production-ready code is generated by AI, a comprehensive DevSecOps platform is essential for secure software development at scale. If you work in a regulated industry, make sure your AI agent solution meets strict security and data privacy rules. Check if it can work offline or in [air-gapped systems](https://about.gitlab.com/the-source/ai/transforming-government-it-ai-for-air-gapped-environments/) if you need that level of security.\n4. **Look for solutions with enterprise control through human oversight**. AI agents should offer clear approval workflows and configurable guardrails that keep humans in the loop. This balance gives you the speed of automation while maintaining proper governance, which is essential for critical systems and strategic decisions.\n\nCompanies that use an end-to-end DevSecOps platform with automated security scanning, compliance guardrails, and standard workflows will be more equipped to harness the benefits of AI agents without adding unnecessary risk. Those without a platform will struggle to manage the complexity and risks of agentic AI while still delivering a safe and reliable customer experience.\n\n## Looking ahead\nWe’re just at the start of the AI agent revolution in software development. As these tools mature, we’ll see even better teamwork between human developers and AI agents, with agents becoming stronger partners in building software.\n\nLooking towards the future, there is significant potential for convergence between code assistants and AI agents. Code assistants will likely evolve to incorporate more advanced AI agent capabilities, such as increased autonomy in handling coding tasks, proactive problem-solving within the development workflow, and deeper integration with other development tools and processes. Future iterations might see code assistants taking on more complex coding tasks beyond simple generation, such as autonomously debugging, testing, and even deploying code based on high-level requirements, effectively becoming more autonomous “code agents.”\n\nSoftware has changed the world over the past five decades, but only a small fraction of people have the skills to build it. Yet these few developers reach billions through smartphones and the internet. Imagine a world where more people can build, secure, and deliver production-ready software. Agentic AI will make that happen.\n\nThe shift from passive assistants to active development partners is a big step forward in software development. As these specialized agents evolve, software development will be faster, more reliable, and more rewarding for developers working with these new AI partners.",[519,522,525,528,531,534,537],{"header":520,"content":521},"What is agentic AI in software development?","Agentic AI refers to autonomous AI agents that can reason, plan, and take initiative across tasks, unlike reactive code assistants that require human prompts. These agents act more like team members, performing complex tasks with minimal oversight and enabling proactive workflows throughout the software development lifecycle.",{"header":523,"content":524},"How do AI agents differ from traditional code assistants?","While code assistants respond to developer prompts, AI agents can independently complete multi-step tasks, coordinate with other agents, and adapt based on project goals. They can handle functions like security scans, test generation, and code reviews without needing manual intervention at every step.",{"header":526,"content":527},"What are the benefits of using AI agents for developers?","AI agents reduce manual workload by automating time-consuming tasks like updating codebases, running compliance checks, and generating documentation. This allows developers to focus on higher-value work such as innovation, problem-solving, and strategic development, ultimately accelerating delivery without compromising quality.",{"header":529,"content":530},"Can AI agents be customized for different levels of human oversight?","Yes. Teams can configure agent autonomy based on task criticality. For routine tasks, agents may operate independently, while for high-risk or business-critical operations, human approval checkpoints can be integrated to maintain governance and compliance.",{"header":532,"content":533},"Are specialized AI agents more effective than general-purpose models?","Specialized AI agents, each trained for a specific function, such as security, testing, or root cause analysis, typically outperform general-purpose models for their targeted tasks. This modular, multi-agent approach improves accuracy and efficiency by leveraging the strengths of domain-optimized models.",{"header":535,"content":536},"What should companies consider when adopting agentic AI?","Organizations should ensure that AI agents align with their security, compliance, and governance requirements. They should be integrated into an end-to-end DevSecOps platform to avoid AI sprawl, maintain control through human oversight, and support enterprise-wide adoption with consistent workflows.",{"header":538,"content":539},"How will agentic AI shape the future of software development?","Agentic AI will democratize software creation by enabling more people to build and manage production-grade software. As agents become more autonomous and integrated, they will drive faster innovation cycles, improve code quality, and make development more accessible, scalable, and secure.","agentic-ai-unlocking-developer-potential-at-scale","content:en-us:the-source:ai:agentic-ai-unlocking-developer-potential-at-scale:index.yml","en-us/the-source/ai/agentic-ai-unlocking-developer-potential-at-scale/index.yml","en-us/the-source/ai/agentic-ai-unlocking-developer-potential-at-scale/index",{"_path":545,"_dir":427,"_draft":6,"_partial":6,"_locale":7,"config":546,"seo":548,"content":552,"type":429,"slug":575,"category":427,"_id":576,"_type":29,"title":549,"_source":30,"_file":577,"_stem":578,"_extension":33,"date":553,"description":550,"timeToRead":442,"heroImage":551,"keyTakeaways":554,"articleBody":558,"faq":559},"/en-us/the-source/ai/reducing-software-development-complexity-with-ai",{"layout":9,"template":431,"articleType":432,"author":27,"featured":6,"gatedAsset":547,"isHighlighted":6,"authorName":11},"source-lp-getting-started-with-ai-in-software-development-a-guide-for-leaders",{"title":549,"description":550,"ogImage":551},"Reducing software development complexity with AI","Discover how a strategic approach can help organizations maximize the benefits of AI without introducing extra complexity into software development.","https://res.cloudinary.com/about-gitlab-com/image/upload/v1751463941/oium18y7jy4pu6nxozkq.png",{"title":549,"date":553,"description":550,"timeToRead":442,"heroImage":551,"keyTakeaways":554,"articleBody":558,"faq":559},"2025-01-28",[555,556,557],"AI offers a promising solution to software development complexity by automating routine tasks like debugging, code reviews, and testing — but it requires thoughtful human oversight to avoid introducing unnecessary complexity.","The successful implementation of AI in software development involves a gradual approach, starting with low-risk areas, continuous adjustments, and the formulation of policies for AI usage.","AI serves a vital role in reducing software development challenges, including improving code reliability, meeting customer expectations, and enhancing cybersecurity measures.","Software development today involves managing an unprecedented number of moving parts. Teams are handling more languages, tools, and dependencies than ever while facing increased security requirements and shorter deployment cycles. AI offers a way to handle this growing complexity - but only with the right approach.\n\nAdopting any emerging technology can sometimes have unexpected consequences. Installing a new app-based home security system may require learning to use different tools or deal with false alarms. If you get an electric vehicle, you may experience lower fuel costs for your commute, but you’ll need to adjust to new controls and be mindful of where and how far you drive.\n\nIncorporating AI into the software development lifecycle takes this concept to the next level. AI can simplify work for developers by eliminating repetitive tasks - but as a leader, you’re likely weighing the promises of AI against very real concerns, from change management to security to the long-term career stability of your developers.\n\n## Understanding where AI fits\nSoftware development is a complex process that involves multiple steps, including coding, testing, debugging, and maintenance. Development teams face increasing pressure from multiple directions:\n\n- **Rapidly evolving technologies**: Developers may struggle to keep up with the latest tools and techniques as technology advances.\n- **Increasing customer expectations**: Customers are becoming more tech-savvy and have higher expectations for software quality, functionality, and security - putting additional pressure on developers to deliver complex and robust solutions.\n- **Legacy systems**: Many organizations still rely on outdated or legacy systems that can be difficult to integrate with newer technologies, resulting in complexity in the development process.\n- **Evolving security landscape**: As new, more advanced cyber threats emerge, developers must be constantly vigilant against security vulnerabilities, adding another layer of complexity.\n\nAs this complexity increases, it's not always clear where AI can deliver on its promises. The key is identifying where AI can have the most impact. One question I hear often is whether AI code assistants will help to reduce this complexity by augmenting developers’ current workflow or whether developers will need to adapt to an entirely new way of working. The answer, as I see it, is both. Human developers will need to get used to working with AI tools and large language models. And AI code assistants will adapt over time, tapping into more specialized models closely tailored to developers’ workflows. Before long, users will engage with [AI agents](https://about.gitlab.com/the-source/ai/ai-trends-for-2025-agentic-ai-self-hosted-models-and-more/#the-future-of-applications-is-intelligent-adaptive-ai-agents) that respond intuitively and learn over time.\n\n## Starting smart with a phased implementation\nImplementing AI isn’t as simple as flipping a light switch. To harness the efficiencies of AI safely and strategically without introducing unnecessary complexity, I’ve found the most successful teams adopt a gradual approach. They start with low-risk areas to avoid pitfalls and allow development teams to experiment with how AI and other tools fit with their workflows. Be prepared: You might see a temporary dip in productivity before realizing long-term efficiency gains.\n\nThink about how changes ripple through your organization. Teams often face initial resistance to change, and in the case of AI, there is also a potential shift in code quality as AI-driven code volume increases. Understanding [how the entire software development lifecycle benefits from AI](https://about.gitlab.com/the-source/ai/overcome-ai-sprawl-with-a-value-stream-management-approach/) is essential for successful adoption.\n\nFrom there, you’ll need clear guardrails and policies for AI usage, including employee guidelines, data sanitization practices, in-product disclosures, and moderation capabilities. What matters is staying focused on real results: continuously evaluating and adjusting how your teams use AI to ensure it's making development more efficient.\n\n## AI and cybersecurity\nStaying focused on what matters doesn’t just apply to automating code review and improving development cycles. Reducing complexity in software development also yields significant security benefits. The ever-increasing volume and sophistication of cyber attacks, combined with the complexity of organizations’ tech stacks, significantly contribute to security frustrations.\n\nBuilding large, multi-faceted software systems will always involve some complexity - that's unavoidable. However, you can take steps to minimize complications like difficult-to-maintain code repositories and redundant dependencies. When you let this unnecessary complexity creep in, you’re not just creating a larger attack surface - you’re giving your teams more security findings to sort through, prioritize, and address.\n\nThis is where your teams can use AI to minimize the potential negative security impacts of AI tools in other parts of the software development lifecycle. For example, [the new generation of AI-powered development tools](https://about.gitlab.com/gitlab-duo/) evaluate AI-generated code to ensure it doesn’t contain vulnerabilities. If vulnerabilities exist, the tool can help explain the expected impact and how developers can address the issue.\n\nAI creates additional security guardrails to prevent problems such as bad actors injecting harmful answers into large language models while helping your teams create more secure and compliant software. With predictive threat analysis, AI can scan code for security threats and automatically apply patches or reconfigure security settings in response to emerging vulnerabilities. Finally, compliance monitoring is another burden that AI development tools can help lift for software engineering teams.\n\n## Measuring success\n[GitLab research](https://about.gitlab.com/developer-survey/) has shown that software developers spend 25% or less of their day writing code; the rest is devoted to fixing errors, resolving security issues, and updating legacy systems. Automating these tasks with AI allows your teams to utilize their expertise more effectively and focus on problem-solving rather than recreating existing code. This reduces complexity, drives innovation by eliminating wasted effort, and enhances job satisfaction.\n\nFrom a business standpoint, objectives such as improving developer productivity and producing better, more secure code are key performance indicators (KPIs) that translate directly to improved cycle times and better results.\n\nThese are tangible improvements, but there are other areas where AI can reduce complexity. For example, code review has been shown to improve code but often creates bottlenecks as developers wait for review. AI streamlines code reviews and creates comprehensive testing scenarios, enhancing code reliability and reducing bugs, leading to improved software quality and higher customer satisfaction.\n\nFurthermore, AI can predict development bottlenecks and automate routine tasks, leading to more predictable release cycles and faster market entry. Its ability to rapidly and accurately tailor software to user feedback ensures that products meet customer needs and expectations more effectively.\n\n## The path forward: Controlling complexity with AI\nThe complexity of software development may continue to ebb and flow as development teams embed new AI technologies more deeply into their workflows and AI tools become more tailored to developers’ specific needs. But as long as your teams can transfer repetitive development tasks to AI while supervising AI output and intervening when needed, your organization should be able to manage complexity over the long term.\n\nBy providing your development teams with the proper training and time to experiment, you’ll find your organization better equipped to handle increasing complexity while delivering better, more secure software more efficiently.",[560,563,566,569,572],{"header":561,"content":562},"What is the best approach for implementing AI in software development?","A phased implementation strategy is ideal. Start with low-risk areas where AI can provide immediate efficiency gains, such as automated testing, documentation summarization, or code reviews. Gradually expand AI integration while monitoring its impact on workflows, ensuring that AI-driven solutions augment human developers rather than replacing essential development expertise.",{"header":564,"content":565},"How does AI impact software security and compliance?","AI enhances security by automating vulnerability detection, security patching, and compliance monitoring. It reduces complexity by scanning AI-generated code for weaknesses, ensuring compliance with industry standards, and providing predictive threat analysis. AI tools can also prevent bad actors from injecting harmful data into large language models while helping teams respond faster to emerging cybersecurity threats.",{"header":567,"content":568},"What are the biggest challenges when implementing AI in software development?","The main challenges include adapting to AI-driven workflows, ensuring code quality, managing security risks, and avoiding AI sprawl — the uncontrolled adoption of disconnected AI tools. Organizations must also establish clear guidelines for AI usage, including data sanitization, compliance monitoring, and AI-driven security guardrails to ensure AI adoption enhances efficiency rather than creating new complexities.",{"header":570,"content":571},"How should organizations measure the success of AI in software development?","Organizations should track key performance indicators (KPIs) such as developer productivity, cycle time reduction, code quality, security issue resolution time, and software reliability. AI’s ability to accelerate development cycles, reduce human error, and improve overall efficiency should be evaluated against real-world business outcomes like faster deployment, lower costs, and increased customer satisfaction.",{"header":573,"content":574},"How can AI help reduce complexity in software development?","AI streamlines software development by automating repetitive tasks such as code generation, bug detection, and security scanning. It helps developers manage multiple languages, tools, and dependencies while improving workflow efficiency. AI-powered tools can also identify bottlenecks, assist in debugging, and predict potential security vulnerabilities before they become major issues.","reducing-software-development-complexity-with-ai","content:en-us:the-source:ai:reducing-software-development-complexity-with-ai:index.yml","en-us/the-source/ai/reducing-software-development-complexity-with-ai/index.yml","en-us/the-source/ai/reducing-software-development-complexity-with-ai/index",{"_path":580,"_dir":427,"_draft":6,"_partial":6,"_locale":7,"config":581,"seo":582,"content":586,"type":429,"slug":610,"category":427,"_id":611,"_type":29,"title":583,"_source":30,"_file":612,"_stem":613,"_extension":33,"date":587,"description":584,"timeToRead":588,"heroImage":585,"keyTakeaways":589,"articleBody":593,"faq":594},"/en-us/the-source/ai/ai-trends-for-2025-agentic-ai-self-hosted-models-and-more",{"layout":9,"template":431,"articleType":432,"author":27,"featured":6,"gatedAsset":547,"isHighlighted":6,"authorName":11},{"title":583,"description":584,"ogImage":585},"Agentic AI, self-hosted models, and more: AI trends for 2025","Discover key trends in AI for software development, from on-premises model deployments to intelligent, adaptive AI agents.","https://res.cloudinary.com/about-gitlab-com/image/upload/v1751464096/twyszwpyraghcxz1bruy.png",{"title":583,"date":587,"description":584,"timeToRead":588,"heroImage":585,"keyTakeaways":589,"articleBody":593,"faq":594},"2024-12-18","3 min read",[590,591,592],"Artificial intelligence is already having a major impact on software development, enhancing code quality and efficiency by removing a wide range of tasks.","Software developers will work alongside AI agents that will facilitate real-time problem solving, rapid optimization of application performance, and overall improvement in software quality, enabling developers to concentrate on strategic decision-making.","Increased use of on-premises AI deployments, particularly in regulated industries, will give companies more control over data privacy and security, as well as the ability to customize their software according to individual needs.","According to [GitLab's 2024 research](https://about.gitlab.com/developer-survey/), 78% of organizations will use artificial intelligence in their software development processes within the next two years - a dramatic shift that's already transforming how teams build and deliver software. The research also shows that the number of organizations actively using AI has jumped from 23% to 39% in the last year alone.\n\nAs software development teams race to integrate AI into their workflows, major shifts are emerging that will fundamentally change how we build software. From intelligent AI agents that adapt in real time to the rise of customized on-premises models, here are three ways AI will significantly alter software development.\n\n## The future of applications is intelligent, adaptive AI agents\nWhile the first wave of AI in software development focused on reactive code assistants for code generation and completion, the future belongs to agentic AI. [Intelligent, adaptable AI agents](https://about.gitlab.com/blog/meet-gitlab-duo-workflow-the-future-of-ai-driven-development/) will surpass the limitations of traditional software. Rather than interacting with fixed interfaces and preset workflows, users will engage with AI agents that respond intuitively and learn over time.\n\nThese AI-powered agents will serve as the application, providing a more interactive and conversational experience. As AI agents can perform complex tasks, offer guidance, and learn from interactions in real time, agentic AI will lead to significantly more personalized and responsive applications, fundamentally reshaping how we use software.\n\n## AI assistants will evolve to become proactive collaborators\n[AI assistants are getting smarter](https://about.gitlab.com/gartner-mq-ai-code-assistants/), moving beyond reactive prompt-based interactions to proactive problem-solvers. As part of this evolution, AI-powered tools will become central hubs for development, anticipating developers’ needs and offering real-time suggestions for optimizing application performance, security, and maintenance. This new generation of AI assistants will tackle complex projects and tasks with little human interaction, accelerating the software development process. This shift will streamline the entire software development lifecycle, making it more accessible through a simple user interface.\n\nThe role of software developers will evolve alongside these advancements. AI will not replace human developers but will augment their capabilities, allowing them to focus on what they love most: solving complex technical problems. By automating routine tasks and providing expert guidance, AI assistants will empower developers to delve deeper into business problem-solving, continuously improve code quality, and explore new technologies and skills.\n\n## More companies will run customized models on-premises\nIn 2025, organizations will shift toward smaller and more specialized AI deployments. As open source models become more cost-effective and accessible, teams will increasingly opt to run customized versions within their own data centers. As a result, it will be cheaper, faster, and easier for organizations to [host their own large language models and fine-tune them to their individual needs](https://about.gitlab.com/releases/2024/10/17/gitlab-17-5-released/#use-self-hosted-model-for-gitlab-duo-code-suggestions). Companies will find they can combine their data with existing models and tailor the customer experience at a fraction of today’s costs.\n\nMeanwhile, increased compliance risks associated with AI will drive regulated industries - like financial institutions and government agencies - to deploy models in air-gapped environments for reduced latency and greater control over data privacy and security.\n\n## Conclusion\nThe future of software development is inextricably linked to AI. These technologies are transforming how software is created, delivered, and maintained. By embracing AI in all its forms - from generative AI to proactive AI assistants to fully autonomous AI agents - organizations can gain a competitive edge, improve efficiency, and deliver innovative solutions that meet customers’ evolving needs.\n\nThis transformation requires thoughtful preparation: strategic planning, investment in talent and infrastructure, and a commitment to continuous learning and adaptation. Organizations that successfully navigate this evolving landscape will be well-positioned to thrive in the digital age.\n",[595,598,601,604,607],{"header":596,"content":597},"Why are companies moving toward self-hosted AI models?","Organizations are shifting toward self-hosted AI models to enhance data privacy, reduce costs, and customize AI solutions for their specific needs. With advancements in open-source AI, companies can fine-tune models in on-premises environments, ensuring compliance with regulations and improving performance while maintaining control over sensitive data.",{"header":599,"content":600},"What are the benefits of running AI models in on-premises environments?","Deploying AI models on-premises offers organizations greater control over data security, improved compliance with regulatory requirements, and reduced latency. This approach is particularly valuable for industries handling sensitive data, such as finance, healthcare, and government agencies.",{"header":602,"content":603},"How are AI-powered coding assistants evolving?","AI coding assistants are transitioning from reactive tools to proactive collaborators. Future AI assistants will anticipate developer needs, provide intelligent recommendations, automate complex tasks, and enhance software security, ultimately making software development more efficient and accessible.",{"header":605,"content":606},"How can organizations prepare for AI-driven software development in 2025?","To successfully adopt AI-driven development, companies should invest in AI infrastructure, upskill developers, implement responsible AI governance, and explore hybrid AI solutions that balance cloud and on-premises deployment. Staying informed about AI trends will help teams leverage AI for innovation and efficiency.",{"header":608,"content":609},"What is agentic AI, and how will it impact software development?","Agentic AI refers to AI systems that operate autonomously, learning from interactions and adapting in real time. Unlike traditional AI coding assistants that react to prompts, agentic AI acts proactively, streamlining software development by automating workflows, improving efficiency, and personalizing user experiences.","ai-trends-for-2025-agentic-ai-self-hosted-models-and-more","content:en-us:the-source:ai:ai-trends-for-2025-agentic-ai-self-hosted-models-and-more:index.yml","en-us/the-source/ai/ai-trends-for-2025-agentic-ai-self-hosted-models-and-more/index.yml","en-us/the-source/ai/ai-trends-for-2025-agentic-ai-self-hosted-models-and-more/index",{"_path":615,"_dir":427,"_draft":6,"_partial":6,"_locale":7,"config":616,"seo":618,"content":622,"type":429,"slug":643,"category":427,"_id":644,"_type":29,"title":619,"_source":30,"_file":645,"_stem":646,"_extension":33,"date":623,"description":620,"timeToRead":624,"heroImage":621,"keyTakeaways":625,"articleBody":629,"faq":630},"/en-us/the-source/ai/6-strategies-to-help-developers-accelerate-ai-adoption",{"layout":9,"template":431,"articleType":432,"author":27,"featured":6,"gatedAsset":617,"isHighlighted":6,"authorName":11},"source-lp-how-to-think-about-developer-productivity-in-the-age-of-ai",{"title":619,"description":620,"ogImage":621},"6 strategies to help developers accelerate AI adoption","AI in software development is here to stay. Here’s how leaders can create an environment that fosters innovation while acknowledging potential concerns.","https://res.cloudinary.com/about-gitlab-com/image/upload/v1751464541/da4tvbmwsisqabz8i0mc.png",{"title":619,"date":623,"description":620,"timeToRead":624,"heroImage":621,"keyTakeaways":625,"articleBody":629,"faq":630},"2024-10-29","7 min read",[626,627,628],"Integrating AI into software development processes can enhance developer productivity by streamlining workflows, allowing teams to focus on innovation over repetitive tasks.","Despite the benefits, successfully integrating AI-powered solutions into workflows can be challenging due a lack of knowledge or resources, workflow adaptation difficulties, and fear of job loss.","Strategies for successful AI implementation include clarifying the goals and objectives of AI, establishing guardrails and workflows, and focusing on talent and culture transformation.","By integrating artificial intelligence (AI) into the coding process, software developers can spend more time on strategic tasks, reduce cognitive load, and deliver greater value.\n\nOrganizations are already making significant investments in AI. According to [GitLab’s 2024 Global DevSecOps Report](https://about.gitlab.com/developer-survey/), 78% of respondents said they are currently using AI in software development or plan to in the next two years, up from 64% in 2023. And organizations adopting AI are already seeing benefits, such as improved developer productivity, better code quality, and more secure code. [Embracing AI](https://about.gitlab.com/the-source/ai/how-ai-helps-devsecops-teams-improve-productivity/) enables development teams to devote more time to creative problem-solving and innovation rather than time-consuming and repetitive tasks such as manually writing boilerplate code.\n\nDespite AI's clear benefits, teams may struggle to integrate AI tools successfully into their day-to-day processes. This challenge can be attributed to various factors, such as a lack of knowledge or resources, difficulty adapting existing workflows and tools, and the fear of losing jobs to automation. Nearly half (49%) of our survey respondents voiced concern that AI will replace their roles in the next five years.\n\nUnderstanding where your team is today is necessary to set them up for success when integrating AI. [Our research](https://about.gitlab.com/developer-survey/2024/ai/) shows that the majority (56%) of organizations are in the Evaluation and Exploration stage - meaning most teams have started to set achievable targets for AI adoption but haven’t actually started using it in their software development lifecycle.\n\nWhether you’re an early adopter or you’re still exploring the idea of AI, here are six strategies you can use to set your team up for success:\n\n## 1. Clarify the goals and objectives of AI adoption\nYour first step should be to create an AI governance model for your organization. What are the goals and objectives of AI adoption? How will it fit into your existing processes and workflows?\n\nIdentifying a leader to oversee AI strategy and implementation is critical. While some companies are beginning to hire a chief AI officer (CAIO), the role doesn’t have to be an immediate addition to the C-suite; it can be a transitional title that a VP assumes to coordinate AI usage across teams.\n\nThe primary goal is to identify and prioritize high-impact AI use cases that directly support business outcomes, focusing on areas where AI can create significant value, such as automation, personalization, or data-driven decision-making. It’s important to remember that AI success isn’t possible without first addressing the privacy, security, and legal requirements your organization might face and how AI adoption plays into continued compliance.\n\n## 2. Establish AI guardrails and workflows\nBefore incorporating AI into your development environment, you'll need to establish guidelines to ensure it is used responsibly and effectively. Set up automated testing, including using a security analyzer, to create a gating mechanism that ensures all AI-generated code is reviewed before being promoted to production. And beware of shadow AI - the latest variation of shadow IT - where workers adopt their own AI assistants while working on your code base, which can lead to the leakage of sensitive information and intellectual property.\n\nYou'll also want to think now about how your teams will use different machine learning models for different types of tasks. One size does not fit all. Large language models (LLMs) are often tuned for specific tasks, meaning teams that are using the same AI models across multiple use cases may not be getting optimal results. As you shop around for AI tools, look for vendors that allow you to use a variety of models tailored to specific use cases - this will save you from having to rip and replace later.\n\n## 3. Build a data-driven AI structure\nThe results that AI can drive for organizations are only as good as the data that AI systems have access to. Feeding data into your AI systems will allow you to tailor the results to your organization’s needs and improve efficiency and productivity across your software development lifecycle. However, long-term success requires a structured approach that allows data to be used across the organization to inform prompts and enhance generative AI outputs.\n\nTo that end, enterprises must:\n\n- Ensure robust data collection, storage, cleaning, and processing mechanisms.\n- Establish clear governance around data access, usage, security, and privacy, especially to ensure compliance with regulations like GDPR or CCPA.\n- Break down data silos to facilitate cross-department collaboration and leverage data across various parts of the organization. Now is the time for developers and data scientists to collaborate on using data warehouses and data lakes to facilitate access to training models and application usage.\n\n## 4. Focus on talent and culture transformation\nContinuous upskilling is critical to safely, securely, and responsibly unlocking AI’s potential. Build a team of data scientists, AI engineers, and other experts to design, develop, and implement AI solutions. Upskilling employees to ensure they can use and maintain AI systems effectively is critical. Finally, embracing AI is a journey, and it will require some [cultural shifts](https://about.gitlab.com/the-source/ai/5-ways-execs-can-support-their-devops-teams-with-ai/). To succeed, it is critical to foster a culture that embraces AI and data-driven decision-making. Encourage experimentation and innovation while addressing fears around automation and job displacement.\n\n## 5. Embrace iteration\nImplementing AI is an ongoing process. Adopt a continuous learning approach, where AI solutions are constantly refined and improved based on feedback, new data, and technological advances. Developers must be given an experimentation period to assess how AI fits into their individual workflows. It’s also important to note that there might be a short-term dip in productivity before the organization benefits from long-term gains. Managers must anticipate this by emphasizing transparency and accountability throughout the implementation and iteration cycles.\n\n## 6. Measure success beyond lines of code\nIt's true that metrics such as the number of tasks completed or lines of code written can be good proxies to help you identify areas where AI is having the biggest impact on your team. However, AI is more than just code generation. What really matters is how AI is driving metrics that are important to the business, such as how quickly teams are able to deliver value to customers, or the code quality of the final product.\n\nKnowing how many lines of code a team produced won’t tell you the full story here. Measuring success in AI adoption requires moving [beyond traditional productivity metrics](https://about.gitlab.com/the-source/ai/4-steps-for-measuring-the-impact-of-ai/) and focusing on KPIs that demonstrate measurable business value, such as faster software delivery, improved developer satisfaction, and higher customer satisfaction scores.\n\n## Conclusion: Empowering developers through AI adoption\nEven if your organization has not fully embraced AI, the time to start is now. According to Gartner®, by 2028, 75% of enterprise software engineers will use AI coding assistants, up from less than 10% in early 2023 [1].\n\nThe adoption curve is steep, but we are still relatively early in the AI hype cycle. In fact, if your team is only just starting to look into adopting an AI code assistant, they may be well-positioned to avoid some of the growing pains early adopters have experienced.\n\nIn addition to the strategies above, adopting an [AI solution integrated into an end-to-end DevSecOps platform](/gitlab-duo/) can jumpstart success by supporting developers at every stage of their workflow.\n\nAs AI transforms the workplace, business leaders should ask how they can harness the power of AI across the software development lifecycle to accelerate innovation and drive tangible benefits for customers.\n\n[1] _Source: Gartner, Top 5 Strategic Technology Trends in Software Engineering for 2024, Joachim Herschmann, Manjunath Bhat, Frank O'Connor, Arun Batchu, Bill Blosen, May 2024. GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. All rights reserved._",[631,634,637,640],{"header":632,"content":633},"How can AI improve software development productivity?","AI enhances software development by automating repetitive tasks, reducing cognitive load, and enabling developers to focus on strategic problem-solving. AI-powered coding assistants can generate code, conduct automated testing, and improve code quality, leading to faster development cycles and more secure applications.",{"header":635,"content":636},"What are AI guardrails, and why are they important in software development?","Success metrics should go beyond traditional coding productivity, such as lines of code written. Instead, organizations should focus on business-impacting KPIs like faster software delivery, improved developer satisfaction, higher code quality, and better customer experience.",{"header":638,"content":639},"What are the biggest challenges in adopting AI for software development?","Common challenges include integrating AI into existing workflows, ensuring data privacy and security, overcoming resistance to change, and addressing concerns about job displacement. Organizations can mitigate these challenges by establishing clear AI governance, upskilling employees, and fostering a culture of experimentation and innovation.",{"header":641,"content":642},"What steps should developers take to prepare for AI-driven software development?","Developers should focus on continuous learning, gaining proficiency in AI tools, and understanding data-driven decision-making. Collaborating with AI engineers and data scientists, experimenting with AI-powered coding assistants, and staying updated on emerging AI trends will help developers maximize AI's potential in software development.","6-strategies-to-help-developers-accelerate-ai-adoption","content:en-us:the-source:ai:6-strategies-to-help-developers-accelerate-ai-adoption:index.yml","en-us/the-source/ai/6-strategies-to-help-developers-accelerate-ai-adoption/index.yml","en-us/the-source/ai/6-strategies-to-help-developers-accelerate-ai-adoption/index",[425,467,503,544,579,614],{"ai":366,"platform":374,"security":370},1758326274309]