Generative AI is reshaping software development: companies are increasingly deploying it as coding assistants — from IDE plugins to autonomous agents. Yet between hype and reality there’s a measurable gap. Those who want to stay ahead by 2030 are laying the groundwork now.

In Germany, 36% of companies are already using AI — nearly twice as many as a year earlier1. Internationally, 84% of developers use AI tools or plan to; 51% of professional developers use them daily2. Measurability remains the key hurdle: 49% cite the difficulty of estimating and demonstrating the value of AI projects as the top barrier3. Organizations that regularly assess their AI systems are more than three times as likely to achieve high GenAI value, according to Gartner4. 63% of leaders from high-AI-maturity organizations concretely measure ROI, risk, and customer impact, and 45% keep AI initiatives in production for three years or more5.

A 2025 LeadDev survey6 of 883 engineering leads confirms the trend: 66% have adopted AI tools for at least some use cases, another 20% are piloting them, and only 2% have no adoption plans. 59% report higher developer productivity.

Coding AI: from autonomous agents to an Agentic AI toolchain

The tools for AI-assisted coding range from GitHub Copilot and Anthropic’s Claude Code to AWS Amazon Q Developer and the AI-native IDE Kiro. In my view, they’re now an integral part of modern toolchains, embedding into IDEs and command-line interfaces to generate, refactor, document, and test code.

The next step is Agentic AI: autonomous AI agents that plan and execute tasks on their own. According to Gartner (June 2025)7, roughly one-third of enterprise software applications will include such capabilities by 2028. But Gartner (June 2025)7 forecasts that over 40% of agentic AI projects will be canceled by the end of 2027 — due to escalating costs, unclear business value, or inadequate risk controls. A symptom of tooling that’s outpaced methodology.

2026 addendum: that methodology now has a name — Agentic Engineering, a structured, verifiable approach that pairs Agentic AI with specifications, guardrails, and architectural control. More in the companion essay to the talk “Coding Is Cheap, Software Is Not” (April 2026).

Frameworks like the Model Context Protocol (MCP) and platforms like Amazon Bedrock AgentCore make orchestrating multi-agent systems easier. Early agent-based tools like the Kiro development environment or the Amazon Q CLI command-line tool (now Kiro CLI8), combined with MCP servers, in my view enable complex integrations in minutes instead of days. In multi-agent collaboration, specialized Agentic AI components work in parallel on subtasks — a defining trend for the years ahead, and one that quickly devolves into mess without disciplined engineering practice.

Governance, risks, and technical debt

Many companies are wrestling with governance: who steers and controls AI usage? 70% of CDAOs own the AI strategy, and 36% report directly to the CEO9. According to McKinsey (The State of AI, 2025)10, direct CEO oversight of AI governance goes hand in hand with higher self-reported bottom-line impact from GenAI — a correlation, not proof of causation.

Risk mitigation is also gaining weight: in my view, AI talent and measures against hallucinations and bias belong near the top of the investment list. How can AI impact be measured? 60% of organizations cite the lack of metrics on AI impact as a key challenge. On technical debt (code quality, maintainability): 41% see no significant impact from AI assistants, 23% feel their technical debt has been reduced — and among companies that measure the impact of AI coding tools, that figure climbs to 54%6.

AI coding doesn’t automatically degrade code quality. What matters is clear ownership, structured evaluation, and targeted training.

My assessment: in 2024, AI served mainly as an assistant for individual tasks — by late 2025, many developers are already coding alongside coding agents and AI-native IDEs.

Team and skill shifts

AI is reshaping team structures and roles. 54% of engineering leads expect fewer junior positions6: AI takes over routine coding while developers focus more on architecture, reviews, and creative problem-solving. Critical thinking (43%) and architectural design (34%) are seen there as the most in-demand competencies over the next three years. 60% of respondents plan to focus on learning how to manage AI agents over the next year, and 53% on prompt engineering. In my view, new roles are emerging as well: context engineer, AI ethics advisor, AI product owner.

According to IDC FutureScape 202611, by 2026 around 40% of all job roles in G2000 companies will involve working with AI agents — redefining entry, mid, and senior positions. As early as its 2023 Future of Work predictions12, IDC pointed to “learning in the flow of work” and personalized technology skills development. This underscores the need for new learning models and formalized mentoring.

Case study: STP One – AI in legal tech

German legal software provider STP.One is developing Legal Twin, an AI for automated case file review, together with Storm Reply (disclosure: I am Managing Director of Storm Reply) and Data Reply. Generative AI analyzes legal case files and produces summaries that give lawyers a quick overview of the entire case. The goal, per the Reply case study13: screen thousands of case files in just a few minutes — saving time, reducing costs, minimizing risks.

Outlook and ROI potential through 2030

A reality check rather than a vision: AI coding is still in its early days, but companies are laying the foundation for major efficiency gains. IDC forecasts14 that global AI investments will rise from USD 307 billion (2025) to USD 632 billion (2028). Successful firms focus on a few prioritized use cases, define clear metrics for productivity, quality, and risk, and run regular assessments — organizations with regular assessments are more than three times as likely to achieve high GenAI value, according to Gartner4.

For organizations, the message is: act now, adapt, and build the foundations. That includes:

Companies that implement AI-powered coding pragmatically and responsibly can unlock significant ROI by 2030: faster release cycles, lower costs, and freeing valuable specialists from routine work.

Sources

  1. Bitkom: “Durchbruch bei Künstlicher Intelligenz”, press release (German), 15 September 2025. Representative survey of 604 companies in Germany with 20+ employees (calendar weeks 27–32, 2025). bitkom.org
  2. Stack Overflow: “2025 Stack Overflow Developer Survey”, section “AI”, survey, 2025. Over 49,000 responses from 177 countries. survey.stackoverflow.co
  3. Gartner: “Gartner Survey Finds Generative AI Is Now the Most Frequently Deployed AI Solution in Organizations”, press release, 7 May 2024. Survey of 644 respondents from organizations in the U.S., Germany and the U.K. (Q4 2023). gartner.com
  4. Gartner: “Gartner Survey Finds Regular AI System Assessments Triple the Likelihood of High GenAI Value”, press release, 4 November 2025. Survey of 360 organisations with at least 250 employees, May to June 2025. gartner.com
  5. Gartner: “Gartner Survey Finds 45% of Organizations With High AI Maturity Keep AI Projects Operational for at Least Three Years”, press release, 30 June 2025. Survey of 432 respondents from organizations in the U.S., U.K., France, Germany, India and Japan (Q4 2024). gartner.com
  6. LeadDev: “The AI Impact Report 2025”, p. 13. Survey of 883 members of the LeadDev community, 30 May to 21 June 2025. leaddev.com
  7. Gartner: “Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027”, press release, 25 June 2025. gartner.com
  8. AWS: repository “aws/amazon-q-developer-cli”, README notice, GitHub; see also Kiro: “Upgrading from Amazon Q Developer CLI”, documentation (Kiro CLI available from 17 November 2025). github.com
  9. Gartner: “Gartner Survey Finds 70% of CDAOs Are Responsible for AI Strategy and Operating Model”, press release, 12 May 2025. Gartner CDAO Agenda Survey 2025, 504 data and analytics executive leaders worldwide (September–November 2024). gartner.com
  10. McKinsey & Company: “The state of AI: How organizations are rewiring to capture value”, survey report, 12 March 2025. Online survey of 1,491 participants in 101 nations (16–31 July 2024). mckinsey.com
  11. IDC: “IDC FutureScape 2026 Predictions Reveal the Rise of Agentic AI and a Turning Point in Enterprise Transformation”, press release (prediction), 23 October 2025. my.idc.com
  12. IDC: “IDC FutureScape: Worldwide Future of Work 2024 Predictions”, report (predictions), October 2023, IDC #US49963723. idc.com
  13. Reply: “Leading German legal tech company revolutionizes file review with AI”, case study, n.d. reply.com
  14. IDC: “AI & GenAI Predictions: Key Insights for 2025 and Beyond”, eBook page, undated (2025 predictions). See also IDC press release “Worldwide Spending on Artificial Intelligence Forecast to Reach $632 Billion in 2028, According to a New IDC Spending Guide”, 19 August 2024. info.idc.com