The AI-Powered Software Development Lifecycle

By now, AI is already becoming mainstream in software development (right?), but are we focusing on the right areas? Developers now write code faster than ever with AI-powered tools like Copilot, yet meetings, requirements engineering, and testing remain time-consuming bottlenecks. What if AI could optimize these less exciting aspects of the Software Development Lifecycle (SDLC) instead?

Take a nurse as an example. She wants to spend more time with patients, yet much of her day is spent on administrative tasks behind a computer. Yet we focus on optimizing her time with patients, while we should focus on reducing the digital overhead instead? The same applies to software teams: developers want to build great software, product owners want to shape the vision, and testers want to ensure quality. All without being bogged down by tedious work.

In this session, we’ll explore the AI-Powered SDLC, moving beyond just coding assistance. We do a quick round through all the steps of the SDLC and how AI can help us to be better, faster and ... less bored by tedious tasks.

Transcript

1. THE AI-POWERED SOFTWARE DEVELOPMENT LIFECYCLE AIW07 RENÉ VAN OSNABRUGGE GLOBAL CONSULTANCY DIRECTOR
2. WELCOME TO SCAMANDER WANDS
3. THE SOFTWARE DEVELOPER LIFECYCLE
4. THE SDLC – WATERFALL (1950-1970) REQUIREMENTS DESIGN IMPLEMENTATION TESTING DEPLOYMENT MAINTENANCE
5. THE SDLC – SPIRAL MODEL 1970- 1990 PLAN RISK ANALYSIS EVALUATE ENGINEERING
6. THE SDLC – SCRUM & AGILE 2000’S
7. THE SDLC – DEVOPS 2010’S CODE PLAN MONITOR OPERATE DEPLOY BUILD TEST RELEASE
8. THE SDLC – AI POWERED CODE PLAN MONITOR OPERATE DEPLOY BUILD TEST RELEASE
9. THE SDLC – AND MORE + SECURITY + INFRASTRUCTURE AS CODE + CI/CD + QUALITY ENGINEERING + CONTAINERS + CLOUD NATIVE + OBSERVABILITY & MONITORING + COMPLIANCE
10. AND BEFORE THIS THERE IS ENVISIONING
11. DEMO ENVISIONING WITH GITHUB SPARK
12. THE NEW CODING LANGUAGE IS ENGLISH
13. WHAT WILL BE THE NEW ROLE OF A DEVELOPER IN THE TIME OF AI?
14. MACHINE CODE AND ASSEMBLY HIGH LEVEL LANGUAGES FRAMEWORKS AND IDES DEVOPS PRACTICES DEVELOPER AS MACHINE OPERATOR DEVELOPER AS CODER DEVELOPER AS ARCHITECT DEVELOPER AS COLLABORATOR DEVELOPER AS DIRECTOR AI ENABLED SOFTWARE ENGINEERING AI SOFTWARE DEVELOPER EVOLUTION TIMELINE DEVELOPER ?
15. THE PRODUCTIVITY PARADOX AUTOMATION ANXIETY EFFICIENCY OBSESSION HUMAN AI COLLABORATION
16. 1ST INDUSTRIAL REVOLUTION Machines will steal our jobs and destroy craftsmanship. Many low-skill jobs were displaced but new, higher- skill roles (mechanics, engineers, managers) emerged. FEAR REALITY
17. 2ND INDUSTRIAL REVOLUTION Being reduced to “cogs in the machine.” Productivity skyrocketed, prices dropped, and new consumer industries emerged FEAR REALITY
18. COMPUTER REVOLUTION Clerks, secretaries, accountants feared job loss as computers could “think” faster Computers replaced repetitive work but created entire new industries . FEAR REALITY
19. INTERNET REVOLUTION Knowledge workers feared outsourcing to cheaper labor markets and “the death of local expertise Global collaboration increased productivity and innovation FEAR REALITY
20. AI REVOLUTION AI will replace developers, writers, artists, even managers We don’t know. AI seems to improve productivity but it depends heavily on the prompting, judgement and validation FEAR REALITY
21. HUMANS ARE POWERED BY AN OBSESSION FOR PRODUCTIVITY
22. EVEN ARTISTOTLE WONDERED… “IF EVERY TOOL, WHEN ORDERED, OR EVEN OF ITS OWN ACCORD, COULD DO THE WORK THAT BEFITS IT... IF THE SHUTTLE WOULD WEAVE AND THE PLECTRUM TOUCH THE LYRE WITHOUT A HAND TO GUIDE THEM, CHIEF WORKMEN WOULD NOT NEED SERVANTS, NOR MASTERS SLAVES.“ ARISTOTLE (384–322 BCE)
23. AND HE WAS NOT THE ONLY ONE.. "MAN-COMPUTER SYMBIOSIS IS AN EXPECTED DEVELOPMENT...“ J.C.R. LICKLIDER, 1960 “THE AUTOMATIC MACHINE... IS THE PRECISE ECONOMIC EQUIVALENT OF SLAVE LABOR.“ NORBERT WIENER, 1950 "THE HOPE IS THAT MACHINES WILL DO THE DRUDGERY, AND HUMANS WILL BE FREE FOR HIGHER PURSUITS.“ JOHN MCCARTHY, FATHER OF AI
24. JOHN MAYNARD KEYNES AND THE 15-HOUR WEEK "WE ARE BEING AFFLICTED WITH A NEW DISEASE... NAMELY, TECHNOLOGICAL UNEMPLOYMENT. THIS MEANS UNEMPLOYMENT DUE TO OUR DISCOVERY OF MEANS OF ECONOMISING THE USE OF LABOUR OUTRUNNING THE PACE AT WHICH WE CAN FIND NEW USES FOR LABOUR.“ "THREE-HOUR SHIFTS OR A FIFTEEN- HOUR WEEK MAY PUT OFF THE PROBLEM FOR A GREAT WHILE." J.M. KEYNES, 1930
25. THE INDUSTRY FOCUSSES ON PRODUCTIVITY McKinsey May 2024
26. BUT DO WE FOCUS ON THE RIGHT THINGS? 28
27. QUIZ HOW MUCH TIME DOES A TEACHER SPEND ON TEACHING? On average across countries, teachers spend half of their working time in non-teaching activities including planning lessons, marking and collaborating with other teachers. Source: How Much Time Do Teachers Spend on Teaching and Non-teaching Activities? | OECD
28. QUIZ HOW MUCH TIME DOES A NURSE SPEND NURSING AT THE BED? Around 35% of time is administration Source: https://aacnjournals.org/ccnonline/article/43/1/10/31956/Documentation-and-Nurses-Time-Caring-for-Patients
29. QUIZ HOW MUCH TIME DOES A DEVELOPER SPEND CODING? Different research. But between 30-70% Source: https://www.software.com/reports/code-time-report https://www.microsoft.com/en-us/research/wp-content/uploads/2019/04/devtime-preprint-TSE19.pdf https://thenewstack.io/how-much-time-do-developers-spend-actually-writing-code/
30. 32 WE TRY TO MAKE SOFTWARE DEVELOPERS MORE PRODUCTIVE IN A VERY TIGHT SCOPE Core Work Context Work Productivity Gain
31. Core Work Context Work Productivity Gain 33 WE NEED TO FOCUS ON THE CONTEXT AS WELL
32. Context Work Core Work WE TRY TO MAKE SOFTWARE DEVELOPERS MORE PRODUCTIVE IN A VERY TIGHT SCOPE 30% 100% Core Work 36% Context Work 70% Core Work 33% Context Work 77% Improving 20% core work efficiency Improving both core & context 10% is easier and gets better results 70% 106% 110% Total productivity increase Total productivity increase
33. 35 TRANSLATED TO OUR WORLD Development Build Pair programming Code generation Documentatio n generation Pipeline creation
34. 36 WE NEED TO FOCUS ON THE ENTIRE LIFECYLE Feature management Planning Development Build Test Deployment Release Monitor Triage Task creation Estimation Duplication removal Pair programming Code generation Documentatio n generation Refinement Creating sub tasks Generating personas Acceptance criteria Log querying Release Notes Incident creation Pipeline creation Generating test cases Building test data Deduplication Disambi- guation
35. 37 AND NOT ONLY FOR THE DEVELOPER ROLE Product Owner Role Develop er Role Tester Role Operatio ns Role Product Owner Role Develop er Role Tester Role Operatio ns Role
36. 38 BUILDING BLOCKS LLM Models AI Enabled Tools Agentic AI MCP
37. PLANNING & REQUIREMENTS CHALLENGES AMBIGUOUS OR SHIFTING REQUIREMENTS CAUSE UP TO 50% REWORK. EXCESSIVE MEETINGS AND MISCOMMUNICATION BETWEEN BUSINESS AND DEVS. MISALIGNMENT: 63% OF DEVELOPERS SAY LEADERSHIP DOESN’T UNDERSTAND THEIR CHALLENGES. AI OPPORTUNITIES LLM FEATURES: SUMMARIZE CUSTOMER FEEDBACK, AUTO- GENERATE SPECS AND ACCEPTANCE CRITERIA. CONVERSATIONAL REFINEMENT AI-POWERED KNOWLEDGE BASES ANSWERING “HAVE WE BUILT THIS BEFORE?” 39
38. DEMO COPILOT STUDIO REQUIREMENT AGENT
39. ARCHITECTURE & DESIGN CHALLENGES TIME-INTENSIVE DOCUMENTATION AND EVALUATION OF ALTERNATIVES. KNOWLEDGE SILOS. RATIONALE OFTEN TRAPPED IN EXPERTS’ HEADS. IMPLEMENTATION IS NOT DESIGN DESIGNERS OUT OF THE SDLC AI OPPORTUNITIES AUTO-GENERATE DIAGRAMS (UML/MERMAID) AND ARCHITECTURAL SKETCHES FROM TEXT. BRAINSTORM SCALABILITY AND DESIGN ISSUES (“WHAT ARE PERFORMANCE BOTTLENECKS?”). SHORTER FEEDBACK LOOP FOR DESIGNERS AUTOMATED DESIGN THREAT MODELING AND RISK FLAGGING. 41
40. DEMO ARCHITECTURE VISUALIZATION
41. DEMO UI DESIGN
42. CODING & IMPLEMENTATION CHALLENGES REPETITIVE BOILERPLATE CODING, ENVIRONMENT SETUP OVERHEAD. DEBUGGING AND CONTEXT SWITCHING. 61% OF DEVS SPEND 30+ MIN/DAY SEARCHING FOR HELP. NOT MUCH DEVELOPER TIME IS ACTUAL CODING (VS. 84% OTHER WORK). AI OPPORTUNITIES CODE AUTOCOMPLETION UP TO 55% FASTER TASK COMPLETION. IN-CONTEXT Q&A (“WHAT DOES THIS FUNCTION DO?”). AUTOMATED REFACTORING, SYNTAX UPGRADES, AND TRANSLATION BETWEEN LANGUAGES. MULTI-FILE CODE GENERATION AND REPO-WIDE REASONING (“ADD CACHING EVERYWHERE”). 44
43. DEMO GITHUB CODING AGENT
44. CODE REVIEW & COLLABORATION CHALLENGES SLOW PR TURNAROUND INCONSISTENT FEEDBACK; SUPERFICIAL OR NITPICKY REVIEWS. COGNITIVE OVERLOAD REVIEWING LARGE DIFFS. AI OPPORTUNITIES AI-GENERATED PR SUMMARIES AND AUTOMATED REVIEW COMMENTS. GUIDED MANUAL REVIEWS. “EXPLAIN THIS FUNCTION” OR “CHECK FOR SECURITY ISSUES.” AUTO-APPROVE TRIVIAL CHANGES AND RUN CONTINUOUS REVIEW ASSISTANTS. 46
45. DEMO GITHUB CODING REVIEW
46. BUILD & CONTINUOUS INTEGRATION CHALLENGES SLOW, FLAKY, OR FAILING BUILDS WASTE DEV TIME. COMPLEX YAML CONFIGS; DEBUGGING CI ERRORS IS TEDIOUS. CONTEXT-SWITCHING BETWEEN CODING AND FIXING BUILD ISSUES. AI OPPORTUNITIES GENERATE CI/CD PIPELINE CONFIGS (“BUILD AND TEST .NET APP”). EXPLAIN BUILD FAILURES AND HIGHLIGHT ROOT CAUSES IN LOGS. SELF-HEALING BUILDS (AUTO- RETRY FLAKY TESTS), PERFORMANCE OPTIMIZATION SUGGESTIONS. 48
47. DEMO CI/CD FIX
48. TESTING (QUALITY ASSURANCE) CHALLENGES WRITING AND MAINTAINING TESTS IS TEDIOUS AND TIME- CONSUMING. FLAKY TESTS AND LONG FEEDBACK LOOPS REDUCE CONFIDENCE. INCOMPLETE TEST COVERAGE CAUSES LATE BUG DISCOVERY. AI OPPORTUNITIES AUTO-GENERATE UNIT TESTS FROM CODE (COPILOT, DIFFBLUE). CREATE TEST DATA, MOCKS, AND INTERPRET FAILING TEST LOGS. DISCOVER MISSING EDGE CASES AND PRIORITIZE TESTS BY RISK. AI-DRIVEN EXPLORATORY TESTING AND SELF-HEALING UI TESTS. 50
49. DEMO USING PLAYWRIGHT MCP
50. SECURITY (DEVSECOPS) CHALLENGES VULNERABILITIES FOUND LATE; HEAVY REWORK. HIGH FALSE POSITIVES FROM SCANNERS . ALERT FATIGUE. MANAGING DEPENDENCY VULNERABILITIES AND PATCHING DELAYS. AI OPPORTUNITIES SMART CODE SCANNING, CONTEXTUAL EXPLANATIONS, AND AUTOFIX FOR VULNERABILITIES (COPILOT + CODEQL). IN-IDE SECURITY HINTS AND NATURAL-LANGUAGE QUERIES (“IS THIS INPUT VALIDATED?”). AUTOMATED PEN TESTING AND POLICY COMPLIANCE VALIDATION. 52
51. DEMO GITHUB ADVANCED SECURITY
52. RELEASE & DEPLOYMENT CHALLENGES MANUAL APPROVAL PROCESSES AND BUREAUCRATIC RELEASE GATES. STRESSFUL ROLLBACKS; CONFIGURATION DRIFT BETWEEN ENVIRONMENTS. MULTI-ENVIRONMENT COMPLEXITY CAUSES “WORKS IN STAGING, FAILS IN PROD” ISSUES. AI OPPORTUNITIES GENERATE RELEASE NOTES AND COMPLIANCE DOCUMENTATION FROM COMMITS. MONITOR METRICS AND DETECT ANOMALIES DURING CANARY DEPLOYMENTS. RELEASE RISK SCORING (“GO/NO- GO” ANALYSIS) USING TEST AND TELEMETRY DATA. AUTO-ROLLBACK AND ROOT CAUSE SUGGESTIONS POST- RELEASE. 54
54. OPERATIONS & SRE CHALLENGES ALERT FATIGUE . TOO MANY NON-ACTIONABLE ALERTS. SLOW ROOT CAUSE ANALYSIS (RCA) AND LOG TRIAGE. REPEATED TROUBLESHOOTING DUE TO POOR DOCUMENTATION. AI OPPORTUNITIES INTELLIGENT ALERT CLUSTERING AND CONTEXTUAL INCIDENT SUMMARIES (E.G., AZURE COPILOT, DATADOG BITS). CHATOPS ASSISTANTS SUMMARIZING INCIDENTS AND RETRIEVING METRICS. ROOT CAUSE CORRELATION ACROSS LOGS, METRICS, AND TRACES. 56
55. DEMO AZURE SRE AGENT
56. Oprah Winfrey 58 • Everybody an AI EVERYBODY AN AI
57. DRIVE – DANIEL PINK AUTONOMY MASTERY PURPOSE
58. PRODUCTIVITY ≠ PURPOSE AS WE GAIN TOOLS THAT MULTIPLY OUTPUT, DO WE REALLY NEED MORE PRODUCTIVITY? OR DO WE NEED MORE MEANING, MORE CREATIVITY, MORE JOY IN THE CRAFT? MAYBE THE TRUE EVOLUTION OF THE DEVELOPER ISN'T ABOUT DOING MORE…BUT ABOUT CHOOSING WHAT’S WORTH DOING.
59. THANK YOU ! RENÉ VAN OSNABRUGGE rene.vanosnabrugge@xebia,com