Share
Architectural Intelligence: Ist AI die bessere Softwarearchitekt:in
Wie heißt es so schön? Nicht die KI wird dir den Job nehmen, sondern diejenigen, die verstehen, wie man sie nutzt. Übertragen auf das Berufsbild der Softwarearchitekt:innen bedeutet das, sinnvolle KI-Szenarien zu entwickeln, die uns dabei unterstützen, unsere Arbeitsweise als Architekt:innen und die von uns zu fällenden Architekturentscheidungen zu verbessern. Leichter gesagt als getan. Denn dazu gilt es zunächst einmal die Möglichkeiten und Grenzen von KI zu verstehen und das, was uns als Mensch grundlegend von ihr unterscheidet. Was kann die KI bereits heute leisten und was eher nicht? Welche KI-basierten Tools können uns bei der Architekturarbeit unterstützen? Inwieweit sind die Ergebnisse der KI dabei belastbar und vertrauenswürdig? Und last but not least die alles entscheidende Frage: Liefern wir Menschen am Ende überhaupt noch den entscheidenden Mehrwert?
Lars Roewekamp PRO
Transcript
124 slides
-
#WISSENTEILEN Software Architecture meets AI Architectural Intelligence Lars Röwekamp |
open knowledge GmbH | lars_roewekamp -
(Architecture, Cloud, AI & ML) Lars Röwekamp in: lars_roewekamp CIO
New Technologies OPEN KNOWLEDGE -
„We are no longer just writing code for the next
developer – we are structuring systems for the next AI agent to execute.“ Arup Saha, Senior Director RBC, Canada -
„The engineers who thrive from here won’t just be the
ones who write the best code; they’ll be the ones who build the best systems for AI to work within.“ Arup Saha, Senior Director RBC, Canada -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation „The six core activities for software architects.“, ISAQB -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Wie kann uns Agentic AI dabei helfen? Was kann Agentic AI gut und was nicht so gut? -
Das neue Normal ein Selbstversuch “Poor Man‘s“ Edition alternativ Aider,
OpenCode, Gemini CLI, Cursor, GitHub Copilot, Cline …. -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation „That sounds like a great project!“ #WTF*, du bist eine Maschine! *(what a terrible future) -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation „…, let‘s define key … .“ Was heißt hier „let us“? -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation potenzielle fachliche Anforderungen benennen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation potenzielle fachliche Anforderungen benennen Unique set of requirements? -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation potenzielle fachliche Anforderungen benennen Some pleasant surprises included, too! -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation potenzielle fachliche Anforderungen benennen Some pleasant surprises included, too! -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation potenzielle nicht-fachliche & technische Anforderungen benennen large, many, fast, … not very meaningful -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation nützliches Bonus Material (du hast zwar nicht danach gefragt, aber …) Reasoning via for, if, … Explaining „business“ -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation OMG! This is great, isn‘t it? nützliches Bonus Material (du hast zwar nicht gefragt, aber …) -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Lücken & Mehrdeutigkeiten in Anforderungen erkennen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Evaluation
of Architectures Communication of Architectures Design of Structures Support during Implementation Building Blocks identifizieren und Strukturen ableiten -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Evaluation
of Architectures Communication of Architectures Design of Structures Support during Implementation Building Blocks identifizieren und Strukturen ableiten Good Service Design? Good Module Design? -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Evaluation
of Architectures Communication of Architectures Design of Structures Support during Implementation Building Blocks und Strukturen visualisieren -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Evaluation
of Architectures Communication of Architectures Design of Structures Support during Implementation Building Blocks und Strukturen visualisieren -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Evaluation
of Architectures Communication of Architectures Design of Structures Support during Implementation Building Blocks und Strukturen visualisieren -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Übergreifende Konzepte „entwerfen“ und dokumentieren -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Communication of Architectures Evaluation of Architectures Support during Implementation Architektur bewerten gemäß einer vorgegebenen Methode 1. Present ATAM 2. Present business drivers 3. Present the architecture 4. Identify architectural approaches 5. Generate quality attribute utility tree 6. Analyze architectural approaches 7. Brainstorm and prioritize scenarios 8. Analyze architectural approaches (with the added) 9.Present results -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Communication of Architectures Evaluation of Architectures Support during Implementation Architektur bewerten gemäß einer vorgegebenen Methode -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Communication of Architectures Evaluation of Architectures Support during Implementation Architektur bewerten gemäß einer vorgegebenen Methode -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Communication of Architectures Evaluation of Architectures Support during Implementation Architektur bewerten gemäß einer vorgegebenen Methode -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Template für Architekturdokumentation vorschlagen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Dokumentation auf Basis vorhandener Quellen erzeugen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Dokumentation auf Basis vorhandener Quellen erzeugen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Dokumentation auf Basis vorhandener Quellen erzeugen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Dokumentation auf Basis vorhandener Quellen erzeugen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Source Code auf Basis von Beschreibungen generieren -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Basis für den ersten Service generieren -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: Value Objects und Builder Pattern -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: Domain Events and Event Publisher -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Bug Fixing: Tests laufen nicht (mehr) korrekt -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Bug Fixing: Tests laufen nicht (mehr) korrekt -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: aktuelle Java Features nutzen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: JPA einbinden (bisher nur In-Memory Persistence) -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: JPA Testing via Quarkus & Testcontainer -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: Bootstraping mit generierten Demo Daten -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: weiteren Service erzeugen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: weiteren Service erzeugen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: weiteren Service erzeugen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: Docker Capabilities hinzufügen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: Docker Capabilities hinzufügen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: Docker Capabilities hinzufügen -
Clarification of Requirements and Constraints Design of cross-sectional Concepts Design
of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Refactoring: POSTMAN Collection -
Das neue Normal ein Selbstversuch Clarification of Requirements and Constraints
Design of cross-sectional Concepts Design of Structures Evaluation of Architectures Communication of Architectures Support during Implementation „The six core activities for software architects.“, ISAQB -
Das neue Normal ein Selbstversuch Clarification of Requirements and Constraints
Design of cross-sectional Concepts Design of Structures Evaluation of Architectures Communication of Architectures Support during Implementation WTF*? Advanced Vibe Coding? Agentic Software Engineering Light? *what a terrific (or terrible?) future -
Das neue Normal ein Selbstversuch Clarification of Requirements and Constraints
Design of cross-sectional Concepts Design of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Egal, das war schon irgendwie cool, aber … -
Das neue Normal ein Selbstversuch Clarification of Requirements and Constraints
Design of cross-sectional Concepts Design of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Egal, das war schon irgendwie cool, aber … ich habe da kein gutes Bauchgefühl! -
Das neue Normal ein Selbstversuch Clarification of Requirements and Constraints
Design of cross-sectional Concepts Design of Structures Evaluation of Architectures Communication of Architectures Support during Implementation Egal, das war schon irgendwie cool, aber … wieso kann die AI das alles überhaupt? -
„AI Agents are systems that combine LLMs for reasoning and
decision-making with tools for real-world interaction, enabling them to complete complex tasks with limited human involvement.“ AI Agents Kind of Definition -
„Ein Agent ist ein LLM, das Tools und Memory in
einer Schleife nutzt.“ Robert Glaser, Head of Applied AI at Exxeta AI Agents Kind of Definition -
„Ein Agent ist ein LLM, das Tools und Memory in
einer Schleife nutzt.“ Robert Glaser, Head of Applied AI at Exxeta AI Agents Kind of Definition -
… Frage? … Antwort! MODEL LLM PLAN ACT REFLECT AI
Agents von GenAI zu Agentic AI -
… Aufgabe! PLAN Agent … Resultat. OK NOT OK PLAN
ACT REFLECT AI Agents von GenAI zu Agentic AI -
AGENT MODEL Tools Long-Term Memory Vector Datastore Execution Loop in
out Plan Action Memory [ST] Tools Content Data Devices Code Services Human APPLICATION Function Calling LLM FLYWHEEL Improve (eval & guard) Reason (plan & reflect) Instruct (Task & Role) Act (Tools & Resources) Remember (Short-Term & Long-Term) AI Agents The Big Picture -
AGENT MODEL Tools Long-Term Memory Vector Datastore Execution Loop in
out Plan Action Memory [ST] Tools Content Data Devices Code Services Human APPLICATION Function Calling LLM FLYWHEEL A #2 A #n ... Scale-up /-out (parallel & specialized) Agentic AI AI Agents The Big Picture -
AGENT MODEL Tools Long-Term Memory Vector Datastore Execution Loop in
out Plan Action Memory [ST] Tools Content Data Devices Code Services Human APPLICATION Function Calling LLM FLYWHEEL A #2 A #n ... Agentic AI Agentic Software Engineering? -
Unser Ziel: Die AI mit dem relevanten Wissen und Tooling
versorgen. (Damit sie unsere Aufgaben wie gewünscht erfüllen kann.) -
Foundation of Agentic Software Engineering Ich habe ein schlechtes Bauchgefühl?
Aber in Bezug auf was eigentlich? • Qualität • Erklärbarkeit • Reproduzierbarkeit • Kostentransparenz • Vendor / Model / Tool Lock-in • Skalierung auf Team-Kollegen -
„Was muss ich dafür tun, um von dem schlechten Bauchgefühl
wegzukommen und Agentic AI produktiv und mit gutem Gewissen in allen Phasen des Software Development Lifecycle einsetzen zu können?“ Foundation of Agentic Software Engineering -
Context Engineering Agentic Validation Agentic Tooling Agentic Codebase Compound Engineering
Gib der KI genau das, was sie tatsächlich braucht. Lass die KI ihre eigene Arbeit überprüfen. Hör auf, dich einzumischen. Fang an aufzubauen. Optimiere für eine AI-basierte Lesbarkeit. Nutze die Arbeit im Team als „Flywheel“. 1 2 3 4 5 Foundation Pillars of Agentic Software Engineering -
Context Engineering Agentic Validation Agentic Tooling Agentic Codebase Compound Engineering
Foundation Pillars of Agentic Software Engineering Gib dem AI Model genau das, was es tatsächlich braucht.* • Leitplanken für Architektur, Security, Observability • „maschinenlesbare“ Notizen und Kommentare im Code • Methodiken, Patterns, Best Practices und guten Beispiel-Code • Kopfwissen und ungeschriebene Regeln • Noise-Elemination ACHTUNG: Weniger ist deutlich mehr! *nicht weniger aber auch nicht mehr -
Context Engineering Agentic Validation Agentic Tooling Agentic Codebase Compound Engineering
Foundation Pillars of Agentic Software Engineering Lass die AI ihre eigene Arbeit überprüfen • AI-basierte flächendeckende Testgenerierung • Visual Checks via Screenshots • Interaction Testing via App / Browser / Device Remote Control • Data Integrity via State-based Semantic Validators • LLM / Agent-spezifische Metrics für Monitoring und Alerting ACHTUNG: Auch ein Agentic Validation System kann lernen! -
Context Engineering Agentic Validation Agentic Tooling Agentic Codebase Compound Engineering
Foundation Pillars of Agentic Software Engineering Hör auf dich einzumischen. Fang an nachhaltige AI-Strukturen aufzubauen. • Bei manuellen Eingriffen wird der Mensch zum Bottleneck. • Nicht das Problem lösen, sondern die Ursache. • Fähigkeiten des Agenten-basierten Systems gezielt erweitern. • Agent-Workflows und Agent-Tooling zur Verbesserung / Erweiterung. ACHTUNG: Total Cost / Time of Fix-a-Problem minimieren! -
Context Engineering Agentic Validation Agentic Tooling Agentic Codebase Compound Engineering
Foundation Pillars of Agentic Software Engineering Optimiere deine Code Base für die AI-Lesbarkeit. • Mehrdeutigkeit im Code verlangsamt und irritiert. • Code um „toten“ Code bereinigen. • Konkurrierende Pattern vereinheitlich. • Golden-Rules des Projektes dokumentieren. • Bilderbuch-Beispiele bereitstellen. ACHTUNG: Die AI lernt von dem, was sie sieht und (nicht) versteht! -
Context Engineering Agentic Validation Agentic Tooling Agentic Codebase Compound Engineering
Foundation Pillars of Agentic Software Engineering Nutze die Arbeit im Team als „Flywheel“. • AI-Tools als Team-Infrastruktur - nicht als persönliche Hacks. • Gemeinsame Governance und gemeinsame Leitplanken. • Gemeinsamer Context sichert einheitliches Verhalten. • Gemeinsames Tooling vermeidet individuelle Reibungsverluste. • Gemeinsame Code-Hygiene liefert Nachhaltigkeit. ACHTUNG: Es wird Zeit für eine Agentic (Internal) Developer Platform! -
Context Engineering Agentic Validation Agentic Tooling Agentic Codebase Compound Engineering
Gib der KI genau das, was sie tatsächlich braucht. Lass die KI ihre eigene Arbeit überprüfen. Hör auf, dich einzumischen. Fang an aufzubauen. Optimiere für eine AI-basierte Lesbarkeit. Nutze die Arbeit im Team als „Flywheel“. Die 5 neuen Tugenden des ASE aka „Das Agentic Software Engineering Framework“ -
Context Validation Tooling Codebase Compound „Das kommt mir irgendwie bekannt
vor.“ Die 5 neuen Tugenden des ASE aka „Das Agentic Software Engineering Framework“ -
„Die 5 Tugenden des Agentic Software Engineerings. Klingt erst einmal
sinnvoll. Aber was muss ich bzw. mein Team dafür tun, um das Framework nachhaltig in unserem Software Development Lifecycle zu etablieren?“ Adoption Process of Agentic Software Engineering -
Efficient ASE Adoption Process A step-by-Step Approcach Experimental Repeatable Defined
Capable Level 1: Level 2: Level 3: Level 4: Level 5: AI Autonomie -
ASE Adoption Process Warum brauchen wir das? Wir schaffen so
eine Standortbestimmung, gemeinsame Sprache und realistische Erwartungshaltung in Bezug auf: • AI-Strategie und -Erfahrung • AI-Governance und Security • Technologie und Daten • Organisation und Kultur • Unternehmensstrategie -
Experimental Repeatable Defined Capable Efficient #1 #2 #3 #4 #5
Agenten-basierte Initiativen sind ungeplant, individuell und stark experimentell. Der Mensch generiert via Vibe Coding Source Code und ggf auch ein wenig mehr (z.B. Dokumentation, Tests). Die Ergebnisse sind inkonsistent, isoliert, nicht reproduzierbar und stark abhängig von einzelnen Personen. Web-basierte AI-Chats, Plug-Ins für IDEs, rechte Maustaste, Coding-Assistants* ASE Adoption Process Level 1 of 5: Experimental aka Freestyle *lediglich als Command Line Interface -
Experimental Repeatable Defined Capable Efficient #1 #2 #3 #4 #5
Gemeinsame Muster, Praktiken und Best Practices entstehen, die im Team untereinander gezielt geteilt werden. Das Team nutzt einheitliches Tooling und einigt sich auf ein minimales Set(up) von „Governance“, welches gelebt und rudimentär Tool-spezifisch hinterlegt wird (claude.md, instruction.md, config, rules). Claude Code, GitHub Copilot, Cursor, Gemini Code Assistant, Open Code, JetBrain Junie ASE Adoption Process Level 2 of 5: Repeatable -
Experimental Repeatable Defined Capable Efficient #1 #2 #3 #4 #5
Das AI-Vorgehen wird gemeinsam klar definiert und von einer ASE-Initiative, Governance und Standards unterstützt. Die gemeinsame Governance wird ausgebaut und für wiederkehrende Aufgaben automatisiert (Refactoring, UI, Documentation, Testing, Security-Checks). Die Eingaben für die AI werden mit Hilfe von AI erzeugt (Spec-Driven Dev). Prompt-Templates, Pre-build oder Custon Skills, SpecKit, OpenSpec ASE Adoption Process Level 3 of 5: Defined -
Experimental Repeatable Defined Capable Efficient #1 #2 #3 #4 #5
Die Nutzung von AI wird deutlich über die Code-Generierung hinaus auf alle Bereiche des SDLC ausgedehnt. Die AI findet Einzug in das Requirement Engineering, in die CI/CD-Pipeline und die Laufzeit (z.B. Observability). Die AI bekommt dazu die notwenigen Zugriffe auf 3rd Party Systeme und Tools Der Mensch wird zum fachlichen* Assistenten und Kontrolleur der AI. 3rd Party Tools via MCP, 3rd Party Agents via A2A. ASE Adoption Process Level 4 of 5: Capable *auch DevOps besitzt eine Fachlichkeit ;-) -
LLM APPLICATION SOME API MCP Sever MCP Sever FILESYSTEM DATABASE
MCP Sever MCP Client MCP Client MCP Client MCP HOST JSON-RPC 2.0 stdio / HTTP + SSE Tools Tools Tools Model Context Protocol aka Reliable Standard -
LLM APPLICATION SOME API MCP Sever MCP Sever FILESYSTEM DATABASE
MCP Sever MCP Client MCP Client MCP Client MCP HOST Context Tools Prompt Context Tools Prompt Context Tools Prompt Roots Sampling Elicitation Roots Sampling Elicitation Roots Sampling Elicitation Model Context Protocol aka Reliable Standard -
Agent2Agent Protocol Open Standard 4 Agent Collabortion Source: https://a2a-protocol.org/latest/topics/what-is-a2a/ Client-2-Remote
Agent 1. Capability Discovery 2. UX Negotiation 3. Task & State Management 4. Collaboration -
Experimental Repeatable Defined Capable Efficient #1 #2 #3 #4 #5
Der gesamte Software Development Lifecyle ist als „Agent-first“ organisiert. Der Mensch gibt zu generierende Fachlichkeiten vor und agiert nur noch über Intents / Goals und auf gezielte Rückfragen der AI. Der Mensch steuert die finale Freigabe der generierten Artefakte / Anwendung. v0, Lovable, Emergent, Replit oder ein sauberes individuelles ASE-Setup ASE Adoption Process Level 5 of 5: Efficient -
Experimental Repeatable Defined Capable Efficient #1 #2 #3 #4 #5
Der gesamte Software Development Lifecyle ist als „Agent-first“ organisiert. ASE Adoption Process Level 5 of 5: Efficient Oha, kann ich das mit dem sauberes individuelles ASE-Setup bitte noch einmal sehen? -
„Menschen wissen, was sie tun. Maschinen tun, was sie wissen.“
Lars Röwekamp, CIO New Technologies, open knowledge -
Der MENSCH bringt Kreativität, Urteilskraft, Verantwortung und ethisches Denken ein
und sorgt so für die gewünschte QUALITÄT. Die MASCHINE liefert Weltwissen, Vernetzung, Rechenpower, Skalierbarkeit und sorgt so für eine nahezu unendliche QUANTITÄT. -
SE for Humans SE for Agents Mensch & Maschine Miteinander
statt gegeneinander -
angelent an: „Agentic Software Engineering: Foundation Pillars and Research Roadmap“
von Ahmed E. Hassan et al. (https://arxiv/pdf/2509.06216) ACE Agent Command Environment AEE Agent Execution Environment Mensch & Maschine Miteinander statt gegeneinander SE for Humans SE for Agents Die Maschine als Coding-Factory Der Mensch als Coach des Agent Mesh. -
angelent an: „Agentic Software Engineering: Foundation Pillars and Research Roadmap“
von Ahmed E. Hassan et al. (https://arxiv/pdf/2509.06216) Agent Command Environment Agent Execution Environment Mensch & Maschine Miteinander statt gegeneinander SE for Humans SE for Agents Humans Agentic Guidance Human Workbench Agent Workbench AI Teammate Lifecycle AI Teammate Infrastructure das „Miteinander“ -
Create Govern based on: „Agentic Software Engineering: Foundation Pillars and
Research Roadmap“ von Ahmed E. Hassan et al. (https://arxiv/pdf/2509.06216) Mensch & Maschine Miteinander statt gegeneinander Humans Agentic Guidance Human Workbench Agent Workbench AI Teammate Lifecycle AI Teammate Infrastructure Agents Briefing Script What? SE for Humans SE for Agents Artifacts as Interface -
Create Create Govern based on: „Agentic Software Engineering: Foundation Pillars
and Research Roadmap“ von Ahmed E. Hassan et al. (https://arxiv/pdf/2509.06216) Mensch & Maschine Miteinander statt gegeneinander Humans Agentic Guidance Human Workbench Agent Workbench AI Teammate Lifecycle AI Teammate Infrastructure Agents Briefing Script Loop Script What? How? SE for Humans SE for Agents Artifacts as Interface -
Create Create Create Govern based on: „Agentic Software Engineering: Foundation
Pillars and Research Roadmap“ von Ahmed E. Hassan et al. (https://arxiv/pdf/2509.06216) Mensch & Maschine Miteinander statt gegeneinander Humans Agentic Guidance Human Workbench Agent Workbench AI Teammate Lifecycle AI Teammate Infrastructure Agents Briefing Script Loop Script Mentor Script What? How? Why? SE for Humans SE for Agents Artifacts as Interface -
Create Create Create Govern Produce Reviewed in based on: „Agentic
Software Engineering: Foundation Pillars and Research Roadmap“ von Ahmed E. Hassan et al. (https://arxiv/pdf/2509.06216) Mensch & Maschine Miteinander statt gegeneinander Humans Agentic Guidance Human Workbench Agent Workbench AI Teammate Lifecycle AI Teammate Infrastructure Agents Briefing Script Loop Script Mentor Script Consultation Request Pack What? How? Why? SE for Humans SE for Agents Artifacts as Interface -
Create Create Create Govern Produce Reviewed in based on: „Agentic
Software Engineering: Foundation Pillars and Research Roadmap“ von Ahmed E. Hassan et al. (https://arxiv/pdf/2509.06216) Mensch & Maschine Miteinander statt gegeneinander Humans Agentic Guidance Human Workbench Agent Workbench AI Teammate Lifecycle AI Teammate Infrastructure Agents Briefing Script Loop Script Mentor Script Merge Readiness Pack Consultation Request Pack What? How? Why? SE for Humans SE for Agents Artifacts as Interface -
Create Create Create Govern Produce Reviewed in Respond with based
on: „Agentic Software Engineering: Foundation Pillars and Research Roadmap“ von Ahmed E. Hassan et al. (https://arxiv/pdf/2509.06216) Mensch & Maschine Miteinander statt gegeneinander Humans Agentic Guidance Human Workbench Agent Workbench AI Teammate Lifecycle AI Teammate Infrastructure Agents Briefing Script Loop Script Mentor Script Merge Readiness Pack Consultation Request Pack Version Controlled Resolution What? How? Why? SE for Humans SE for Agents Artifacts as Interface -
„The traditional path is faster on day one. The agentic
path pays back every day after.“ -
„This isn‘t about replacing yourself; it‘s about building systems that
make you dramatically more capable.“ -
„The software architect will not be replaced by AI but
by another software architect who makes smart use of AI.“ -
#WISSENTEILEN Vielen Dank! #WISSENTEILEN by open knowledge GmbH @_openKnowledge |
@mobileLarson Lars Röwekamp, CIO New Technologies -
#WISSENTEILEN #WISSENTEILEN BILDNACHWEIS Folie 21: © photoplotnikov - istockphoto.com Folie
23: © Mix und Match Studios - shutterstock.com Folie 23: © Mix und Match Studios - shutterstock.com All other pictures, drawings and icons originate from • pexels.com, • pixabay.com, • unsplash.com, • flaticon.com or are created by my own.