Software Architecture Doesn't Matter Any More - Does It?

A talk given as a keynote at the European Conference on Software Architecture, September 2025, in Limassol, Cyprus.

Contemporary software development presents a paradox: while architectural decisions have become increasingly abstracted from developers, their consequences have never been more significant. Modern software delivery stacks embody significant pre-made architectural choices, agile working emphasises adaptability over upfront design, and patterns like microservices can make detailed models redundant. AI-assisted code generation can further distance practitioners from architectural concerns, contributing to a situation where many practitioners may question the importance of software architecture.

This talk explores how software architecture relates to modern software delivery practice. By considering how we got here, I argue that software architecture remains not only relevant but increasingly critical, although its nature and application may need to evolve significantly from some of the fundamentals of historical approaches. This has significant implications for the focus of software architecture research and how we conduct it, in order to have significant industrial impact.

I will discuss what today's practitioners require from software architecture and why it is more essential than ever. I will also explore how we might align software architecture research with contemporary practitioner needs to maximise its contribution in an environment where architectural decisions need to be made quickly in the face of rapidly changing requirements.

Transcript

1. SOFTWARE ARCHITECTURE DOESN’T MATTER ANY MORE - DOES IT? EOIN WOODS ARTECHRA & IMPERIAL COLLEGE LONDON
2. THE SOFTWARE ARCHITECTURE PARADOX
3. Many developers don’t know much about software architecture Yet today’s digital systems often involve complex architectural tradeoffs … it doesn’t tend to be a focus in professional training … and sophisticated abstractions make them difficult to reason about This is a dangerous disconnect
4. Eoin Woods • Independent consultant • Academic Visitor at Imperial College London • Ex-Chief Engineer at Endava based in London (2015-2025) • 10 years in capital markets - UBS and BGI • 10+ years in products - Bull, Sybase, InterTrust ADDRESSING ENERGY EFFICIENCY IN SYSTEM DESIGN: A JOURNEY FROM ARCHITECTURE TO OPERATION EOIN WOODS A thesis submitted in partial fulfilment of the requirements of the University of East London for the degree of Doctor of Philosophy December 2018
5. DISCLAIMERS ¡ These are the views of one practitioner, not of “industry” ¡ My experience is with information systems of varying types and scales … much of this may not apply to nuclear power station control software
6. So why does this paradox occur?
7. WHY CAN ARCHITECTURE FEEL IRRELEVANT?
8. WHY ARCHITECTURE CAN FEEL IRRELEVANT Cloud Services Cloud Infrastructure • Abstract away complexity • Impose design patterns • Allow building without understanding • Infrastructure becomes coding • Malleable on demand • Perception of being infinitely flexible and changeable
9. WHY ARCHITECTURE CAN FEEL IRRELEVANT “Simple” Styles (microservices, serverless) Standardised Styles • “Simple” styles hide a lot of complexity in the network • Traditional “architecture” diagrams aren’t that useful • Cloud vendors provide architectural styles • Popular industrial styles and fashions (microservices, EDA)
10. CQRS W/MICROSERVICES ON AWS
11. WHY ARCHITECTURE CAN FEEL IRRELEVANT https://www.uber.com/en-in/blog/microservice-architecture
12. WHY ARCHITECTURE CAN FEEL IRRELEVANT AI Generated Code Opensource Frameworks • Fantastically powerful and quick • Sometimes correct • Allow building without understanding • Hide a lot of complexity • Implement a style & patterns • Allow building without understanding
13. WHY ARCHITECTURE CAN FEEL IRRELEVANT Cloud Infrastructure AI Generated Code Opensource Frameworks Cloud Services “Simple” Styles (microservices, FaaS) Standardised Styles ?
14. https://leanpub.com/visualising-software-architecture/read
15. THE CRITICALITY OF SOFTWARE ARCHITECTURE
16. CLASSICAL CONCERNS Stakeholders Quality Attributes Styles and Technologies Decision Tradeoffs Skill & techniques needed to address unclear, conflicting, inconsistent needs Well-understood tactics to achieve QAs, tradeoffs need to be understood and balanced Implications, applicability, tradeoffs need to be understood Need to be assessed accurately in a specific context
17. EMERGING CONCERNS Scale and Complexity LLM based AI Tools Consequences of decisions amplified so decision validation is ever more critical Structure and constraints greatly improve predictability and reduce “errors”
18. THE CRITICALITY OF SOFTWARE ARCHITECTURE Software Architecture Stakeholders LLM-based AI Tools Scale and Complexity Quality Attributes Decision Tradeoffs Styles & Technologies
19. SOFTWARE ARCHITECTURE EVOLVES
20. EVOLVING PRACTICE ”Up Front” Design Incremental Modelling Continuous Architecture Document centric Inhibited agility & adaptation Insights often buried Architect centric Too expensive & slow for fast moving change Few real insights “for free” Team centric Principles, styles, “C4” models, decisions Adaptive & quick(er) Insights often still need “mined”
21. Yet we are still missing some things that developers need …
22. WHAT ARCHITECTURE & ARCHITECTS NEED
23. GETTING MORE VALUE FROM ARCHITECTURE Evidence Techniques Insights Experience-based evidence of what works (or doesn’t) in specific contexts (avoid general results) Simple, robust techniques for assessing and capturing the implications and tradeoffs of decisions Applicability, tradeoffs & implications of architectural tactics, styles and technology choices
24. HOW WE MIGHT DO IT
25. INSIGHTS Gain Test emerging techniques and technologies. What are the tradeoffs? How do they work? (e.g. Jepson) Communicate Practitioner conferences, LinkedIn, accessible open publications, blogs, … Automate …
26. INSIGHTS: AI-POWERED ARCHITECT’S ASSISTANT Credit: Claude.ai (link)
27. INSIGHTS: COMMUNICATION https://www.ozimmer.ch (Olaf Zimmerman) https://jepsen.io/ (Kyle Kingsbury)
28. TECHNIQUES Education Architecture thinking & techniques threaded through CS & SE programmes (not just 4th year electives). Ever more important as AI tools develop. Research Develop (or find) techniques for tradeoff analysis, can be applied, incrementally, to parts of a system, without needing a comprehensive model (or a “tool”)
29. EVIDENCE Case Studies Study popular industrial techniques & technologies • What are they? How do they vary in use? • Who uses them? • What makes them work? And where? Communities Harnessing the “wisdom of crowds” • OWASP have done so quite well, also IASA • Community knowledge of (e.g.) patterns & styles • Researchers bring skills in validation, organisation, communication
30. TO CONCLUDE
31. CYCLICAL INTEREST IN ARCHITECTURE ¡ Software architecture has come into and gone out of “fashion” over the years ¡ At present it seems to be gaining interest again
32. ARCHITECTURE CAN SEEM IRRELEVANT ¡ Mainstream, prescriptive technologies and styles can make it appear that the decisions are all made ¡ Particularly for early career practitioners
33. YET IT’S STILL CRUCIAL ¡ Classical concerns are still important ¡ Scale and complexity have increased ¡ Expectations for reliability, security, speed of delivery have increased
34. A COLLABORATION OPPORTUNITY Industrial Software Architecture Evidence Techniques Insights Rigour Measurement Independence Skepticism Research
35. TIPS FOR SUCCESS ¡ Start with a (real) problem ¡ Constantly check applicability (with relevant people) ¡ Communicate widely in practitioner terms ¡ Make it context specific
36. Thank You – Questions? Dr Eoin Woods Artechra and Imperial College London eoin.woods@artechra.com and e.woods@imperial.ac.uk www.eoinwoods.info