Andrew Harmel-Law presented the architecture advice process at GOTO Copenhagen. How we decide is at the core of architecture, and the architecture advice process is a way to decentralize architectural decisions. It needs to be supported by Architecture Decision Records because of the speed at which technology and systems move, and can be complemented by a weekly architecture advice forum.
Writing is thinking, and therefore writing an Architectural Decision Record (ADR) helps those deciding in this practice, Harmel-Law argued. They help deciders make small, reasoned decisions, supported by the expertise of their peers. That also helps those less experienced in deciding to learn this art.
Lightweight ADRs allow the right people to be involved in the right decision-making at the right time. Offering written advice asynchronously allows different minds time to think in the way that suits them best, Harmel-Law said.
When ADRs are used to back up the architecture advice process, they combine to form an immutable change log for the overall system architecture, Harmel-Law explained:
This represents both a granular record of the path taken to get to where the architecture is now, with all its gory and perhaps unpalatable context, as well as a learning resource for those both onboarding into a specific organization as well as for anyone desiring to learn how to think and reason architecturally.
The architectural advice forum is a regular session open to all, which primarily serves to build trust in the advice process, make explicit the decisions that are happening following it, and to allow offering advice and discussing synchronously, Harmel-Law mentioned. It allows those less experienced in deciding to see this process (which is in older approaches frequently adversarial and undertaken behind closed doors) happening in front of them.
The architectural advice forum shines a light on the practice of architecture and democratises it. Not everyone will be inclined to attend, but they can, Harmel-Law explained:
Not all attendees will actively participate, but they are able to. Only some will step up and present decisions and ADRs, but those that do will be able to harvest the knowledge and support of the collective. In a world that needs more architecture, not less, this is incredibly powerful.
Failures with the advice process and lightweight ADRs arise mainly because of old "top-down architectural practice" mental models and a valuing of knowledge stocks ("what I know") over knowledge flows ("what everyone who needs it knows"), Harmel-Law explained:
The issue that underpins them all is a lack of trust. In order for the advice process to work and prevent architects from becoming bottlenecks, the power to decide needs to be spread more widely. If those with that power are unwilling to share it, then none of the benefits can accrue.
Harmel-Law mentioned that he sees traditional architects worrying about "bad" decisions, being the ones they disagree with. This is a failure to acknowledge that no one knows whether a decision will turn out well or poorly until it is put into action.
Feedback is the only way to know, and so deciding small and fast, and getting feedback on the outcomes of that decision rapidly, are key to avoiding this issue, Harmel-Law concluded.
InfoQ interviewed Andrew Harmel-Law about the advice process, ADRs, and doing architecture.
InfoQ: How can approaches like the advice process and lightweight ADRs fail?
Andrew Harmel-Law: I frequently see issues where those whom the advice process now allows to decide, choose not to exercise this power as they are afraid of the ramifications. In these circumstances, those more experienced need to not take up the space the advice process opens up, and instead work with those less experienced to build their confidence and skills.
I also see decisions that fly under the radar. This can be for various reasons (typically those deciding don’t realise it’s a decision, or they do, but are afraid of being told "no" by others, despite the fact that the architecture advice process prevents that – they only need to seek advice, they maintain full responsibility (and accountability) for their decisions.
This can be remedied by those with more experience in the practice of architecture mentoring others. This is a great way to again turn knowledge stocks (the experience of the traditional architect) into knowledge flows (sharing that experience with others as and when they need it.).
InfoQ: Is there still value in doing architecture?
Harmel-Law: The value of architecture, at all scopes and scales, is far more important than ever. With the technologies and tooling we have these days (even before LLMs came on the scene) and the number of people working on these systems, both within and outside of our firewalls, architecture (including sociotechnical architecture - who works on what) is becoming more important than ever.
We feel it when it goes wrong - just look at the mess many of us got ourselves into with microservices. It was supposed to enable team autonomy, but instead led to way too many distributed monoliths. Or think about ensuring our systems embody the systemic qualities we expect. Or the fact that our systems are constantly evolving or being put to uses that were never initially imagined.
Software is very powerful; it is capable of solving almost any problem we can throw at it. But making sure it’s a good fit for that is a lot of work. That’s architecture.