As AI Advances, Who Is Looking to Its Architecture?

By Andras R. Szakal, VP and CTO, The Open Group

Most enterprise architects will be familiar with – and will probably spend a fair portion of their lives thinking about – some version of Conway’s Law.

Broadly speaking, the law suggests that the structure of any particular system will echo the structure of the organization that designed it. If two distinct teams own the financial data and contact data elements of a customer relationship management platform, for instance, those parts of the system will tend to involve different logic, require different processes, and offer different end-user experiences.

As a way of understanding the relationship between an organization’s structure and its outputs, Conway’s Law might be too simplistic – but, if there is any truth to it, that truth has never been more important than now, when the systems we are designing involve ever more advanced artificial intelligence.

The prizes on offer

The revolutionary, transformative potential of AI has been endlessly discussed over the last year or so. It is not just about an ability to generate competitive advantages or business efficiencies: AI offers a powerful new tool for tackling some of the most significant challenges we face as a species.

It may deliver significant progress in how we monitor and understand the impacts of climate change, where predictive computation is famously complex, resource-intensive, and difficult to verify. Indeed, AI models are already generating promising results in near-term weather forecasts.

It will, likewise, create whole new approaches to reducing the emissions that are causing those impacts in the first place. In some cases, this will be by eliminating the need for emissions-causing activity, such as improving remote conferencing to reduce the incentive of business travel. In others it will be by creating greater efficiency in essential processes, such as routing shipping more effectively.

It can have an important role in both medical diagnosis and treatment development, analyzing data more quickly and exhaustively than human experts to identify likely diagnoses and candidate molecules for pharmaceutical treatment.

It can make the world more accessible through tools like speech recognition and machine translation, and make education more effective by dynamically personalizing learning materials. It can help us to deliver humanitarian aid more effectively and manage energy networks more resiliently.

The list is long – but it is also a work in progress, and not a guaranteed outcome of recent innovations and breakthroughs in the AI space.

A role for standards

There are as many debates about the potential harms of AI as there are predictions about its transformative benefits. However, beyond considerations of things like its military applications, its power to erode societal trust, or its chance of becoming misaligned with our desires, we should also consider how the structures that are designing and deploying it will influence its impact on the world.

Again, if there is any truth to Conway’s Law, great care should be taken now – not when it is too late – to make sure that the structures of the organizations pushing forward what AI can do, how it functions, and where it is applied truly echo the consequences we would like to see AI have.

At The Open Group, we manage and oversee many important standards frameworks and references. Ways of leveraging the power of AI in this space are already well underway: the Data Integration Working Group, in particular, has explored the idea of ‘standards as language models’, in which the breadth and complexity of mature standards is made more accessible and navigable by embedding them in interrogatable conversational interfaces.

There will also be a role for enterprise architects to play as organizations seek to mitigate or minimize the emissions impact, in particular, of AI applications, making sure that the high-power computing resources that AI requires are being used in the most appropriate, efficient way.

There is a case to be made, though, that enterprise architects have a much more fundamental role to play in our current phase of technological evolution than simply implementing its advancements into our workflows. AI solutions must seek to enhance the role of the enterprise architecture and their productivity, not attempt to supplant it.

Standards are important not just because they enable collaboration, but because they build consensus. A successful standard draws on the insights and expertise of the whole community of practitioners which needs to use it. In that process, many conversations are had – and occasionally quite fraught ones – in the interest of finding a common understanding of what a good, mature, responsible, successful approach looks like. One that puts the human at the center of the decision loop.

The point of listing so many of AI’s potential positive outcomes earlier in this article was not just to emphasize how dramatic and wide-ranging its impact could be. It was also to establish how collectively agreeable those impacts are: these are outcomes which transcend competitive disagreements, which, in short, everyone can get behind.

That gives enterprise architects, and the standards community more broadly, something very important to consider going forward.