Information architecture (IA), originally coined by Richard Wurman (1976) during an address at the American Institute of Architecture conference, can be referred to as a professional practice and field of studies focused on solving the problems of accessing, using, and gaining value from the vast amounts of data and information available today. It is also recognized that there is much debate about the definition of and distinction between information architecture and data architecture. While opinions vary, a common trend in the literature suggests information architecture is a discipline that encompasses data architecture which is consistent with the views of Resmini and Rosati (2012) and Downey and Banerjee (2010).  

Perspectives on the components of information architecture can fluctuate given 1) the broad definition, “the need to transform data into meaningful information for people to use” Dillon and Turnbull (2005), and 2) the organizational scope spanning people, process and technology.  Not to mention the intersection and overlap with related disciplines such as enterprise architecture and data architecture.  

Given the rapidly evolving nature of technology and associated disciplines today, information architecture can be described as encompassing a number of the following functions (please note, this is just one of many perspectives):

• Data architecture (master, reference and meta data management, data security, data quality, data orchestration, storage, processing and presentation)

• Analytics architecture (model management, AIOps)

• Infrastructure architecture (technical capabilities to enable the data and analytics lifecycle, e.g., storage, identity, monitoring and service management)

• Data governance (data and information lifecycle management, responsible data uses, privacy, security, policies)

• Knowledge management (data availability, search and retrieval, including communities of practice) 

These foundational components of information architecture are keystone enablers for data and analytics from which organizations derive value, be it accelerated innovation through a mesh or energy line inspection using AI.   

The benefit of artificial intelligence is being observed in many industries (such as KPMG using AI to detect potential fraud orAzure Farmbeats to improve agriculture intelligence), and the potential of capabilities such as Open AI ChatGPT has reinvigorated interest. However, the fundamental capabilities of information architecture and data management practices are often not in place to enable success.  It is also widely acknowledged that the organizational data maturity (Uren, 2020, Carruthers and Jackson, 2022), data management and infrastructure are critical success factors (Westenberger et al., 2022), which is why it is advocated that information architecture should be considered an enabler for artificial intelligence, namely IA before AI. 

But organizations often need to consider the information architecture foundations to support the delivery of AI into production before they embark on AI initiatives. Consider the IA fundamentals, such as data quality. Is the data fit for purpose? Is there a data dictionary? Are provenance and lineage established? 

Other notable AI challenges are observed when IA is an afterthought. Data exploration (to determine AI feasibility and value) is typically the first step. Still, efforts to go beyond proof of value/proof of concept are often impeded by a lack of IA capabilities. For example, failure to ensure a scalable and agile architecture and to address the practicalities of AIOps — the continuous integration and continuous delivery lifecycle, from embedding and integrating AI models into production workflows to AI operationalization and monitoring (security, model telemetry, retraining, and rollback strategy).   

Disregarding these critical IA capabilities (data readiness, dataOps, AIOps, regulatory and ethical concerns, to name but a few) can ultimately result in failure to realize the value of AI. It is, therefore, essential that the data and information foundations, along with the accompanying release management processes and organizational governance frameworks, are established from which organizations can accelerate AI initiatives.  

This is not to say that an organization should embark on a monolithic multi-year data modernization program before realizing the value of AI. Often, these are parallel programs, IA activities being addressed in conjunction with AI exploration — think of this as a bimodal relationship whereby IA and AI resources and workstreams inform the requirements and activities. It is also equally reasonable to approach the problem with a use-case lens which narrows the scope and focus of the work required across IA and AI.  It has been observed that this is a more common tactic since it can accelerate the time to value using a lower-risk lean and agile approach. As such, IA capabilities can organically develop and adapt in tandem with the AI use case exploration and development.  

Let’s also not forget the people and process aspect of information architecture. It is easy to think about the technology enablers, but the supporting functions to manage information architecture must also be present to facilitate AI success. From a practical perspective, this should encompass the following core organizational capabilities: 

• Ensuring that a scalable and flexible infrastructure is established and available

• DataOps and AIOps skills, capabilities and practices are understood and available.  

• Data is cataloged and documented (using a catalog such as Purview or Collibra) 

• Data governance board and processes are established (e.g., privacy, security, and sensitive uses are reviewed. Take a look atMicrosoft’s framework for building AI systems responsibly, which incorporates processes such as impact assessment to identify and mitigate AI risks).  

• Roles and responsibilities are understood; for example, data ownership.  

All these considerations are amplified by the fact that most organizations have a hybrid multi-cloud strategy which introduces numerous complexities such as technology integration, data flows and data estate management across public cloud, private cloud and edge.  

If your information architecture capabilities are unknown, you may want to consider self-assessing using one of the many maturity frameworks such as EDMCouncil DCAM (and CDMC if you are considering cloud adoption) or Gartner Enterprise Information Management Framework – there are others available.  These frameworks address a spectrum of capabilities, from data strategy to analytics management (AI/ML), but evaluating and determining which framework meets your needs would be pertinent. 

In summary, a number of key points are highlighted below to reinforce the importance of information architecture and its relationship to AI:

1. It is critical that organizations address these information and data foundations while also balancing the need to provide agility to deliver and operationalize AI for the business.  

2. Focus on the business outcome(s), engage, align and adapt the information architecture to support the strategy and requirements to drive continuous value and innovation for the business. 

3. Success with AI depends on the use case, data quality and the relevant IA capabilities (infrastructure readiness, data maturity), and the relevant organizational mindset and operating model (Responsible AI framework as an example) to deliver AI success.

4. Where there is uncertainty regarding information architecture, consider assessing the organizational data and information maturity parallel to AI exploration before AI operationalization.

5. Look beyond AI exploration to consider the capabilities required to integrate, deploy, and operationalize your AI models – are DataOps and AIOps well understood and managed today in your organization? 

About the Author

Wright-Jones is a Senior Solution Architect on the Microsoft Industry Solutions Engineering team, working with global organizations to deliver value through cloud computing. He also serves as the worldwide community lead for data and AI solutions, helping others succeed through knowledge sharing and collaboration. Wright-Jones holds a bachelor’s in information systems and a master’s in analytics, and is undertaking a doctorate in data science. Outside of work, Ben enjoys family time with his wife and two eight-year-old children. He is passionate about education and serves as a school governor, ensuring high achievement standards for children ages 2 to 11.