The 90% AI Problem: Why Business Architecture Comes First
Many organisations struggle with Enterprise Architecture; it is often stated that only 20% really succeed. Many struggle just as much with Artificial Intelligence (AI); it is often stated that less than 10% really get the intended benefit. Yet the real challenge does not lie in either discipline on its own. It emerges when both are expected to work together. Without a shared foundation of understanding, intent, and structure – then two wrongs will not make a right.
More than nine out of ten AI initiatives fail to deliver real business value. AI is a tremendous opportunity to scale – or fail. Historically, Enterprise Architecture has focused heavily on models and frameworks, often overlooking the delivery of stakeholder and business value in ways that are easy to understand. But when properly combined, when architecture precedes intelligence, then the success rate changes dramatically.
That is where a new generation of approaches, often described as Architecture as Code, begins to matter as opposed to documentation and designs.
The Illusion of Comparability
One of the most persistent challenges in Enterprise Architecture is the tendency to compare fundamentally different tools as if they solve the same problem. Industry analysts rank platforms along standardised dimensions in magic quadrants, creating the impression that selection is simply a matter of scoring highest. It is not.
It is akin to stepping into a kitchen and comparing a spoon, a refrigerator, a grill, and an oven on a single scale of value. The exercise can be performed. Criteria can be defined. Rankings can be produced. Yet the premise remains flawed.
These architecture tools exist to fulfil different purposes at different stages of value creation. Without clarity on intent, whether the goal is to understand, to design, to govern, to document, or to execute – then any comparison becomes detached from reality.
Modern platforms such as Next-Insight take a fundamentally different stance. Rather than positioning themselves purely as modelling tools, they operate as decision-centric environments where architecture is not documented after the fact, but actively used to steer business outcomes with one consumable portal for stakeholders of mid to large companies.
EA Has Changed Fundamentally
Enterprise Architecture today is not what it was ten years ago. Historically, EA was often perceived as a documentation exercise, stacks of diagrams around a repository of processes, systems, and capabilities. Valuable, but rarely decisive.
Today, EA has become the foundation for understanding the modern digital enterprise. It is no longer about describing how the organisation works in drawings and documents. It is about digitally enabling how the organisation should operate in a world defined by faster changes and increasing amount of structured data, and intelligent systems that can learn from this to augment decision-support. This shift is subtle but profound.
Understanding the enterprise is not the end state. It is the entry point. Value emerges only when that understanding is translated into prioritisation, decision-making, and governed execution. It is the usage of structured information to align decisions that really allows execution of business value.
The Second Misunderstanding: Treating AI as One Thing
At the same time, organisations oversimplify Artificial Intelligence. AI is often discussed as if it were a single capability. In reality, it is a collection of fundamentally different forms of intelligence:
- Predictive intelligence (forecasting outcomes)
- Classificatory intelligence (structuring information)
- Optimisation intelligence (evaluating trade-offs)
- Generative intelligence (producing plausible content)
- Autonomous intelligence (acting within defined constraints)
These are not variations of one toolset. They are distinct ways of shaping decisions and outcomes. When organisations fail to distinguish between them, they apply the wrong type of intelligence to the wrong problem, and that is where most initiatives lose impact.
From AI Technology to Business Intelligence Design
The real shift occurs when organisations stop discussing AI as technology and start defining it as business behaviour. Instead of asking “Can we use AI?”, the question becomes:
“What type of intelligence is required, in which business capability, to achieve which strategic outcome?”
This is precisely where Enterprise Architecture becomes critical and foundational for AI. Platforms like Next-Insight enable this transition by embedding intelligence design directly into architectural thinking. They allow organisations to connect strategy, capabilities, and AI behaviour in a single coherent structure that is the core knowledge hub; ensuring that initiatives are not isolated experiments, but aligned transformations.
Where the Few Succeed
The organisations that succeed with EA-guided AI share a set of common characteristics. They begin with Business Architecture, first identifying where work accumulates, where processes repeat, and where decisions can be measured and improved.
They then determine which type of intelligence best fits each context.
- Stable, repetitive domains are optimised through automation and predictive models
- Structured complexity is improved through classification
- Decision-heavy environments are supported — not replaced — by optimisation and augmentation
Critically, they also understand where not to apply AI. Enterprise Architecture operates in domains defined by ambiguity, trade-offs, and long-term consequences. In these areas, the misuse of AI does not simply reduce efficiency, it easily distorts accountability.
The best organisations design AI to augment human judgement, not replace it.
The Role of EA in an AI-Driven World
Enterprise Architecture is no longer a supporting discipline. It is the mechanism through which intelligence is governed. It defines:
- Where different types of intelligence belong
- How they should behave
- What outcomes they are expected to produce
- How those outcomes are monitored and controlled
Without this structure, AI remains experimental. With it, AI becomes strategic.
This is why the principle of “EA before AI” is increasingly decisive, and why integrated platforms such as Next-Insight are emerging at the centre of modern transformation efforts. They operationalise architecture, turning it into a live, decision-driving capability rather than a static repository or isolated tech experiments with little business outcome.
What Differentiates the Minority
The minority that succeeds understands something fundamental. They do not treat AI as a goal. They treat it as a consequence. They do not start with AI tools. They start with intent. They do not stop at understanding. They continue until that understanding produces measurable outcomes.
They know whether they are describing, designing, or delivering value, and they align both architecture and intelligence accordingly.
The Transformation Bottom Line
AI is a great opportunity for transforming businesses. It is a loss not to use it, but the question is type and where and how much. Enterprise Architecture exists to shape AI. In the age of AI, its role is to ensure that intelligence is applied deliberately, appropriately, and with accountability.
That is what separates experimentation from execution. That is what turns AI initiatives into business outcomes. That is why organisations that combine modern Enterprise Architecture with a clear intelligence design approach, supported by platforms like Next-Insight and experienced advisors, are far more likely to achieve a safe and successful AI transformation. They use structured enterprise knowledge as the foundation and augment it with prediction, recommendations, and gap analysis rather than replacing it.
If you want to move beyond isolated AI experiments and towards strategic, outcome-driven transformation, the journey starts with Business Architecture and an understanding of how to use it effectively. Let’s connect if you require assistance getting started with EA for AI and for governing AI at scale. Book a talk here.
