Why is focusing on implementation tools (the 'hype') the wrong approach?
The standard initial conversation clients have is often, 'What AI tools should I implement in my organisation?' However, experts suggest this is the incorrect question. The correct question is "How to create the right ecosystem for AI to thrive and evolve?' We call it AI organisational resilience.
Focusing solely on tools overlooks the fact that AI is not something you can simply 'bolt on'. AI is best viewed as the tip of a pyramid, where everything underneath must be sturdy, well-designed, and properly operated for AI to reach its full potential.
Want to learn more about how to use AI to grow your business?
How does the current AI revolution fundamentally differ from the evolutionary IT development of the past 20 years?
For the past two decades, IT development has mostly been evolutionary, rewarding the slow and steady approach of building upon past achievements.
The introduction of AI marks a paradigm shift, representing a technological revolution that will significantly change the landscape of how work is performed. It requires companies to be open to operating in a new reality.
Key differences include:
- Technology as innate business: Tech is no longer merely complementary or a back office function; it becomes an innate part of the business ecosystem. Every company is now a tech company. Companies must adapt tech to their day-to-day operations to remain competitive.
- Need for managed risk: Previously rewarded approaches favored building on the past; now, companies must be open to operating in a new reality where managing risk responsibly is a part of the business. Success requires balancing high-risk and low-risk baskets simultaneously.
- Non-deterministic systems: Unlike previous tools (like JIRA or coding support) where a defined input yields the exact same output, AI is a non-deterministic system. With the same input, you won’t always get the same results, making scaling AI implementations extremely challenging without proper guardrails. That makes scaling AI implementation extremely difficult, necessitating the establishment of guardrails within the company to manage this unknown factor.
Transform into an AI-boosted business.
Discover how our services will cut costs, improve productivity, test your ideas, and maximise ROI.
What specific foundational elements are necessary to ensure our IT architecture is truly AI-ready?
The necessary groundwork is often referred to as ‘grunt work’ or ‘cleaning up your house’. Building a solid foundation involves establishing good data, good governance, and good boundaries.
These strong foundations are crucial because if you want to integrate AI, you need to reduce the risk of AI multiplying existing organisational mistakes, not only on initial implementation but also in the future once it’s widely adopted.
If companies try to implement AI before achieving this foundation, they risk falling victim to the AI hype wave without achieving sustainable business wins.
A tailored measurement strategy aligning Key Performance Indicators (KPIs) with organisational objectives, system architecture, and maturity levels is a must.
What is the ‘Total Metrics-Driven Approach’ and why is it critical for achieving AI impact?
The Total Metrics-Driven Approach (TMD) means that every bit and every element of your business processes can actually be measured.
A common mistake is that while boards have high-level KPIs (e.g., for revenue), the underlying business processes that generate that revenue often lack metrics.
Without establishing metrics and KPIs on the process level, improvements are often based merely on intuition (‘I do feel that this process is working better’) rather than sheer numbers.
Failure to implement a proper baseline measurement (a starting ‘zero point’) is a major reason why 72% of modernisation projects fail.
The Total Metrics-Driven Approach is essential for:
- Defining the starting point: If the starting point is undefined, you cannot accurately measure success or know if you’ve reached your goal. TMD ensures that improvements are grounded in concrete financial metrics.
- Building safeguards: You should set both positive KPIs (the desired goal) and negative KPIs (safeguards). Negative KPIs determine when the transformation is ‘on the wrong way,’ allowing the organisation to pivot if things go south.
For AI to ‘truly move the needle,’ KPIs must connect technical performance with business impact.
How should we manage the varying speed of AI adoption and risk across different departments?
You should assume that you will have varying levels of adoption speed within your teams. Different departments have different risk tolerances.
- Risk tolerance: Departments like Sales often have a margin for error and can adopt AI more swiftly. Conversely, departments like Compliance or Insurance cannot afford very risky moves, as a failure in compliance systems can be catastrophic.
- KPI alignment: It is wise to have conservative KPIs for critical, risk-averse departments. Having clear KPIs that vary for each department, based on their specific needs and risk approach, is crucial.
- Balancing risk: Successful adoption requires the organisation to have its hands in two baskets: the high-risk basket (where you cannot stay safe) and the low-risk basket (where you cannot go too risky).
By cleaning everything up and building a solid foundation first, you ensure that you can launch independent or semi-independent streams that adopt AI at their own, appropriate pace.
Take a look at these related topics as well:
What is the immediate, practical first step a company should take to begin its journey to AI readiness?
The first practical step is to conduct an AI readiness assessment or AI maturity assessment. This is particularly important for business leaders like CIOs who receive rapid directives (‘we need to be using AI next quarter’).
This initial assessment provides a diagnostic of the current situation and establishes a realistic timeline. It is comprehensive, assessing not only the technical state of IT systems but also the business state, service state, processes, and people behind the systems.
Knowing the state of your inventory and reserves allows leaders to make informed moves rather than falling victim to the AI hype wave.
The AI maturity assessment must be comprehensive, evaluating the state of the organisation across multiple critical areas, not just the technology.
What specific KPIs should we implement to manage ethical risks, bias, and compliance in our AI development?
Responsible AI requires intentional governance structures and embedding accountability at every stage.
Neglecting ethical KPIs exposes the organisation to risks such as data privacy breaches, unintended bias, and compliance gaps.
You must embed dedicated risk management KPIs aimed at prevention and early detection. Essential metrics include:
- Bias detection rates: tracking how often the AI flags and mitigates unfair patterns. Integrating bias detection helps future-proof compliance.
- Audit trail completeness: ensuring every decision and change is documented for accountability and transparency.
- Failure recovery time: measuring how fast the organisation bounces back from glitches or outages.
- Regulatory compliance adherence: showing how consistently the AI meets legal standards.
Continuous KPI monitoring is necessary to catch issues before they snowball. Transparent risk metrics build confidence, helping the company earn a reputation as a responsible, trustworthy AI partner.
Transform into an AI-boosted business.
Discover how our services will cut costs, improve productivity, test your ideas, and maximise ROI.
FAQ
Why can AI investments fail even when technically sound?
Because many organisations treat AI models as a technology experiment rather than a business initiative. Without architecture and KPIs aligned to strategy, AI can deliver pilots but not sustainable value.
What does it mean to “align IT architecture with KPIs’ in the context of AI?
It means designing your system (data pipelines, API layers, model serving, scalability) so that every technical component supports measurable business outcomes – KPIs like revenue lift, cost reduction, customer retention, or process automation.
How do you choose which KPIs to link with AI initiatives?
Start with strategic business goals (e.g. reduce churn, increase sales), then identify processes where AI might drive uplift, and define KPIs that capture that uplift.
Always connect the technical metric (e.g. model accuracy, latency) with a business impact (e.g. fewer support calls, more conversions).
How can you evolve architecture as AI matures?
Use iterative architecture – start with a clean, modular core and expand. Introduce feedback loops, refactor bottlenecks, and remove initial “hacky’ shortcuts when validated. Always keep architecture goals in view (scalability, resilience, observability).
What warning signs show architecture and KPI misalignment?
Signs include slow performance when scaling, KPI improvements in isolation (e.g. model accuracy improves but business impact unchanged), fragmented systems where AI outputs can’t feed into operations, or cost overruns.
What are the risks of overfocusing on hype rather than impact?
Risks include expensive failed projects, wasted engineering time, lost credibility among leadership, morale hit, and inability to scale AI into production.
