“Forget artificial intelligence. In the brave new world of big data, it’s artificial idiocy we should be looking out for.” 

– Tom Chatfield

Blindly delegating critical human judgment to systems that excel at computation but fail at true understanding and flexibility will usually lead to systemic, automated mistakes (artificial idiocy) rather than intelligent outcomes. Tom Chatfield, a British author, journalist, and technology philosopher, warned in 2014. Despite this, most organisations today are still operating at the “crawl” stage of generative AI, using data to generate business insights through an arbitrary chat (prompt). There are, however, higher stages to aim for. 

If you are still at the prompts level, the next stage to aim for is “walk”. This is where AI is designed to take specific actions on behalf of the user, such as starting a workflow or obtaining approval within systems you already have access to. Beyond this is the “run” phase. This is where AI is built with a full-fledged agentic architecture using several task-specific agents equipped with web-enabled Application Programming Interface (API) configurations. These agents communicate with each other about your data and can make highly complex decisions at scale and speed, giving your organisation a disruptive, competitive advantage.

Reaching this stage, however, is a journey and a process, guided by global best practices & standards, some of which we shall discuss. As Andrew Ng of Google Brain advised, “Humans are not perfect, and neither is AI, but together, we can create something extraordinary.”

If you have been following the steps outlined in the previous two articles, your AI initiative should no longer be just a technological experiment. It should have grown into a potentially strategic financial asset. This is to say, as your organisation matures its AI project, you might qualify to recognise your in-house AI systems not merely as operational expenses, but as intangible assets on the balance sheet. This shift, governed by developments in IAS 38, marks a turning point in how businesses value innovation. Reclassifying AI initiatives from short-term “project expenses” to long-term balance sheet “assets” can offer several significant strategic and financial advantages for corporates, which include: 

Improved financial ratios: By capitalising AI development costs, companies can defer expenses and improve ratios. This will enhance key financial metrics such as Earnings Before Interest, Tax, Depreciation and Amortisation (EBITDA) and Net Income. This accounting approach can boost investor confidence and positively influence company valuation.

Strategic signalling: A firm treating AI as an asset communicates a long-term commitment to innovation and technological maturity. This positions the organisation as being forward-thinking and tech-driven, which can boost stakeholder trust and its market and brand perception.

Better budgeting and Return on Investment (ROI) tracking: AI assets are managed through depreciation schedules, enabling clearer, longer-term ROI analysis. This structured approach supports stronger governance and more effective asset performance reviews.

In contrast, treating such an AI initiative purely as a project expense may obscure its long-term value. Further, it can reduce visibility into the strategic impact of such AI initiatives and limit the firm’s ability to measure and communicate its full potential.

If you wish to determine whether your AI project is viable and can be legitimately recognised as an asset, you must:

• First, establish its technical feasibility. This means that you need to show that the AI system can be successfully developed and will operate as intended. 
• Next, you must show a genuine intention to complete the AI asset. This involves having a formally documented plan in place, with allocated resources such as budget, personnel, and infrastructure. It is important to note that global regulations, such as the EU AI Act, require inadequately documented corporate AI systems to be identified as Liabilities. 
• Finally, your AI project must be expected to deliver probable future economic benefits. This means that your AI system must demonstrate a contribution to the organisation’s financial performance by reducing costs, increasing revenue, providing a strategic edge, or some combination of these.

The CMM Model as a Roadmap to AI Excellence

As Jaron Lanier, the American computer scientist, visual artist, computer philosopher, and pioneer of virtual reality (VR), stated in the mid-2010s, “The best art of the future will be a fusion of human imagination and AI precision.” Borrowing a leaf from this thought, as humans, we must guide our AI initiatives with well-thought-out, systematic steps. To achieve this for your AI project, you can adopt various standards and frameworks. One global best practice is the Capability Maturity Model (CMM), which offers a structured framework for project transformation. This model outlines various levels, from chaotic experimentation to continuous optimisation, and applies to most project implementations. This model guides the assessment and improvement of your AI project maturity across five distinct levels:

Level 1

Initial (Chaotic): At this foundational stage, processes are ad hoc and reactive. Project success hinges on individual brilliance rather than repeatable systems. Organisation projects at this level often lack key documentation, version controls, and sufficiently skilled professionals. As an analyst from Gartner put it, “AI maturity at this level is like giving a toddler a chainsaw; it’s technically possible, but not advisable.”

Level 2

Repeatable: At this level, basic project management processes begin to emerge. Organisations can replicate past project successes fairly well, though innovation remains limited and scalability is constrained. For instance, a mid-sized insurer might use its in-house AI to process claims using consistent workflows, but that system may lack the flexibility to scale or adapt to new challenges as they arise.

Level 3

Defined: At this level, your AI development processes become standardised and documented across the entire organisation. AI initiatives deliberately align with broader business goals, enabling more strategic deployment. An example is a multinational bank establishing a centralised AI governance board that oversees standardised agent-model development policies, ensuring consistency and compliance across all its departments globally.

Level 4

Managed: At this stage, your organisation will collect detailed metrics to understand your AI’s performance to improve it quantitatively. Two critical metrics you need to track include: 

• Learning Velocity (LV): Here, you will measure how quickly your AI system improves in quality and accuracy of its outputs relative to time and budget. 
• Model Degradation Velocity (MDV): Here, you will track how fast your AI system loses accuracy in its outputs over time without retraining.

These metrics are essential for maintaining the performance of your AI project and ensuring a reasonable return on investment. This level also highlights the criticality of an ongoing maintenance process for your AI initiative, covering its entire life cycle. 

Level 5

Optimising: This is the highest project maturity level, where continuous improvement becomes the norm across your entire organisation. Here, all your AI systems are continually refined using carefully designed feedback loops and innovative techniques. Your internal corporate data is also iteratively tagged and catalogued according to ever-changing industry-specific standards and terms to sharpen the AI’s insights, even though this process regularly increases operational costs. You do this while appreciating that sharpening AI with insufficient data is like tuning a piano with a sledgehammer. The payoff of a well-optimised AI initiative is a self-sustaining system that delivers high-quality, actionable strategic business intelligence. At this point, you will have moved your project from the “crawl” stage to the “run” stage.

It is also paramount to remember that “optimisation” in an AI initiative isn’t just about refining algorithms; it requires aligning your business data, governance, and long-term strategies. High-maturity organisations also understand that the real power of such an AI project lies in its sustained impact over time. Such firms keep AI projects operational for at least three years to ensure that they deliver lasting business value as they evolve into strategic assets. Therefore, in addition to the direct monetary value potential, consider other gains from this initiative, such as improved data quality and increased employee productivity.