The AI capability lag. Turn AI into an everyday edge

• Pilot paralysis and misspent potential: While there is widespread experimentation, reports from the Australian market show that 53% of companies have slowed or stopped GenAI initiatives due to difficulties proving business value or managing data challenges². We are stuck in pilot mode because the infrastructure for continuous scaling—the people and the process—was treated as an afterthought.
• The Agentic disconnect and over-engineering: As Mantel research highlights, while the hype suggests “Agents are all you need,” the reality is that many complex, expensive Agent deployments could be solved with simpler, guided LLM workflows⁹. The drive to deploy the “biggest weapon in the arsenal for every task”—often resulting in high cost Agents for use cases like simplistic manual operations—is leading to misallocated budgets and unnecessary complexity.
• Unmanaged risk amplified: The shift to Agentic systems introduces a new layer of risk, requiring full autonomy for decision making. If governance structures are static, they fail instantly when faced with systems designed to act independently.

1. Misallocated budget risk: The project model, driven by the pressure to deliver a headline grabbing “solution,” encourages the deployment of complex, high cost technologies even when simpler deterministic rules would suffice. For instance, using a complex, expensive Agent for internal digital assistant scheduling and bookings (a common use case observed in the market) is often a gross overspend when a guided LLM workflow could achieve the same outcome with far less governance overhead⁹.
2. Unaccountable complexity: Agents are “complicated and hard to evaluate” ⁹. The hardest work—long term monitoring and evaluation of an Agent deployed for complex research reasoning—begins only after project closure. The project model guarantees this crucial sustainment phase is starved of funds and resources. The failure rate of AI initiatives is exceptionally high, with some reports suggesting 95% of GenAI pilots bring zero return ¹ and up to 90% of analytically immature organisations fail to achieve their project goals⁸.
3. Episodic investment: Funding stops when the technology is launched. This starves the necessary sustaining activities—the continuous policy updates, the dynamic reskilling, the workflow maintenance—guaranteeing that the system of capability collapses under the weight of continuous technological change.

• Perpetual funding: Capability building must be funded as a standing operational expense (OPEX). This financial commitment signals that human evolution is inseparable from technology acquisition.
• A living strategy: The capability roadmap is a living document, reviewed and pivoted quarterly, informed by technology updates, regulatory shifts, and internal data on proficiency. It exists in constant dialogue with the business and technology roadmaps.
• Measuring velocity, not milestones: Success is measured by the continuous velocity of adaptation—tracking sustained proficiency, ethical usage rates, and the verifiable business value derived from human-AI augmentation.

Failing to make this distinction is not a change management oversight; it is a strategic failure to adapt to the speed of the modern economy.