AI Copilots Widely Used, Financial Impact Unproven

Artificial intelligence copilots are now widely used across organisations, yet their financial impact remains largely unproven. Nearly eight in ten companies have deployed generative AI, but a similar number report no material effect on earnings. This indicates the tools were designed to address an incorrect problem, failing to move the economic needle.

Credits: UNSPLASH

Most organisational work does not require human judgement. Knowledge workers spend significant time on administrative tasks, status tracking, and data entry, activities of low value. Current AI systems assist humans in making improved decisions on tasks often requiring no human touch.

But the primary economic prize lies in automating 80 per cent of activity. This includes scheduling, routing, subsequent outreach, and data reconciliation. Copilots have mainly optimised the remaining 20 per cent, assisting individuals with thinking, writing, and information retrieval. These broader tasks demand coordination and reliable execution, capabilities now within AI's scope, according to Forbes.

The bottleneck has shifted from information scarcity to execution inability. Organisations are now overwhelmed, not by lack of insights, but by failure to implement recommendations at scale. New data often stalls due to overworked staff, disparate systems, and workflows not designed for autonomous action. Providing further insight only widens this gap.

And McKinsey refers to this as the "Gen AI paradox." Organisations deploy tools but do not capture value. Their research estimates agentic AI could automate USD 2.9 trillion in US economic value by 2030. This requires addressing the knowing doing gap as a design challenge, not an AI capability issue.

The necessary design shift is vertical AI execution. This AI is built for specific purposes. It coordinates workflows with multiple steps, completes tasks across systems, and adapts to unusual circumstances. Such systems maintain human involved guardrails and ensure auditability, vital for regulated industries.

So, copilots became the default architecture due to caution when AI systems were newer and less reliable. This caution remains valid. It no longer argues against AI execution. It instead advocates for building execution systems embedding accountability from inception. Every action must be traceable, every decision recorded, and every outcome reviewable.

Early automation often operated as a black box. Vertical AI execution adopts a distinct method: auditability is a core design principle from the beginning. In any regulated industry, this means confidently explaining AI actions, reasoning, and consequences. Organisations overlooking this risk restricting autonomous AI's potential in areas of greatest value.

And pressure mounts for AI investments to demonstrate their worth. Workforces are leaner, budgets tighter. Patience for AI initiatives producing dashboards without outcomes is expiring. Metrics must change. Prediction accuracy or active copilot users do not represent return on investment.

True return on investment is evident when workflows complete faster, exceptions resolve without human escalation, and outcomes improve. It requires demonstrating which bottlenecks AI eliminated and what new capacity resulted. McKinsey estimates the overall productivity opportunity from AI at USD 4.4 trillion annually.

So, bridging the gap between current applications and this potential requires a willingness to redesign for direct execution. Copilots were a valuable initial chapter, accustoming organisations to working with AI and delivering genuine productivity gains. This groundwork holds importance.

Organisations defining the next phase will be those systematically identifying where human judgement is genuinely required and automating all other processes. Leaders must assess what percentage of work truly needs human involvement. For most organisations, the answer will be surprising, potentially uncomfortable. Closing this gap marks the start of real AI value creation.

• Generative AI deployment is widespread but shows limited economic impact.

• Copilots primarily optimise a small portion of organisational work.

• The real economic value lies in automating tasks not requiring human judgement.

• Vertical AI execution, which directly completes tasks with built in auditability, is proposed as the next step.

• Measuring return on investment requires new metrics focused on workflow speed and outcome improvement.

Source: Forbes