Next wave of AI will be physical: What Gulf business leaders must learn now

We are entering the era of embedded AI: artificial intelligence that can perceive, reason, plan, and act in the real world. AI began with perception — recognising images, words, and sounds, and moved to generation — creating text, code, and media. Now it is becoming embodied: machines that navigate warehouses, operate on patients, inspect infrastructure, and manufacture goods alongside humans.

The embodied AI market is projected to grow from $4.4 billion to $23 billion by 2030, at 39 per cent annually. Morgan Stanley forecasts the humanoid robot market alone reaching $5 trillion by 2050. SAP’s Embodied AI initiative reports up to 50 per cent reduction in unplanned downtime in early enterprise deployments. Goldman Sachs notes that humanoid manufacturing costs dropped 40 per cent between 2023 and 2024 alone.

For the Gulf — which is investing billions in logistics hubs, smart city infrastructure, autonomous systems, and advanced manufacturing — this represents the defining economic opportunity of the next decade. The UAE AI Strategy 2031 projects that AI will add over $96 billion to the nation’s GDP. But that figure assumes effective deployment at scale, including in the physical and industrial domains where the largest value creation will occur. The question is whether it has the leaders who can convert that access into operational advantage before the window narrows.

The market for individual AI literacy has been comprehensively served. Courses on what AI is, how large language models work, and how to use ChatGPT for personal productivity are abundant, affordable, and increasingly free. The EY 2025 Work Reimagined Survey, covering 15,000 employees across 29 countries, confirms this: 88 per cent of employees now use AI in their daily work. Individual adoption is no longer the bottleneck.

The bottleneck is the distance between what AI does for one person and what it does for an entire organisation. Only 5 per cent of employees use AI in ways that fundamentally transform how they work. Companies are leaving up to 40 per cent of potential productivity gains unrealised. MIT’s 2025 Project NANDA study of 300 enterprise implementations found that 95 per cent of AI pilots produce zero measurable impact on profit and loss. The constraint, consistently, is not model quality, budget, or regulation. It is organisational design: workflows that remain unchanged, feedback loops that do not exist, and AI tools that are purchased but never woven into the fabric of how work actually gets done.

This matters more as AI moves into the physical world. Screen-based AI is forgiving — an employee can experiment with a chatbot, get useful results, and share a tip with colleagues. The value accrues one user at a time. Physical AI operates differently: it requires models running with extreme efficiency at the edge, making millisecond decisions on a factory floor rather than round-tripping to the cloud. It demands integration across production systems, quality controls, safety protocols, and sensor networks. The organisations that spent years running chatbot experiments without restructuring their operational DNA will not be ready when physical AI arrives in their sectors. The ones that built institutional AI capability will move first and compound their advantage.

The World Economic Forum reports that 94 per cent of leaders face AI-critical skill shortages, with one in three reporting gaps of 40 per cent or more. The skills in shortest supply are not traditional engineering competencies. They are AI governance, agentic workflow design, and human-AI collaboration — capabilities at the intersection of technology, operations, and leadership. IDC estimates the global cost of this gap at $5.5 trillion by 2026.

The successful 5 per cent of organisations partner with specialised vendors — external partnerships reach production 67 per cent of the time, versus 33 per cent for internal builds. They decentralise implementation: domain experts and line managers surface problems, select tools, and lead rollouts. And the strongest deployments began with power users — employees who had already experimented with AI for personal productivity and became champions of formal adoption. When this bottom-up capability met top-down strategic accountability, adoption accelerated while preserving operational fit.

Neither direction works in isolation. Top-down education alone produces strategic frameworks that never reach the operational floor. Bottom-up adoption alone produces scattered tool usage that never compounds into enterprise capability. The organisations that cross the “GenAI Divide” are the ones where both movements converge: leadership that creates the right conditions, and operational teams that surface the right use cases and feed results back into strategy.

The UAE has made genuinely forward-thinking investments in AI leadership infrastructure. An IBM–Dubai Future Foundation study found that 33 per cent of UAE organisations have appointed a Chief AI Officer, above the 26 per cent global average. Ninety per cent of UAE CAIOs report strong CEO support. As Minister of State for AI Omar Sultan Al Olama wrote, “AI is not a singular breakthrough — it’s 10,000 small shifts. It’s cultural. It’s institutional. It’s a habit.” But 76 per cent of UAE organisations remain at pilot stage, compared to 60 per cent globally. What is missing is a specific set of leadership capabilities that translate architecture into outcomes – and it is precisely this gap that dedicated executive AI programmes, such as the Executive Program for Chief AI Officers developed by Polynome AI Academy and Abu Dhabi School of Management, are designed to close.

First: distinguish a pilot from a product. Most AI pilots are demonstrations of technical possibility, not evidence of business value. Leaders need the fluency to ask whether an AI initiative reduces a measurable cost, accelerates a revenue-generating process, or merely produces an impressive demo. MIT found that the biggest predictor of failure is not technology choice but the absence of business-outcome metrics from the outset. If a leader cannot articulate — before deployment — which line on the P&L an AI system is meant to move, the pilot is already in the 95 per cent of failure.

Second: design workflows around AI, not the reverse. The default approach in most organisations is to take an existing process and add an AI tool to one step. This is why most AI usage remains basic. Transformative deployment requires redesigning the workflow itself — rethinking what is automated, what requires human judgement, how decisions are routed, and where feedback is captured. This is an operational design skill, not a technology procurement skill.

Third: evaluate AI systems by operational efficiency, not just accuracy. As AI moves into physical environments, the ability to assess whether a model can run at the edge with the required latency, power, and reliability becomes a leadership-level decision. A model that is 99 per cent accurate but too computationally expensive to run in real time on a factory floor is not a viable product. Leaders making investment decisions in logistics, manufacturing, and infrastructure need to understand this trade-off — and it is the defining technical challenge of physical AI.

Fourth: build feedback loops between strategy and operations. The organisations that scale AI successfully are the ones where insights from the operational floor reach the C-suite quickly, and strategic priorities are translated into actionable deployment plans just as fast. This requires deliberate organisational design: governance structures, reporting cadences, and incentive systems that reward learning speed, not pilot volume. The IBM–DFF study found that where CAIOs drive a hub-and-spoke model connecting executive strategy to operational teams, AI returns increase by up to 36 per cent.

Fifth: manage AI as a portfolio, not a project. The most effective organisations treat AI like a capital allocation problem. They decide consciously where to build internally and where to partner — MIT shows partnerships outperform internal builds two to one. They balance investment between digital AI that improves productivity today and physical AI that will define competitive advantage tomorrow. And they sequence deployments based on where workflow integration is easiest and business impact is most measurable, typically starting with back-office operations before moving to customer-facing systems.

The next wave of AI will not live on screens – it will move through warehouses, across factory floors, and into the physical systems that power economies. The Gulf has the infrastructure, the capital, and the political will to lead in this era. What it needs now is not more technology adoption but a generation of leaders who think in systems rather than tools, who can operate at the intersection of strategy and engineering, and who understand that the distance between AI ambition and AI execution is closed not by procurement but by organisational design. The region's early investments in CAIO appointments, sovereign AI infrastructure, and executive AI education are the right foundation. That is the real race — not for the best model, but for the fastest institutional learning loop. The organisations that build it will capture the value of physical AI. The rest will watch from the pilot stage.

- The writer is CEO, Polynome and Polynome AI Academy