BY SRIJITH KN & SUBRATO BASU

As AI spending rises and experimentation gives way to accountability, the real challenge for business leaders is no longer access to technology, but proving value, controlling cost and redesigning how organisations work.

Every year, the technology industry promises a breakthrough that will change everything. New models arrive, computing capacity expands and investment reaches new highs. Then, somewhere between the product demonstration and the quarterly earnings call, someone asks the question that matters most:

What did all of this actually improve?

That question came into sharper focus in 2026.

Artificial intelligence moved deeper into everyday business. Companies deployed assistants, experimented with agents and embedded AI across customer service, software development, compliance, finance and operations.

Yet adoption turned out to be the easy part.

Installing an AI assistant is relatively straightforward. Redesigning an organisation around it is not.

Fragmented systems, poor-quality data, unclear ownership, weak governance and sceptical employees do not disappear because a company has purchased another software licence.

Businesses do not need another year of impressive demonstrations. They need to know whether AI is helping them make better decisions, serve customers more effectively and run their operations more reliably.

That shift is turning AI from primarily a technology question into a management one.

From experimentation to accountability

The adoption numbers appear impressive.

McKinsey found that nearly nine in ten survey respondents reported regular organisational use of AI in at least one business function. Gartner, using a different measure, found that only 22% of organisations had successfully scaled AI across multiple business units or adopted an AI-first approach.

The methodologies are different, but together they point to an important distinction: using AI is not the same as changing how a business operates.

The economics reinforce that concern.

In McKinsey’s Technology Trends Outlook 2026, one survey found that 93% of respondents had exceeded their AI budgets. Another found that only 37% of companies could attribute a positive EBIT impact to their AI programmes.

Those numbers create an uncomfortable boardroom question.

If spending continues to exceed expectations while measurable returns remain limited, what exactly are organisations paying for?

The clearest returns are likely to emerge where companies begin with a defined business problem: reducing customer-service resolution time, accelerating code review, simplifying invoice handling or strengthening compliance.

What becomes harder to justify is the proliferation of pilots with no accountable owner, no performance baseline and no clear route to economic value.

Automating a broken process can simply make the breakage happen faster.

Agentic AI raises the stakes

Agentic AI made this management challenge more significant in 2026.

Traditional generative AI largely responded to requests. Agents can increasingly perform multistep activities, interact with systems and execute actions.

That creates more opportunity, but it also introduces more cost, operational risk and responsibility.

Gartner has forecast that more than 40% of agentic AI projects will be cancelled by the end of 2027 because of escalating costs, unclear business value and inadequate risk controls.

It is a forecast rather than an observed failure rate, but the warning matters.

Giving software more autonomy only increases the need for organisations to define who owns the outcome, who monitors the process and who can intervene when something goes wrong.

The question for leaders is therefore becoming less about what an AI agent can do and more about what the organisation should allow it to do.

The economics nobody puts on the slide

AI is also forcing companies to think differently about technology costs.

The International Energy Agency reported that capital expenditure by five major technology companies exceeded $400 billion in 2025 and was expected to rise by another 75% in 2026.

The physical foundations of AI require enormous amounts of computing infrastructure, electricity, cooling, networking, land, semiconductors and highly specialised talent.

An individual AI task may become cheaper while the total technology bill continues to grow simply because usage expands.

That is not necessarily a problem if the additional activity creates measurable value.

But somebody needs to be able to explain what the extra expenditure produced.

Did it generate more revenue?

Did customers receive better service?

Did employees become more productive?

Did the organisation reduce risk?

Technology and finance leaders increasingly need to measure the cost per useful outcome, rather than concentrating only on the price of individual models, tokens or software licences.

Energy, computing capacity and infrastructure resilience now belong in the same boardroom conversation as digital transformation.

AI also creates a wider economic question

There is another issue that sits beyond corporate productivity.

A business may be able to reduce certain labour requirements through automation, but an economy still depends on people earning, consuming, creating businesses and paying taxes.

If productivity gains do not translate into new industries, new forms of employment or greater economic opportunity, the benefits become harder to sustain.

Reskilling cannot therefore be treated simply as a corporate social-responsibility programme.

It increasingly belongs at the centre of technology strategy.

The real measure of intelligent technology should not be how many jobs it eliminates or how many tasks it completes. It should be whether it creates durable value for enterprises while expanding economic opportunity around them.

The Gulf moves from technology buyer to builder

This is particularly relevant in the Gulf.

For years, the region was often described primarily as a major buyer of global technology. That description increasingly feels incomplete.

Across the UAE, Saudi Arabia and the wider Gulf, governments and businesses are investing in AI infrastructure, cloud capacity, digital government, data centres and technology ecosystems.

Stargate UAE, the planned one-gigawatt AI infrastructure cluster in Abu Dhabi, illustrates the scale of that ambition. Its first 200 megawatts were scheduled to come online during 2026.

Abu Dhabi’s ambition to build an AI-native government by 2027 and Dubai’s support for digital entrepreneurship point in the same direction.

But infrastructure alone does not create prosperity.

The more important question is whether that infrastructure lowers the cost of innovation and makes advanced technology accessible to the companies that drive the region’s everyday economy.

Can a UAE SME use AI to compete more effectively with a multinational?

Can a local healthcare provider deploy specialised intelligence without building an enormous internal technology operation?

Can a logistics company use automation to improve reliability rather than simply add another layer of software?

The Gulf has invested heavily in technological ambition. The next chapter will be about making that ambition useful.

That means moving from importing technology to developing capabilities, intellectual property, businesses and talent around it.

Innovation moves into the physical world

The technology conversation is also becoming broader than generative AI.

Physical AI is bringing intelligence into robotics, logistics, manufacturing, healthcare and mobility.

Digital twins, edge computing and intelligent energy systems are connecting digital decisions to physical operations. Smaller, specialised AI models may also offer better economics and governance than enormous general-purpose systems for narrowly defined industry tasks.

Quantum computing continues to advance, although business leaders should remain cautious about claims of immediate universal commercial impact.

The more urgent issue for many organisations is post-quantum cybersecurity.

Confidential information protected today could potentially become vulnerable as computing capabilities evolve, meaning organisations handling long-lived sensitive data need to begin thinking about cryptographic resilience well before large-scale quantum computing arrives.

Across these technologies, the competitive advantage is unlikely to come from collecting the greatest number of tools.

It will come from integrating the right technologies into an operation that works.

Trust becomes part of the business model

As AI systems gain access to more sensitive information and perform more consequential activities, cybersecurity becomes inseparable from business strategy.

The World Economic Forum’s Global Cybersecurity Outlook 2026 described an environment shaped by AI-enabled threats, geopolitical fragmentation and uneven cyber resilience.

AI agents interacting with financial systems, customer information or operational infrastructure raise the importance of identity controls, authorisation, monitoring and human intervention.

Cybersecurity can therefore no longer be treated simply as the responsibility of the security department.

It is increasingly about business continuity, reputation and customer trust.

The companies that can demonstrate that intelligent systems are secure, explainable and accountable may find that trust becomes a genuine competitive advantage.

2027: the numbers will have to add up

If 2026 was the year organisations experimented with how far AI could go, 2027 is likely to bring much tougher questions about whether those investments make economic sense.

A good demonstration will matter less than what a company can responsibly allow a system to do.

Agents will begin taking responsibility for defined activities, but each will require boundaries, monitoring and a human being who ultimately owns the consequences.

Boardroom conversations should shift from:

• How many AI tools have we deployed?
• to:
• What does this process cost now?
• How much faster is it?
• Has the error rate fallen?
• Are customers noticing the difference?

That scrutiny will extend beneath the software layer.

Energy availability, sovereign computing capacity, regulatory requirements and infrastructure resilience will increasingly influence where companies choose to invest.

The same scrutiny will reach the workforce.

Employees will need technical fluency, but they will also need business judgement: the ability to understand exceptions, question automated recommendations and take responsibility when systems get something wrong.

Access to another AI model will not be the defining advantage.

An organisation that operates better — at a cost that survives examination — will be.

The question worth taking into 2027

As 2026 closes, business leaders should temporarily set aside the question of how much technology their organisations have adopted.

Instead, ask what the investment actually changed.

Did serving a customer become simpler?

Did an expensive process become more efficient without becoming less reliable?

Did employees gain the ability to spend more time on valuable work?

Can the improvement be seen in the numbers?

A technology budget is ultimately a statement of belief about the future.

In 2027, belief will not be enough.

AI adoption alone will no longer be impressive. The organisations that stand out will be those that can explain, in plain business terms, what they spent, what improved and who benefited.

And that may be the clearest indication yet that AI is becoming less of a technology problem — and much more of a management one.

Subrato Basu is an author, adviser and speaker on AI strategy and technology cost management. With nearly three decades of experience advising C-suite leaders, he helps organisations across North America, Asia Pacific and the Middle East turn technology investment into measurable business value.

Srijith KN is Senior Editor at Integrator Media, who is hands-on in covering enterprise technology, artificial intelligence, cybersecurity and the digital economy across the Middle East.

Abu Dhabi’s data center industry is entering a critical phase in which hyperscalers and organisations building and running its capacity must work as one with government to create the foundations needed to build AI infrastructure at scale. How the industry moves from individual projects to a unified approach will be high on the agenda at Data Center World Middle East, the premier global event for data center, IT infrastructure and innovation for more than four decades, which makes its Middle East debut in Abu Dhabi next week.

Abu Dhabi is already the Middle East’s largest data center market, with 374MW of IT capacity, according to the EMEA Mid-Year Data Center Report 2026 from JLL, the global commercial real estate and investment management firm. The Department of Energy expects AI infrastructure to need around 2GW of power in a first phase, rising to 5GW, and is thus investing AED 250 billion in energy and water infrastructure between 2025 and 2035. Mehdi Paryavi, Chairman and CEO, International Data Center Authority (IDCA), who speaks on the opening panel, “AI, Power & Sovereign Demand,” said: “The biggest gap in making Gulf data centers AI-ready is the lack of a coordinated approach to scale. The region needs adaptable standards, a skilled workforce, long-term regional planning, reliable energy at the scale AI requires, and greater stability. Developers must look beyond individual facilities and work with governments, utilities, and industry partners to build an ecosystem that can support sustainable, large-scale AI infrastructure.”

For the companies running Gulf capacity, that work starts long before a facility opens. Khazna Data Centers, one of the world’s fastest-growing hyperscale data center platforms, with 30 data centers in operation and 673MW of capacity, is central to the UAE’s ambitions.

Morten Tidemand, Senior Director of Operations, Khazna Data Centers, who speaks in the fireside chat “Ambition vs Availability: Supply Chain Realities,” said: “The biggest adjustment for operations teams as AI workloads become more demanding is recognising that operational readiness cannot begin at handover. Higher densities require us to rethink cooling strategies, maintenance regimes and emergency procedures well before new capacity goes live. Equally important is ensuring our teams understand how these systems work together, not in isolation.”

Data Center World Middle East organised by inD – an Informa/DWTC company – runs on 17–18 November at Dusit Thani Abu Dhabi and is built around those delivery questions. Technical sessions will cover liquid cooling at the limit and mixed-density halls running racks from 10kW to 100kW+ and delegates will hear from the UAE Ministry of Energy and Infrastructure and the UAE Ministry of Investment, alongside hyperscalers AWS, Google, Microsoft and Oracle.

Tara Gibb, Senior Event Director, Data Center World said: “Data Center World Middle East is where the people who build, power, finance and regulate the region’s AI infrastructure meet to make decisions. Abu Dhabi has made the commitments. Turning them into capacity that runs reliably at 50°C depends on everyone working from the same plan, and that is the conversation this event is built for.”

Tabreed, the National Central Cooling Company, whose district cooling networks have 1.58 million refrigeration tonnes of connected capacity, will be represented by its CEO, Dr Yousif Al Hammadi, and Chief Operating Officer, Antonio Di Cecca. Other confirmed speakers include Sameer Cassim, Senior Vice President, Business Development, BDx Data Centers; Khaled El Bedri, Director of Cloud Solutions Architecture, Microsoft; Magide Sebtioui, Director, Global Strategic Accounts & Growth, Honeywell; Abdulrahman Khalid Almazroa, Director of Cybersecurity, Johns Hopkins Aramco Healthcare; Aida Elbouanani, Business Development Director MEA, Liquid Cooling, Castrol; and Dr Umaima Haider, Principal Analyst, Omdia.

Delinea, the identity security platform that continuously controls what AI agents, humans and machines can do and for how long, today announced its participation at GITEX GLOBAL 2026. Under the theme “From Access to Action: Governing AI at Runtime”, the company will demonstrate how organisations can put enforceable controls behind their AI ambitions, reducing the risk of rogue actions, while giving businesses the confidence to deploy autonomous agents across critical systems.

In recent research, 98% of UAE respondents admitted that their organisations had experienced confirmed or suspected instances in which an AI tool accessed sensitive data beyond what its task required. Concerningly, only 16% said they could immediately revoke an AI tool’s credentials and terminate its session, highlighting the gap between granting access and retaining control. These findings come as regional frameworks, including Saudi Arabia’s National Cybersecurity Authority Essential Cybersecurity Controls (NCA ECC) and UAE government cybersecurity standards, reinforce the need for effective, auditable access controls. Through its participation, Delinea will show how runtime authorisation can help organisations apply these principles to autonomous agents by enforcing policy at the moment of action.

“This year has challenged regional organisations to build resilience against unexpected disruption while responding to AI-enabled threats and the shift from conversational AI to agents that act across business systems,” said Mortada Ayad, VP – META, Delinea. “These monumental shifts create opportunities to automate more complex work, but also make control over autonomous actions essential. GITEX’s December timing gives leaders the opportunity to reflect on these pressures and decide what they need to move forward. Identity security and runtime authorisation give them a foundation for that transition by empowering them to know who or what is acting, limit actions to the task, and retain the ability to intervene.”

Runtime authorisation will be at the centre of Delinea’s product demonstrations, showing how the Delinea Platform brings together enterprise privileged access management and just-in-time controls for cloud-native and DevOps environments. At dedicated demo stations, visitors will explore Secret Server, privileged access management and identity threat detection, to understand how to discover identities, assess their access and reduce reliance on standing credentials. The platform’s Cedar-based policy engine enforces least privilege throughout the session, translating access policies into controls that limit what an identity can do and for how long.

Building on these capabilities, the demonstrations will show how runtime authorisation extends control to AI agents working across databases, SSH hosts, Kubernetes clusters and cloud consoles. Visitors will see how task-scoped access and real-time policy enforcement can allow, block or escalate an action before it executes, helping address the excessive access identified in Delinea’s research. The focus of these product showcases will be on helping teams govern agent activity, as it happens, alongside the credential and session controls needed to intervene when access must end.

“As CISOs and IT decision makers shape their 2027 budgets, GITEX is an opportunity to translate the year’s lessons into practical conversations with security partners such as Delinea. It’s not just an exercise in defence, but an opportunity to explore how security can enable the next growth stage of the business. If leaders are being asked to deliver agentic AI outcomes, they need to understand how to give agents enough freedom to do useful work while retaining the ability to govern their actions. We look forward to welcoming visitors to our stand and helping them turn those ambitions into a practical plan for secure AI adoption,” said Ayad.