The gap between deploying AI and redesigning for it
Artificial intelligence is making employees faster. Whether it makes organisations better depends on a decision many leaders have yet to take
26 October, 2026
TT
16
The most revealing number in artificial intelligence (AI) right now is not about adoption. In McKinsey’s latest State of AI survey, published in August, four in five respondents said AI makes them personally more productive. Only just over a third said their organisation is seeing any impact on earnings, and that share has barely moved in a year. People are working faster. Most companies are not yet performing better.
The productivity paradox
That gap matters more in the Gulf than almost anywhere else, because expectations here run higher. KPMG’s 2026 technology reports found that 76 per cent of Saudi organisations expect AI to deliver measurable returns at enterprise scale within 12 months, the highest confidence of any market surveyed. In the UAE, 96 per cent of organisations believe managing AI agents will become a critical workforce skill within five years. The region has the ambition and the investment. The open question is whether it has the operating model to match. In boardrooms across the region, I see a familiar pattern. An organisation deploys an AI assistant or agent, runs a successful pilot, and then watches the gains flatten once it meets the real business. The technology rarely fails. What stalls is everything around it: approvals that still route through four people, customer data split across systems that don’t talk to each other, and job roles designed for a world where a person handled every step. This is the gap between deploying AI and redesigning for it. Deployment adds a capable tool to an existing process. Redesign asks what that process should look like when much of the routine work no longer needs a person, and what people should do with the time and judgement that frees up. Recent studies and data point the same way: the small group of companies drawing real profit from AI are nearly three times as likely as others to have fundamentally redesigned their workflows.
Deployment is not redesign
It helps to be precise about why deployment alone falls short.
A general-purpose AI model is highly capable and knows nothing about your organisation. It does not know your customers, your pricing rules or who is allowed to approve what. Without that context it remains an articulate stranger, impressive in conversation and unreliable in operations. Closing that gap means deciding which information is trusted, who owns it, and which decisions an agent may take on its own. Those are leadership choices before they are technical ones.
Our own research reaches the same conclusion. In Salesforce’s 2026 Agentic Enterprise study, 94 per cent of business leaders said AI creates more value when it is built into core workflows than when it runs as a separate tool, and organisations that prepared their data before deploying agents saw stronger returns. The same thinking sits behind our expanded partnership with Anthropic, which featured prominently at Dreamforce this September. The premise is simple: pair a leading model’s reasoning with the business context it would otherwise lack, from customer history to the rules and approvals that turn a good answer into a correct action. Capability is no longer scarce. Knowing where to apply it is. For leaders in the region, two things follow.
The first is to treat redesign as a leadership question. Deciding which decisions stay with people, which move to agents, and how the two hand work to each other is a matter of accountability, and it belongs with the executive team rather than on the IT roadmap.
The second is to measure outcomes rather than adoption. The number of employees using an AI tool says very little. The time taken to resolve a customer issue, the cost to serve, or how long a client waits for an answer says a great deal. None of this takes people out of the picture. It changes what their work is. The Gulf’s young, digitally fluent workforce is a real advantage here, provided organisations invest in the judgement and oversight skills that redesigned roles demand, and give employees a clear stake in the change rather than leaving them to guess at it.
The region’s early commitment to AI gave it a head start. Keeping that lead will depend less on how quickly organisations adopt the next tool and more on how willing they are to rebuild the work around it.
