This year I joined Mantiq in Dubai, the technology team of a group in real estate and marine. Its ERP runs finance, HR, legal, procurement and the factory. My job was to bring AI into that work.
Half of the people who work in it. That is the number I watch. Volume is easy to inflate: one enthusiastic user can make any chart climb. How many people reach for the tool on an ordinary week is harder to fake. Getting there took time next to them: sitting with each team, watching how the work is really done, and putting AI only where the hours were going.
And 34 only counts the people who reached for an AI tool themselves. Much of the AI I built runs inside the pipelines: documents read and pre-filled before anyone opens the form, checks gathered before a request reaches its approver, scheduled agents preparing work in the background. The other half of the company gets that speed too, without ever opening a chat.
A manager showed me how they prepare approvals for expense notes and purchase requests: check the budget, compare outside prices, spot the recurring purchases. Two to three hours a week, spread across every day. Now an agent gathers those checks, the requests come in batches, and the manager still makes every decision.
That last part is the rule I keep everywhere. The AI can only do what the person using it is allowed to do, the risky steps stop and ask, and every call is logged. People trust a tool they can stop. Then they use it.