Dubai’s DEWA moves 76% of services to agentic AI as processing time falls 95%

What happened

Dubai Electricity and Water Authority (DEWA) has converted 16 of its 21 customer-facing services to agentic AI, equivalent to 76% of its service portfolio. The utility reports that processing times for some services have fallen by as much as 95%, with technical case assignment among the functions now handled through autonomous execution rather than manual routing.

The move signals a shift beyond conventional automation or chatbot-style assistance: agentic AI systems are being trusted to make decisions and execute multi-step workflows — such as assigning technical cases to the right teams — without step-by-step human intervention.

Why it matters

This is fundamentally a technology and operating-model story. Agentic AI differs from earlier generations of automation in that it doesn't just flag or recommend an action — it completes the task end-to-end, adapting to context within defined parameters. For a utility handling high volumes of routine and time-sensitive requests, that distinction is significant: it changes headcount allocation, staff roles, and the speed at which customers receive resolution.

For other public-sector and infrastructure operators across the region, DEWA's rollout offers an early, large-scale reference point for what agentic AI adoption looks like in practice — not as a pilot, but as the operating default for the majority of a service catalogue.

By the numbers

• 16 of 21 DEWA services have been transformed using agentic AI.
• 76% of DEWA's total service portfolio now runs on agentic AI.
• 95% reduction in processing time reported for some transformed services.

The Renascence take

The headline number here is speed, but the more interesting story is trust: DEWA has moved technical case assignment — a decision point that used to require human judgement — into autonomous hands. That's a bigger organisational leap than the percentage gains suggest.

Most coverage of agentic AI fixates on time saved, but the real design question is what happens at the 5% of cases that don't fit the pattern — the exceptions, disputes and edge cases where a customer needs a human to exercise judgement, not just execute a workflow. Utilities and government services moving at this pace should be publishing their escalation paths as visibly as they publish their efficiency gains; a 95% faster process that quietly fails the remaining 5% will erode trust faster than it builds it. The operators who get this right will treat autonomous execution and human escalation as one designed system, not a fallback bolted on afterwards.