One principal, many agents: what consolidation says about the next developer operating model

On 5 November, Emaar asked its minority shareholders to sell their stake in its listed development business so the group would bring it fully back in house. Read alongside what Aldar and Dubai Holding have done over the past two years, the move says as much about the future developer operating model as it does about capital.
Consolidation is a signal
Emaar's offer values Emaar Development's minority stake at about AED 12 billion, at AED 15 a share. According to the company's filing, as reported by The National, the transaction would let Emaar Properties simplify and consolidate the group's structure and run the UAE development business on a fully integrated basis, under a single capital allocation framework. The offer closes on 10 December.
Since early 2024, Dubai Holding has used Palantir's AI and data integration platforms across its businesses, including Nakheel, Meraas and Jumeirah. Aldar reported AED 3.3 billion of strategic acquisitions in 2025, and the same results release described Digital Spine, its core transformation programme, as having managed more than 50 active projects digitally in 2025, with lower costs and faster approvals.
Each move was explained in terms of capital, efficiency or growth, and none of the companies has linked them to AI. The connection is our reading. Taken together, they point towards fewer principals, one decision structure and one data estate across businesses that used to run separately. Consolidation answers the capital question first. It also reopens the operating model question: who decides what, at which threshold, on which data, and with what visibility from project team to board. That question is about to acquire a new kind of participant.
The agents are already at work
In September, Aldar and Globant introduced two artificial intelligence (AI) agents inside the Hello Aldar platform. The Construction Progress Agent translates technical construction data into plain-language updates for buyers. The Post-Sale Service Dispatcher categorises issues, routes them to the right technician, sets priority and books the appointment. According to the announcement, both agents operate autonomously, and both query a unified customer layer that Globant created from customer, service and property records that had previously sat in separate systems.
In May, Dubai Holding announced a partnership with Microsoft to bring AI into operational processes and decision-making frameworks across its businesses, including real estate and community management, giving employees AI tools through a unified interface and deploying AI agents to automate routine tasks. The cross-brand data platform came first, and group-wide agents are following two years later.
Regulators are moving in step. The Dubai Land Department's Initial Registration platform, launched in September, uses AI to read submitted documents, and registration transactions that comply with its business rules are eligible for approval on submission. Its Project 360 view brings unit status, escrow accounts and project records together, and a single developer account can manage several companies, with defined permissions and separate roles for submitting and reviewing transactions. At federal level, the UAE Cabinet announced in April a framework to move 50% of government sectors and services onto agentic AI within two years, under the supervision of Sheikh Mansour bin Zayed and built on integrated federal data. Dubai Municipality is developing an AI agent to issue permits for private and investment villas automatically, with manual review limited to exceptional cases. When the regulator's side of an approval runs as an agent, the developer and its consultants have to answer for the quality of what they submit.
Read as an operating model, the federal programme has three parts: one accountable principal, an integrated data foundation and delegated execution. Developers face the same design question.
Many agents, one accountable principal
McKinsey, writing in September 2025 from its work with early adopters, reports that a human team of two to five people can already supervise 50 to 100 specialised agents, with humans holding final accountability. The ratio changes the shape of an organisation. The accountability does not move.
Clyde & Co's September 2026 analysis of AI liability in the Middle East concludes that courts in the region are unlikely to treat AI as a legal actor capable of assuming responsibility in its own right, and that a deploying organisation may face liability if it implements AI irresponsibly. In the Dubai International Financial Centre, Regulation 10 of the data protection regulations deems the deployer of an autonomous system processing personal data to be the controller, and lists the appointment of an Autonomous Systems Officer among the conditions for high-risk processing.
Government announcements speak of agents acting without human intervention. In practice, oversight becomes a design decision: which actions an agent can take alone, and which still need a person who can answer for them.
An organisation that deploys many agents therefore needs a principal that can answer for all of them. A group of semi-independent entities, each with its own approvals, data and capital rules, struggles to play that role. Consolidation can settle the identity of the principal. It leaves open how that principal delegates. In operating model terms, that is a question of delegation of authority, role clarity and decision rights, the same instruments a developer already uses for its people.
The gap sits in the organisation
In a McKinsey survey of 139 GCC executives and board directors, published in November 2025, 60% said their organisations use AI agents to some extent, yet only 11% qualified as value realisers, and most of those had a well-established technology foundation and strong data fundamentals. The Royal Institution of Chartered Surveyors (RICS) reported in August 2026 that construction data resets at each project boundary, and that fewer than 1% of respondents to its global construction survey reported full integration of AI across their organisation.
The governance picture is similar. BCG found in September 2026 that 42% of companies expect to give agents genuine autonomy by 2030, while only 5% have the full set of controls in place today. Gartner expects 40% of enterprises to demote or decommission autonomous agents by 2027 because of governance gaps found only after production incidents. It also warns against applying one uniform governance regime to every agent, and sets out four levels of autonomy: observe, advise, act with approval and act autonomously. That cuts against an easy reading of consolidation. One accountable principal does not mean one rulebook: a consolidated group needs a single line of accountability and differentiated authority beneath it.
Read together, these findings place the constraint in data that persists across projects, decision rights explicit enough to delegate, and controls that remain meaningful under pressure. Each of those is an operating model question.
The operating model becomes the agents' brief
The structures that govern a developer organisation now carry a second job. They become the specification that agents work within.
A delegation of authority already defines who approves what, at which threshold. In an agentic organisation it also defines which level of autonomy each agent holds, and where a human approval remains a real control.
Stage gates do the same. A decision-led gate sets out what must be true before a project moves forward. An agent can assemble the gate evidence and flag exceptions, while the decision stays with an accountable lead. The Dubai Land Department's platform already builds a separation between submitting and reviewing roles into the developer's own account.
A single source of truth stops being a reporting convenience. Agents act on the data they can reach, and data that resets at each project boundary will produce agents that do the same. Role clarity extends to agents: each one needs a named human owner, and each owner needs to know where their authority ends. Portfolio visibility changes too. A board that sees project status from its teams also needs to see what agents have decided, at which level of autonomy and on whose authority.
One major Abu Dhabi master developer we worked with had been running 12 projects without a single version of the truth. It moved to one governance framework, one source of truth and decision-led stage gates. Foundations of that kind are where an agentic operating model starts. Agents deployed without them risk automating fragmentation at the speed of software.
For developer leadership, the governing question now has a second layer: is the organisation built to perform at every level when part of its workforce is software acting on its authority?
- The National: Emaar plans to delist construction unit after board approves $3.3 billion buyout, November 2026
- Emaar Properties: Emaar Development acquisition offer, November 2026
- Government of Dubai Media Office: Dubai Holding and Palantir launch Aither, November 2025
- Aldar Properties: Q4 and FY 2025 earnings press release, February 2026
- TradeArabia: Globant, Aldar Properties induct AI-powered agents, September 2026
- Gulf News: Dubai Holding partners with Microsoft to scale AI adoption across operations, May 2026
- Government of Dubai Media Office: Dubai Land Department launches Initial Registration, September 2026
- Government of Dubai Media Office: Mohammed bin Rashid chairs UAE Cabinet meeting, April 2026
- Government of Dubai Media Office: Mohammed bin Rashid reviews progress of AI integration, May 2026
- Government of Dubai Media Office: Dubai redefines the future of building permits with an AI system that issues licences automatically, August 2026
- McKinsey: The agentic organization: contours of the next paradigm for the AI era, September 2025
- Clyde & Co: AI liability in the Middle East: who pays when the algorithm gets it wrong?, September 2026
- Mayer Brown: AI regulation in the DIFC, January 2026
- McKinsey: The state of AI in GCC countries, November 2025
- RICS: AI in commercial property and construction report 2026, August 2026
- BCG: AI is starting to pay off, September 2026
- Gartner: Applying uniform governance across AI agents will lead to enterprise AI agent failure, May 2026
Consolidation decides who the principal is. The operating model decides what its agents are allowed to do.


If you want to understand how these frameworks apply to your portfolio
Contact usDownload this quote
The selected quote appears here.