In many property firms, AI adoption has started with a pilot, a few licences and a promise to come back to the board with a plan. Turning that activity into a lasting change in how people work is proving harder. The difficulty extends beyond property: JLL’s global 2026 Future of Work research found that 42% of companies had moved beyond their initial pilots, but only 15% had progressed with AI-driven organisational change.
Closing that gap requires someone with the time, authority and budget to change how the firm works. For some UK property businesses, that should be a Chief AI Officer (CAIO). My view is that most can assign the responsibility elsewhere.
Where a dedicated role is justified, I would give it a three-to-five-year mandate, with a planned handover from the outset. Its purpose should be to establish the tools, skills and working practices that allow the business to take responsibility for AI itself.
Which firms need one?
The case depends on the shape of the business. A property owner with a small team and a large balance sheet has a different opportunity from an agency employing hundreds of fee earners. The owner may see limited scope for staff efficiencies, although better investment analysis or asset management could still be valuable. An agency or consultancy has more opportunities to improve how work moves between people, teams and service lines.
Large agencies, multidisciplinary consultancies and investment managers with complex operations are therefore the strongest candidates. Even then, size alone is insufficient. A dedicated executive becomes more useful when several parts of the business need to change together and nobody currently has the capacity and authority to make that happen.
Before recruiting, the board should be able to identify a programme of work substantial enough to occupy a full-time leader and justify the supporting investment. If an existing COO or CTO can deliver it with additional resources, that may be the better answer. A new title will achieve little if the underlying problem is an unwillingness to fund or support change.
Smaller firms may be better served by a Head of AI or a COO supported by a part-time adviser. An AI-literate non-executive director can help the board challenge proposals and scrutinise risk, but someone inside the business still needs to deliver the work.
The financial backdrop helps explain the hesitation. Boston Consulting Group’s May 2026 report, The AI-First Real Estate Company: An Opportunity for Structural Advantage, estimates that real estate’s AI investment is roughly half the cross-industry average, behind even utilities. The market also leaves boards cautious: CBRE’s UK Real Estate Market Outlook Midyear Review 2026 reported softer UK transaction volumes and continuing uncertainty over financing conditions. In that environment, another senior appointment needs a convincing commercial case.
There is also a question of how much change leaders believe is necessary. Many see client relationships and professional judgement as protection against disruption. In CBRE’s 2025 Annual Report, published in 2026, its chairman and chief executive described transactional activities as the
part of the business most protected from AI disruption, while also identifying opportunities to improve efficiency and services. That distinction is useful: confidence in the value of advice can coexist with a strong case for changing how it is delivered.
Established working habits reinforce the caution. Property has generally been more comfortable adopting technology once others have demonstrated its value. AI asks experienced people to reconsider processes they know well, often while they are under pressure to win and deliver
instructions. Whoever takes responsibility will need to make the case in terms colleagues recognise: better delivery, stronger margins and more time for clients.
Give the role authority to deliver
The CEO should set the ambition and remain accountable for the investment. The CAIO should lead implementation with the CTO and the heads of each business unit.
That division reflects the work involved. Someone needs to choose which processes to improve, assess tools, organise training, resolve objections and check whether anything has changed. A CEO can sponsor that work, but will rarely have the time to manage it personally.
The CTO is a plausible owner, particularly where the role already includes business transformation. In other firms, infrastructure, cyber security and existing systems consume most of the technology team’s attention. Adding AI to that remit without changing its capacity or priorities is unlikely to produce much progress.
HSBC’s March 2026 announcement of its first CAIO offers one example of how to divide the work. David Rice took up the role in April after serving as chief operating officer of its corporate and institutional bank. At the same time, HSBC expanded its CTO’s remit to strengthen the technology foundations and build a central AI platform. Business adoption and technical delivery received distinct, connected responsibilities.
In a property firm, the CAIO should agree priorities with business leaders, commission improvements and be accountable for their adoption and results. The CTO should own infrastructure and technical security. Together with the relevant risk and professional leads, they should establish rules for data use, testing and human review. Business heads must remain accountable for the quality of the services their teams deliver.
The reporting line would usually be to the CEO or COO, ideally as a peer of the CTO, with regular reporting to the board. The mandate should give the CAIO authority to allocate the agreed AI budget, require business units to nominate responsible owners and stop projects that fail to meet agreed criteria. Disputes over priorities and risk need a named executive to resolve them.
The budget must cover implementation, training and maintenance as well as the appointment itself. A CAIO with a title, a mandate and no money is an expensive way of producing slides.
External activity should follow internal delivery. Comparing experience with peers can be useful, but nothing loses a room of surveyors faster than a leader on a conference stage describing programmes that have not yet reached their own desks.
The right candidate needs commercial credibility and enough technical depth to challenge proposals and understand their consequences. Gleeds’ leadership profile of its Head of AI and Data, James Garner, describes the quantity surveying background he brings to the role. That combination of professional knowledge and technical understanding is a useful model for the sector.
Start with the people doing the work
Boards may expect resistance from technology teams. Fee earners have their own reasons to hesitate, and those deserve attention before a firm starts handing out licences.
Liability is one. A valuer asked to complete more instructions without extra review time or a change in reward may reasonably see it as more work, more liability, same pay. An adoption programme that overlooks that concern will struggle to win support.
Client relationships are another. In agency and advisory work, an individual’s value often rests on contacts, knowledge and a personal way of working. Making that knowledge easier to share benefits the firm, but may feel less attractive to the person whose bargaining power depends on it.
Pricing adds a further complication. Where fees are agreed per transaction or instruction, more efficient delivery can improve margins, provided quality is maintained and the gains are not wholly passed on through lower fees. In time-based consultancy, the firm also needs to consider how it charges for the resulting work.
These concerns should shape implementation. Start with tasks fee earners find burdensome, involve experienced practitioners in designing checks and review incentives so that quality and productivity are rewarded alongside fees. People need to see how the change helps them as well as the business.
I would favour broad access to approved tools for appropriate tasks, supported by training and clear limits on data use. Give teams room to discover useful applications, then review usage and results after six months. More sensitive uses and internal applications should earn wider deployment through testing and professional review.
Some spending will be written off. The board should agree in advance how much it is prepared to commit to experimentation and what evidence will determine continued funding. There is little value in prescribing a universal percentage: the right amount depends on the firm’s size, existing systems and proposed uses.
Providing useful, approved tools also helps address shadow AI: the use of AI tools without the firm’s approval or oversight. Staff may already be using personal accounts to get work done. A prohibition on its own gives them little reason to bring that activity into view.
IBM’s Cost of a Data Breach Report 2026, based on research by Ponemon Institute into 602 organisations that had suffered a breach, found that 43% reported security incidents involving shadow AI, up from 20% in the previous year’s study. Breaches involving shadow AI cost an average of US$5.39m, compared with the global average of US$4.99m. These are international findings across industries, but the exposure is familiar.
In property, the information involved might include a rent roll, an unpublished valuation, deal terms or a tenant’s personal details. The firm needs to know which tools handle that information and under what conditions.
Publish a short policy explaining what information can be used, where restrictions apply and how to request a new tool. Training should use situations people recognise from their own work. A surveyor needs a clear answer about a valuation report, not an abstract lecture on responsible AI.
Look beyond the software licence
Approved tools may meet many of those needs. Buying licences for tools such as Claude or ChatGPT can be a sensible starting point, but the CAIO’s remit should also cover building and managing systems around the firm’s own information and working practices.
Where the business case warrants it, that can include deploying an open-weight model: a model available to download and run within infrastructure the firm controls, subject to its licence. The team can then test whether fine-tuning it on carefully selected examples improves its performance
on particular tasks. For a property business, those tasks might include extracting information from leases or producing consistent first drafts of standard documents. The starting point is an existing model that the firm adapts. Mistral AI described this approach in its announcement, My Tailor is Mistral, explaining how organisations could fine-tune its models on their own infrastructure.
Fine-tuning and access to current information serve different purposes. A model might be trained to follow a particular format, while the surrounding system retrieves the latest approved documents for each instruction. The CAIO needs enough understanding of both to commission useful work and question unnecessary complexity.
The team also needs to build the “harness”: the software around the model that supplies context, connects it to tools and manages the steps it takes. That might determine which files an assistant can retrieve, which calculations it can run and when it must stop for a surveyor’s approval. Anthropic’s April 2026 article, Trustworthy agents in practice, explains why the model, its instructions, tools and operating environment need to be considered together.
These decisions create an ongoing responsibility. Internal models and applications need testing against real tasks, monitoring, maintenance and controlled updates. The CAIO should be accountable for their usefulness and performance, working with the CTO on reliable and secure operation.
The team can start small, but its capacity must match what it is expected to run. Building internal systems requires continuing access to AI, software and data engineering expertise, whether employed directly or provided by partners. A firm should compare the full cost and performance of that approach with buying an established service. Deciding what to buy, what to build and what the business can maintain is a central part of the CAIO’s job.
Measure what changes
The CAIO should report progress at three levels:
- Adoption: who uses the tools, how often and for which tasks.
- Capacity and quality: what time has been released, how it has been used and whether
accuracy, rework and turnaround times have improved. - Commercial outcomes: changes in margins, fee income, instructions per head and client
satisfaction, assessed against the full cost of the programme.
The board should expect early evidence of useful adoption. Within six months, it should be possible to show which processes have improved, what remains unresolved and which investments should stop. Larger commercial effects may take 12 to 24 months to assess.
Hours saved do not automatically become a financial return. If a team spends less time preparing reports, management still needs to decide whether that capacity will support more instructions, better analysis, faster delivery or less overtime. Each may be valuable, but they produce different results.
Some savings can arrive sooner. The CAIO and CTO should review the software estate together, looking for duplicate subscriptions and tools that no longer justify their cost. Any replacement should be assessed against the cost of migration, integration and ongoing support.
A baseline matters throughout. Fee income can rise because the market improves, and staff can become busier without becoming more productive. The CAIO should distinguish measured improvements from estimates and wider changes in trading conditions.
Design the role to end
The lasting work of a CAIO is to change everyday practice: how instructions are handled, where professional review sits, how staff learn and how managers assess performance.
Those responsibilities need to become part of running the business. That is why I would set an initial mandate of three to five years, with a handover plan reviewed annually.
By the end of it, each business unit should own its AI uses and results. Training should be part of normal staff development. Approved tools, data rules and review procedures should be established. Internal models and applications should have named technical owners, maintenance budgets and a process for testing changes.
The specialist work will continue. Depending on the firm, it may sit with the CTO, the COO or a permanent Head of AI. Ending the CAIO role means transferring that work and its accountability into an organisation equipped to sustain it.
Any extension should be justified by specific unfinished work and a revised timetable. A board should be able to explain what still requires a separate executive and what would allow that responsibility to be handed over.
Boards considering an appointment can start with three questions:
- Who owns AI adoption, by name?
- What budget and decision rights do they have?
- When will their job be done?
