Enterprise AI rollouts can fail for the same reason restaurant openings fail: The concept is exciting, but the operation behind it cannot execute consistently. In the same way, an AI assistant cannot create dependable value when the company’s data is fragmented, its systems do not communicate, and employees do not trust the output.
For real estate professionals, the equivalent “kitchen” may include everything from listings platforms to customer relationship platforms, transactions platforms, property-management systems, accounting databases and individual team spreadsheets.
At Langham Hospitality Group, Sean Seah, Senior Vice President of Strategy, Technology and Innovation, and his team identified a foundational problem: Core systems (property management, CRM, point of sale, housekeeping) did not communicate effectively and they involved slow and siloed data. Rather than placing AI above those systems, the company addressed the underlying environment before developing its AI toolkit.
Announced on November 27, 2025, the toolkit consists of three coordinated agents for guest service, employee knowledge and commercial insight, to be phased across Langham’s 31-property portfolio. These agents would be in service to these objectives: increase occupancy and average daily rate, improve employee productivity and reduce inefficiencies, and elevate guest services and experiences.
The toolkit is still in its early days; its long-term effects on its goals are not yet established. So right now, the plan serves as an instructive example of how to prepare an organization to reach transformation. The lesson is this: Before selecting ambitious AI use cases, determine whether the underlying operation can support them.
Menu first or kitchen first?
Should a restaurateur design a dream menu and build a kitchen around it, or assess the kitchen’s capabilities and create a menu it can execute consistently? Langham faced the same choice with AI and chose the pragmatic route. It assessed its technology stack, record quality, integrations, security requirements and workforce readiness before finalizing the toolkit’s functions.
This became the first phase of a three-phase methodology: Ideate, incubate and industrialize.
The ideation step brought together a cross-functional group from legal, finance, IT, information security, branding and sales and marketing to evaluate initiatives and map dependencies.
The initial priority was consolidating records into a faster, more accessible cloud environment.
After that groundwork, Langham defined its menu — the Experience Agent, Knowledge Agent and Insight Agent.
Scope and scale
The Experience Agent lets guests ask practical questions through email and social media platforms while preserving the ability to speak with hotel staff. Questions range from broad to specific, and the agent can respond in more than 50 languages.
The Knowledge Agent is an internal resource. Employees can ask about operating procedures, housekeeping standards, brand practices or HR policies in natural language rather than searching across multiple documents.
The Insight Agent gives commercial and revenue teams immediate access to booking patterns, demand signals and aggregated guest behavior, surfacing recommendations about timing, pricing, audiences and campaigns without routing every question through an analyst.
Local design and governance
Instead of imposing one generic implementation, Langham gave each hotel a locally configured “digital co-worker.” Property teams add specific hotel and market context, encouraging ownership, while corporate retains the brand voice, service rituals and governance. This preserves consistency while reflecting the unique differences among the properties.
Langham introduced information-governance protocols during development. Two non-negotiable concerns were brand alignment and security. The interface had to reflect Langham’s positioning and the technology required vulnerability assessments, guardrails and training.
For a real estate organization, that governance layer should also address data licenses, privacy, source verification, fair housing and client communications, as well as human review.
Piloting, iterating and scaling
Langham’s second phase, incubation, is the soft opening: At a technology-ready property, Seah’s team directs a small portion of web-chat traffic (as little as 1%) to AI agents. The teams monitor stability, guest sentiment, employee use and operational performance over 90 days, refining the system with frontline feedback.
A pilot should be small enough that errors can be contained, but real enough to test the full workflow.
The evaluation should separate activity from value. Chat volume, logins and questions answered are activity metrics. Accuracy, response time, successful human handoffs, labor saved, lead conversion, service recovery, satisfaction, revenue and operating costs are business measures.
The third phase, industrialization, begins once the pilot produces sufficient evidence. It may include expansion to additional properties, standardized training, formalized governance, stronger integrations and ongoing monitoring.
Scaling is a decision, not an automatic next step; define in advance what justifies expansion, what requires another pilot and what stops the project.
Training
Different roles require different levels of instruction, and employees need a safe environment to practice in before relying on a system with a guest. Langham addressed this through its three-tier AI Academy. “AI for Everyone” introduces foundational concepts: “AI for Leaders” focuses on strategic and managerial responsibilities; “AI for Experts” provided the technical depth to configure and optimize tools.
Managers need to understand performance measures and legal exposure; technical specialists own data connections, access controls, testing, and incident response. Training is not a launch event; it is part of the system.
The guest experience
Great hospitality feels effortless, and Langham wants the mechanics to fade into the background. The agents handle practical information and repetitive retrieval so employees can focus on the judgment, empathy and personal interactions that define hospitality.
The feedback loop
Guest feedback reveals how the experience felt; employee feedback reveals where the workflow failed; operational and financial measures show whether the tool changed the business.
Langham’s 90-day monitoring cycles combine visitor usage, employee commentary and operating measures to identify where AI interactions succeed, stall, or affect revenue, while central oversight watches for drift across the portfolio and properties maintain local knowledge. The toolkit is a living infrastructure. It requires maintenance, updated information, renewed training and constant security.
Create the feedback loop before the pilot. Who reviews inaccurate responses, how corrections are made, when conversations escalate, how to report bias or compliance, and who owns final decisions.
Final thoughts
Langham’s toolkit is purpose-built, but its defining test is ahead — employee adoption, reliable answers, frictionless human handoff and measurable gains that justify the costs. The practical takeaway for a brokerage, operator or investor is straightforward: Do not begin with the AI menu. Begin with the kitchen. Audit one high-friction workflow. Identify its dependent systems and records. Establish the current cost, response time, error rate or conversion baseline. Pilot one narrow use case with a clear human handoff. Then decide, using evidence over enthusiasm, whether to refine it, scale it or shut it down.
Brandon Stoller is a principal at Current Lodging Advisors, LLC.
The opinions expressed in this column do not necessarily reflect the opinions of CoStar News or CoStar Group and its affiliated companies. Bloggers published on this site are given the freedom to express views that may be controversial, but our goal is to provoke thought and constructive discussion within our reader community. Please feel free to contact an editor with any questions or concern.