In Part 1 of this series, we covered the business case: why the traditional guest services model has a structural ceiling, how AI changes the cost and quality of shopper interactions, and what Cadillac Fairview has been building toward over the past six years. Check How AI Is Transforming Guest Services in Shopping Centres
This is the part most articles skip. Not why AI guest services work, but what they actually look like when you build them at a retail property. The queries, handoffs, technology, data, and what your team does differently once the system is running.
What Shoppers Are Actually Asking
The first thing most property teams want to know is whether an AI system can handle their specific queries. The answer, almost always, is yes, because the query distribution at a shopping centre is heavily concentrated in a small number of repeatable categories.
ICSC's research on AI in shopping centres identifies the most common AI-handled mall queries as store hours, in-mall directions and wayfinding, product information ("does store X carry Y?"), and general FAQs (ICSC Exchange, 2024). Across publicly documented guest services structures at major North American malls, the full list extends to: parking availability and directions, lost and found, accessibility services including wheelchair and stroller rental, security and vehicle assistance, upcoming event information, employment inquiries, transit and getting-here questions, and gift card and promotion queries.
These are informational and transactional queries. They follow predictable patterns. They don't require judgment, de-escalation, or local knowledge. They require accurate, current information delivered immediately. That's exactly what an AI system does well, and exactly what a staffed desk does inconsistently, especially at peak hours and outside operating hours.
A peer-reviewed deployment study published in December 2025 tracked what happened when a retail service operation introduced AI to its customer query workflow. Before implementation, the average response time was 118 minutes, and CSAT scored 3.73 out of 5. By month three post-launch, the AI was handling 85.4% of all queries automatically, average response time had dropped to 43 minutes across all interactions, including escalated ones, and CSAT had climbed to 4.4 out of 5. No staff were replaced. The human team's workload shifted (MDPI, Information Journal, Vol. 16, No. 12, December 2025).
How the Handoff Actually Works
The most common objection to AI guest services is the handoff. What happens when a shopper's question is too complex, too emotional, or too specific for the AI to handle? Who takes over, and does the shopper have to repeat themselves?
A properly designed system has four layers working simultaneously. The detection layer monitors every message for escalation signals: rising frustration in the text, repeated failed attempts by the AI to understand the query, explicit requests for a human agent, or query types flagged as requiring human judgment from the start. The decision layer applies routing logic to determine who handles the escalation and at what priority. The handoff layer transfers the full conversation transcript, sentiment history, and an AI-generated summary to the receiving agent instantly. The feedback loop captures the outcome of every escalated interaction to improve the trigger logic over time (SearchUnify, 2025; BlueTweak, 2025).
When this is done well, 92.6% of chatbot-to-agent handoffs result in satisfied interactions, an all-time benchmark high as of the 2026 Comm100 AI Live Chat Benchmark Report. When it's done poorly, the shopper repeats themselves, loses confidence, and CSAT on the escalated interaction drops sharply. 96% of customers who experience high-effort service interactions become disloyal (Gartner, cited in Lorikeet, March 2026).
The design question isn't whether to have a handoff. It's what triggers it, how context transfers, and who receives it. 98% of customer service leaders say smooth AI-to-human handoffs are essential. 90% admit they struggle to make them work consistently (Nextiva 2025 CX Trends Report). The gap is in design, not in the technology.
The Technology Stack
At the property level, the platform choice matters less than the integration quality. An AI guest services system that can't access live data, current store hours, real-time event schedules, tenant directory changes, or parking status is limited to static FAQ responses. The moment the information it holds is out of date, the system becomes a liability rather than an asset.
The major platforms relevant to retail property deployments include Zendesk, Freshworks, Intercom, and Salesforce Agentforce, each offering omnichannel AI with native handoff capability and CRM integration. For shopping-centre-specific deployments, ICSC has identified Mall Maverick as a platform built specifically for generative AI-ready website ecosystems at mall and mixed-use properties (ICSC Exchange, 2024).
On the channel side, the deployment should cover web chat, WhatsApp Business, Apple Messages for Business, and Facebook Messenger at a minimum. Apple Messages for Business is particularly important for Canadian retail properties because it surfaces directly in Apple Maps, Safari, and Siri, meaning a shopper researching your property before their visit can initiate a conversation without finding a separate contact page. 63% of consumers across generations now prefer messaging a business over calling or emailing (Meta, cited by RingCentral, 2024). The channel has to be where the shopper already is.
A professional-assisted implementation typically takes two to three weeks from kickoff to launch, with AI resolution rates reaching 85% or higher within 90 days of go-live (Fin.ai, 2026). The single biggest determinant of launch performance is knowledge base readiness: the quality and currency of the information the AI is trained on before it goes live.
What Your Team Does Differently
The staff reallocation question deserves a direct answer. AI does not replace a guest services team. It eliminates the part of the job nobody wanted: answering the same ten questions on rotation for eight hours while a queue builds behind them.
83% of contact centre leaders say their agents spend too much time on simple and repetitive interactions (CCW Digital Market Study, 2024). When AI absorbs that volume, the human team shifts to the interactions that actually require them: complex tenant relationship management, proactive shopper engagement, event and activation support, accessibility and special needs assistance, and complaint escalation. IBM's research found that when agents were given access to AI tools, their average productivity increased 14% (IBM Consulting, 2024). McKinsey found AI reduces total service interactions for human agents by 40 to 50% (McKinsey, 2025).
The team doesn't shrink. It changes what it does. And what it does becomes harder to automate and more directly connected to the experience that differentiates your property.
The Data Your Guest Services System Should Be Producing
Every AI interaction at your property is a structured data collection event. Query type tells you what shoppers don't know when they arrive. Visit timing tells you when engagement peaks. Channel preference tells you how different demographics want to communicate. Escalation reasons tell you where your service model has gaps. Sentiment signals across thousands of interactions tell you more about aggregate shopper satisfaction than any point-in-time survey.
72% of consumers say they don't care whether their query is handled by a bot or a human, as long as their issue is resolved efficiently (Genesys). What they do care about is that the property knows them well enough to be useful the next time they interact. That's the data dividend that a properly integrated AI guest services system starts building from day one.
The question isn't whether your property can afford to build this. It's whether you can afford to keep answering the same questions manually while the data that would make every other marketing decision more accurate goes uncaptured.
Interested in what an AI guest services program would look like at your property? Book a meeting now
