Beyond the Algorithm: Why AI Still Fails at Trust, Culture & Human Connection
Summary
This panel session, moderated by Dr. Jessika Weber-Sabil (Professor of Digital Transformation for Cultural Tourism, Breda University of Applied Sciences), examined why AI continues to fail at building genuine trust, cultural understanding, and human connection in travel and hospitality — despite rapid adoption. Five panelists contributed from industry and academia: Laurenz Schwarzhappel (CEO, Global Living Apartments), Claire Robinson (Co-Founder, Region Lovers AI), Katerina Volchek (Professor, Deggendorf Institute of Technology), Nicholas Hall (CEO, Digital Tourism Think Tank), and Klaas Koerten (Researcher, Hotelschool The Hague).
The session opened with a core distinction: AI does not create trust on its own — it either strengthens or weakens the trust that already exists within an operation. Schwarzhappel reported that Global Living Apartments operates at a 90% automation rate, with only 10% of guests ever requiring human interaction — not because human access is withheld, but because fast, accurate AI responses satisfy the vast majority of queries. The consensus framing was that AI works best for simple, transactional tasks (check-in, checkout, FAQs, parking queries, towel requests) while human touch remains essential for emotional complexity, unexpected situations, and culturally nuanced connection.
A central theme was transparency and the right to know whether you are speaking to a machine or a human. Hall called this 'a simple basic right' that is being ignored in the rush to deploy AI. He praised Slovenia Tourism's chatbot 'Alma' and Germany's 'Emma' as positive examples of clearly disclosed AI interfaces. He revealed that his organisation, Digital Tourism Think Tank, launched a transparency model on the day of the panel. He also shared a striking data point: when Anthropic refused the US government's request for unrestricted military access to its AI capabilities while OpenAI initially agreed (then backtracked), Anthropic saw a 60% growth in users almost immediately — a concrete demonstration that trust and stated values drive adoption.
Koerten shared a personal anecdote from an Airbnb stay: he received messages in Dutch from a French host, felt a genuine emotional connection, then discovered the language had been auto-translated by Airbnb without disclosure. The moment the automated mediation was revealed, the sense of authentic connection evaporated entirely, illustrating how undisclosed AI mediation damages trust even when the underlying intent was positive.
Robinson reframed authenticity: it does not mean human — it means genuine and not copied. She argued that feeding an LLM with a property's own data, local knowledge, and brand voice produces 'authentic' AI responses. Her company, Region Lovers AI, collects granular on-the-ground data — for example, sensory details about museum spaces suitable for visitors on the autistic spectrum (lighting levels, acoustic reverberation, quiet rooms) — that general models like ChatGPT or Gemini cannot provide because they only surface generic or review-averaged information.
Volchek introduced 'cognitive engineering' as a trust layer beyond data: monitoring the words and rhythm a user employs during an AI conversation to detect frustration, then routing them to a human agent. She also argued that websites, which cost approximately 25,000 euro to build five years ago, can now be created for free using tools like Canva or ChatGPT if the user has a clear strategic brief — representing a revolutionary levelling of the playing field for small and medium enterprises.
Hall noted that Amazon's policy of automatically refunding orders under approximately 50 euro — without verifying non-delivery — is cheaper than running human customer service, illustrating how even AI dead-ends are sometimes bypassed by radical cost logic rather than genuine service improvement.
Koerten closed with a presentation of the Hotelschool The Hague 2025 Outlook Report on AI in hospitality (34 pages, available via QR code). He cited that in the Netherlands, hospitality uses AI in under 10% of companies — versus over 50% in information and communication businesses — yet a quarter of Dutch hospitality revenue already comes from companies that have adopted AI. The report identifies five future scenarios: (1) AI Future Baseline — agentic AI on both guest and hotel sides creates seamless but commoditised service, leading guests to become more demanding; (2) Platform Power Shift — guests use AI agents to shield personal data from hotels, leaving hotels blind to their guests' identities; (3) Customer AI Counterforce — hotels use AI to polish imagery and generate positive reviews while guests use AI to see through these manipulations; (4) Nightmare of Modern Times — AI manages employees like cogs in a machine, dictating scripts and efficiency metrics, referencing Chaplin's 'Modern Times'; (5) Worker Empowering Path — AI automates dull and repetitive tasks, freeing humans for genuine connection, the panel's preferred and most aspirational outcome.
The Q&A introduced the question of who owns AI infrastructure and who benefits. Schwarzhappel advised smaller hotels against building proprietary LLMs, recommending they use existing platforms instead. Robinson advocated combining a commercial LLM's language capability with a company's own proprietary data and adding a second LLM as a verification loop to increase answer reliability. Responsibility and accountability for AI outputs — analogous to the US regulatory framework holding self-driving car manufacturers liable for autonomous vehicle actions — was identified as the most pressing governance gap.
Key Takeaways
- •Global Living Apartments operates at a 90% AI automation rate: only 10% of guests ever interact with a human, not because access is restricted but because AI answers are fast and accurate enough that guests do not escalate.
- •Transparency is a basic right being systematically violated: Nicholas Hall stated that almost no company is currently giving users the right to know whether they are speaking to a machine or a human. Slovenia Tourism's 'Alma' and Germany's 'Emma' were cited as rare positive exceptions.
- •Anthropic's refusal of US government requests for unrestricted military AI access produced a 60% surge in users, demonstrating that clearly stated ethical values and governance commitments translate directly into consumer trust and business growth.
- •Undisclosed AI mediation destroys authentic connection: Klaas Koerten's Airbnb anecdote showed that a perceived personal gesture (host writing in Dutch) became a breach of trust the moment the Airbnb auto-translation was discovered — the experience flipped from warm to manipulative without any change in the underlying data.
- •Authenticity does not require human delivery — it requires genuine, non-copied content. Claire Robinson argued that an LLM trained on a hotel's own data, values, and local knowledge produces an authentic response; the key is the quality and specificity of the data fed in ('garbage in, garbage out').
- •Cognitive engineering adds a third trust layer beyond data and language: monitoring a user's word choice and conversational rhythm to detect frustration and triggering automatic human escalation at the right moment makes customers feel genuinely heard.
- •A 'dual LLM' architecture — one LLM constrained to proprietary company data, with a second LLM verifying its outputs — increases answer reliability without requiring the cost or engineering capacity to build a custom model from scratch.
- •Building a proprietary LLM is not recommended for most hospitality businesses: Schwarzhappel and Hall both advised against it due to cost, engineering retention challenges, and lack of strategic necessity — using existing LLMs with proprietary data is sufficient.
Transcript
Hello everyone. Um I am very excited to introduce our next session for today. We are going to be looking at beyond the algorithm why AI still fails at trust culture and human connection. And representing and moderating this session um is a fellow representative uh at a school in the Netherlands. uh she is professor of digital transformation for cultural tourism at Brida University of Applied Sciences. Please give a warm welcome to Jessica Va Zabil, Dr. Professor Doctor. >> Hi, good afternoon. Um...
