build_circle
Organiser · The Advisory Clinic
AI Interfaces & User Experience
calendar_today Wednesday 3 June · Afternoon
schedule 40 minutes
groups Two facilitators
James Arnold
Digital Trends Analyst
Digital Tourism Think Tank
Digital Tourism Think Tank
Fábio Caldeira
Digital Trends Analyst
Digital Tourism Think Tank
Digital Tourism Think Tank
1 / 8
Patterns across the submissions
A snapshot of the priorities surfaced in the pre-event Typeform, drawn from eight respondents. The Clinic will focus on the highest ranked themes.
PRIORITY 01
smart_toy
75%
Visitor experiences
The top ranked priority, using AI to improve the visitor journey.
PRIORITY 02
security
50%
Brand voice
Keeping AI interfaces sounding like your destination.
PRIORITY 03
integration_instructions
25%
Data sources
Connecting the inputs an interface draws on.
PRIORITY 04
query_stats
25%
Functional tools
Building interfaces that solve a clear visitor need.
PRIORITY 05
person_pin
13%
Personalisation
Tailoring the visitor journey without losing accuracy.
PRIORITY 06
hub
0%
Native integrations
Embedding AI inside the website rather than bolting it on.
Personal AI maturity (3.1 / 5)
62%
Organisational AI maturity (2.1 / 5)
43%
Maturity gap between individual and organisation
20%
2 / 8
Easy steps to help them implement this
STEP 01
Define the visitor problem you are solving before choosing an interface. Starting from the technology rather than the need produces something impressive briefly and useful rarely.
STEP 02
Audit your data layer before building anything visitor-facing. An interface is only as useful as the information it can draw on.
STEP 03
Plan for native integration from the start. Interfaces bolted on as features underperform those embedded in the visitor journey.
STEP 04
Build a review rhythm into the planning from day one. AI capability changes quickly and an interface that is not actively maintained becomes outdated faster than most other digital investments.
Tools worth knowing
Claude and ChatGPT
Use either to prototype and test an AI interface concept with real visitor queries before committing to any technical build.
Satisfi Labs
The most established AI interface platform built specifically for the destination sector. Powers visitor assistants across web, mobile and in-destination screens, with a live city-wide network of connected agents in Columbus.
Mindtrip
Integrates directly into DMO websites, turning destination content into personalised itineraries for visitors. Active partnerships with Brand USA, The Bahamas, Visit Costa Rica and others.
3 / 8
01
Challenge One
Personalisation without hallucination: designing safe AI interfaces
An interface that gives inaccurate information damages trust that is hard to recover, so personalisation is only worth building on verified, current destination data.
7:30
7 min 30 sec per challenge
Diagnose
Why does this matter for your destination?
- An AI interface that gives visitors inaccurate information damages trust in ways that are hard to recover from, however impressive the experience feels.
- Personalisation is only valuable when the recommendations it produces are grounded in verified, current destination data.
- The boundary between helpful personalisation and harmful inaccuracy is a data and design question as much as a technology one.
Diagnose
What have you tried so far?
- Have you tested any AI interfaces with live visitor queries and reviewed the outputs critically for accuracy?
- Is your destination data structured and current enough to underpin a personalisation layer reliably?
- Where have you already seen AI produce inaccurate or misleading outputs about your destination?
Constraint
What is the biggest constraint holding you back?
- Is it the quality of the underlying data, the design of the interface, the absence of a review process for outputs, or defining the boundaries the system should operate within?
- Is there internal appetite to accept some level of AI error as a managed risk?
Next move
What is the first step you can take this week?
- Define what your interface should and should not do before building anything. A tight scope produces fewer inaccuracies than an open-ended one.
- Audit the destination data that would underpin the interface and assess honestly whether it is accurate and complete enough to support what you have in mind.
4 / 8
02
Challenge Two
Reaching niche audiences through AI-mediated channels
Niche audiences carry high intent and high value, so reaching them needs content specific enough to match the precise questions they ask.
7:30
7 min 30 sec per challenge
Diagnose
Why does this matter for your destination?
- Niche audiences often carry high intent and high value, so reaching them through AI needs content specific enough to match the precise questions they ask.
- General destination content gets absorbed into general AI responses, while niche-specific content surfaces in niche-specific queries.
- The move from broad messaging toward targeted content is one that AI both accelerates and rewards.
Diagnose
What have you tried so far?
- Have you mapped the niche audiences most valuable to your destination and the queries they are likely to put to AI systems?
- Is your content specific enough to surface in AI responses to niche queries, or does it sit at a general destination level?
- Have you tested how your destination appears when niche-specific questions are put to AI systems?
Constraint
What is the biggest constraint holding you back?
- Is it producing enough niche-specific content to be credible, knowing which niches to prioritise, or understanding what technical specificity AI visibility needs?
- Is there internal resistance to moving from broad destination messaging toward more targeted content?
Next move
What is the first step you can take this week?
- Pick one niche audience and run a set of queries in their language across two or three AI systems. Note whether your destination appears and how it is described.
- Identify whether the content to speak to that audience exists on your site in a form specific and detailed enough to be cited.
5 / 8
03
Challenge Three
Maintaining brand voice and consistency across AI interfaces
An interface that does not sound like your destination undermines brand work done everywhere else, since voice here is a data question as much as a tone one.
7:30
7 min 30 sec per challenge
Diagnose
Why does this matter for your destination?
- An AI interface that does not sound like your destination creates a disconnected experience that undermines brand work done everywhere else.
- Brand voice in an AI context is a data question as much as a tone one, since the system reflects the content and guidance it is given.
- Visitors move between channels, so a chatbot, a content tool and a social AI feature should all feel like they come from the same place.
Diagnose
What have you tried so far?
- Has your brand voice been documented in a way that could guide an AI system, or does it exist mainly as abstract brand principles?
- Have you tested any AI-assisted outputs against your brand guidelines and compared them honestly?
- Where has voice inconsistency already shown up in AI-assisted content or interface outputs?
Constraint
What is the biggest constraint holding you back?
- Is it that the brand voice is not defined precisely enough for an AI system to replicate, that the team building AI tools sits apart from the team that owns the brand, or that holding voice at the volume AI enables feels unmanageable?
- Is there a tendency to accept voice drift because the alternative feels too labour-intensive?
Next move
What is the first step you can take this week?
- Write a one-page voice guide with specific examples, vocabulary and tone notes and test it as a system prompt or project knowledge input.
- Take one AI-assisted output currently in use and assess it honestly against your brand guidelines.
6 / 8
04
Challenge Four
Challenge title to be added once submissions are reviewed
The fourth challenge will be confirmed once the final wave of participant submissions has been reviewed. The slide structure mirrors the other three and will be populated with the same depth of facilitator notes.
7:30
7 min 30 sec per challenge
Diagnose
Question to be confirmed once challenge is finalised.
- Facilitator notes will be added once the challenge is confirmed and aligned with submissions.
Diagnose
Question to be confirmed once challenge is finalised.
- Facilitator notes will be added once the challenge is confirmed and aligned with submissions.
Constraint
Question to be confirmed once challenge is finalised.
- Facilitator notes will be added once the challenge is confirmed and aligned with submissions.
Next move
Question to be confirmed once challenge is finalised.
- Facilitator notes will be added once the challenge is confirmed and aligned with submissions.
7 / 8
favorite The Advisory Clinic
Thank you.
lightbulb
Key pattern
The one theme that surfaced across all four challenges.
arrow_forward
First move
The smallest action you can take back to your organisation this week.
block
Common trap
The mistake most organisations make with this topic.
8 / 8