For years, travel executives have called their decision-making "data-driven." Artificial intelligence (AI) is about to reveal how much of it was actually instinct-led, dressed up with dashboards.
AI won't just automate tasks. It will surface insights that challenge assumptions, expose where intuition has been substituted for evidence and enable a level of operational intelligence that will make us wonder how we ever managed without it.
Yet nearly one third (32%) of travel companies are not using agentic AI at all, according to Phocuswright research.
On the multi-day travel front, brands are still trying to figure out where AI fits. Executives see impressive demos of agentic platforms, while frontline teams still wrestle with quoting, rooming lists and supplier emails.
The following phased model for AI adoption reflects how tour and cruise operators actually work. The brands who want to lead, not follow, will need to think differently about how AI-first applications are built from the ground up.
Phase 1: Assistive AI—the gateway drug
Most organizations start here: leveraging general-purpose AI tools like ChatGPT to work as a digital assistant.
Real-world applications include:
- Sales teams using AI to draft personalized quote emails and itinerary summaries
- Product teams generate tour descriptions in minutes versus hours
- Chatbots handling basic traveler FAQs, from visa requirements to insurance questions.
For example, at leading rail operator Railbookers, teams analyze anonymized call transcripts and interaction logs to surface trending questions and the "why" behind demand spikes on specific routes. Those insights flow back into merchandising and training. Early gains are real. This phase builds confidence, creates shared vocabulary and helps identify where automation delivers impact.
But here's the trap: It feels like transformation without being truly transformative. Teams get excited. Leadership checks the "AI strategy" box. And then nothing scales, because prompts alone can't rewire how a business operates.
The most forward-thinking leaders use this phase to define guardrails and start asking harder questions about what their systems can actually support.
Phase 2: Embedded AI—where implementations quickly die
This is where productivity gains become meaningful, but also where most AI initiatives stall.
When AI is woven into the systems that teams already live in, it can act autonomously inside workflows: reducing manual input, improving collaboration and actually doing work rather than just acting as a human assistant.
Use cases in multi-day travel include:
- Automated triggers: Sending supplier requests or custom traveler emails from customer relationship management (CRM) events like booking confirmations or deposits received.
- Dynamic itinerary creation: Generating draft itineraries using pre-set rules, content blocks, and pricing logic, ready for advisors to refine.
- Smart summaries: Capturing phone calls or form submissions and converting them into CRM records.
Virgin Voyages has partnered with Google Cloud to deploy more than 50 specialized AI agents using Gemini Enterprise. One, known internally as “Email Ellie,” autonomously generates marketing campaigns in the brand’s voice, optimizes send times, and routes distribution.
Similarly, Tour Partner Group built a custom AI tool to process inbound hotel availability emails—categorizing, extracting and generating suggested replies. What once took 30 minutes per query now takes 15 seconds.
This is AI embedded directly into operational workflows. But phase two is where I've watched implementations quietly die.
Operators spend considerable resources wiring AI into itinerary-building workflows, only to discover a fundamental integration challenge in the user experience: legacy itinerary building systems were never designed for AI-style interaction patterns or AI load.
The result is operational risk, personified by systems that can't handle the real-time data requirements that AI-driven workflows demand. What worked reliably for years under human control suddenly becomes unstable when AI agents start firing requests at scale.
Most multi-day travel brands don't yet have the infrastructure—or in-house expertise—to scale this on their own. And they shouldn't be ashamed of that. The question is whether you're building on a foundation that can grow, or bolting artificial intelligence onto something that was never designed for it.
Phase 3: Decision intelligence from dashboards to debate partners
If the first two phases are about efficiency, phase three is about using AI to actually challenge how decisions get made.
For years, we've called our decision-making "data-driven," but what we really had was gut instinct validated by selective dashboard views.
AI changes that equation entirely. It can synthesize context across every touchpoint, surface contradictions between what we say we prioritize and what our data shows we prioritize, and force conversations we've been avoiding.
But here's what I haven't seen yet: an operator that has done a full rethink of their decision intelligence layer. Not just bolting AI onto existing dashboards, but building a contextual layer that connects pricing, supplier performance, customer behavior, margin trends and operational constraints into a system that can genuinely debate strategy with leadership.
Most business intellligence (BI) stacks today are archeological sites: Layers of reports built for questions someone asked three years ago, sitting alongside dashboards no one trusts, fed by data pipelines no one fully understands.
The operators who will lead in the next decade aren’t just adopting AI. They’re rethinking the foundation and rebuilding their platforms around unified data, event-driven workflows and decision logic designed for AI from day one.
An uncomfortable truth: You can't out-prompt a broken foundation.
Here's the real problem.
When you engineer AI onto fragmented, aging systems, you inherit their blind spots. The AI might be brilliant, but it can only see what you've designed it to see.
The operators who will lead in the next decade aren't just adopting AI. They’re redesigning their platforms so data, workflows and decision logic live in one integrated system built for AI from day one.
About the author ...
Ragnar Fjölnisson is the co-founder and CTO of
Kaptio.