Everyone’s talking about agentic AI. ChatGPT, Claude, Gemini. Conversational AI is everywhere. Luxury resorts are being pitched by chatbots by vendors, watching OTAs build AI-powered booking experiences, and feeling overwhelmed by the pressure to “do something with AI.”
Hospitality leaders must slow down, and be deliberate to understand how you build agentic AI matters more than when you build it.
The approach you choose determines not just you’re able to do with AI now, but what type of overall future you will have. Will agentic AI be a differentiator and accelerator for you or once again create that trapped feeling so many feel with the rest of their technology stack.
After guiding enterprise transformations for 25+ years when it comes to agentic AI strategy in hospitality and travel, there are 3 distinct approaches. Each has profound implications for your property’s long-term technology maturity.
The Three Paths to Implement Agentic AI
1. DIY (Build It Yourself)
This is the “we’ll build it ourselves from scratch” approach. It’s often the desire of “in-the-trenches” part of the team. They are capable and want to do “cool stuff”. Your team researches, designs, prototypes, builds, tests, configures, deploys, hosts, maintains, and evolves the solution. Most developers use open-source frameworks like LangChain, Llama, CrewAI, AutoGen, and Haystack as building blocks rather than starting from zero.
The Appeal:
- Maximum control and customization for unique requirements
- Complete ownership of features, data, and integrations
- Flexibility to innovate without vendor constraints
- No vendor lock-in or licensing dependencies
The Reality:
- Requires substantial development resources and AI expertise
- High ongoing maintenance and support burden
- Complex scaling and upgrade challenges
- Significant time investment before seeing results
Who This Works For:
The big chains (ala Marriott, Hyatt, etc) and large hotel groups with dedicated development teams, deep technical expertise, and budget for long-term internal AI capability building. Think major brands with innovation labs, not individual properties or small luxury groups.
The Hospitality Context:
Almost no luxury resort has the internal technical resources to take this approach. If you were to attempt it I’d suggest you’re underestimating the complexity and will end up with abandoned projects or perpetually “in development” solutions.
2. Extensible Platform (The Composable Middle Ground)
This is where platforms like Botpress, Voiceflow, Rasa, and others live. Some of the big names like Google Vertex, Microsoft Copilot Studio, and Salesforce AgentForce are beginning to emerge to help create agentic AI as well. They provide the foundation including the hosting, security, pre-built integrations, development frameworks all while allowing extensive customization and control.
Think of these as sophisticated development platforms rather than point solutions. They offer the router-like functionality, highly configurable layers, and integration capabilities that let you build what you need without starting from scratch.
The Appeal:
- Balance of customization and ease of deployment
- Faster time-to-value than DIY
- Professional support and guidance (CSM teams, documentation, community)
- Can extend across departments and use cases as needs evolve
- Typically includes pre-built integrations with common platforms
The Reality:
- Still requires developer expertise to realize full potential
- Learning curve for the platform itself
- Some platforms may constrain certain customizations
- Multiple vendors in your stack to manage
Who This Works For:
Organizations that need sophisticated, customized AI but lack the resources for full DIY. Properties that understand they need strategic partners to guide implementation while maintaining ownership of the outcome.
The Hospitality Context:
This is the sweet spot for most luxury resort groups. You get enterprise-grade capability without enterprise-scale development teams. You can start with one use case (guest-facing chatbot), prove value, then expand to operations, reservations, and beyond using the same platform foundation.
The Strategic Alignment:
Extensible platforms align with composable architecture principles which aligns with the same philosophy I advocate for your entire technology stack. Best-of-breed tools, open APIs, loosely coupled systems. Your AI platform should fit this model, not force you back into monolithic thinking.
3. Closed Proprietary Solution
These are pre-built, use-case-specific AI solutions. In hospitality, they’re typically “AI chatbot for hotels” or “AI concierge” solutions.
The Appeal:
- Quick implementation (weeks, not months)
- Cost-effective for the specific use case
- Minimal technical demands on your team
- “Turnkey” experience
The Reality:
- Limited expansion beyond predefined use cases
- Less adaptable to your specific needs and brand voice
- High risk of vendor lock-in
- Cannot easily extend to other departments or channels
- Often can’t integrate deeply with your systems
Who This Works For:
Small properties with single, well-defined needs and limited technical resources. Properties that want to experiment with AI at low cost and aren’t concerned about long-term strategic AI capability.
The Hospitality Context:
Most hospitality AI vendors sell closed solutions. They’re optimized for the illusion of plug-n-play for hospitality but at the cost of strategic flexibility. You get a chatbot that answers FAQs on your website, and that’s largely where it ends.
The Strategic Risk:
If AI transformation is strategic to your future (and it should be), closed solutions create the same dependency problems as legacy hospitality vendors. You’re locked into what they build, on their timeline, with their limitations.
The Composable Architecture Connection
Here’s what most hospitality leaders miss: Your agentic AI approach for hospitality should align with your broader technology architecture philosophy.
If you’re building (or should be building) a composable technology stack with:
- Data warehouse at the strategic center (not PMS)
- Best-in-class tools for each function
- Open APIs and integration flexibility
- Your organization governing the architecture
Then why would you accept a closed, proprietary AI solution that doesn’t fit this model?
Extensible platforms support composable thinking. They let you:
- Connect to your data warehouse and CDP
- Integrate with your content platform
- Deploy across any channel (website, mobile, WhatsApp, ChatGPT, Perplexity)
- Build once, extend everywhere
- Maintain strategic control while leveraging expert partners
This is the MACH principle applied to AI: Microservices, API-first, Cloud-native, Headless.
What Luxury Hospitality Actually Needs from Agentic AI
Let me be direct about what works for most luxury resorts:
Start with extensible platform, guided by expert partners.
Here’s why:
- You lack the internal AI expertise for DIY
That’s not a criticism. It’s reality. I’m working at this every day I it’s hard for me to keep up. Building AI agents from scratch requires specialized knowledge most hospitality organizations don’t have and shouldn’t need to acquire. - You need more than a point solution
If you’re thinking strategically (and you should be), you’ll want AI that spans guest-facing, operations, and back-office. Closed solutions can’t scale across these needs. - You need to own the outcome
Extensible platforms let you maintain strategic ownership while partnering with experts. You’re not captive to a vendor’s roadmap or locked into their limited feature set. - You need it to work now AND evolve
Quick wins matter (90-day deployments prove value), but you also need a foundation that grows with your AI ambitions over years, not months.
The Questions That Reveal the Right Choice
Not sure which approach fits your property? Ask yourself:
About Resources:
- Do we have dedicated developers with AI expertise?
- Can we commit to long-term maintenance of custom-built solutions?
- Do we have budget for extended development timelines?
About Strategy:
- Is AI tactical (solve one problem) or strategic (transform how we operate)?
- Do we need AI in multiple departments and channels?
- How important is it that we maintain control and flexibility?
About Risk:
- Can we afford to be locked into a vendor’s capabilities and timeline?
- What happens if the vendor pivots, gets acquired, or exits hospitality?
- How critical is it that our AI integrates deeply with our other systems?
If you answered “no” to the Resources questions, “strategic” to the Strategy questions, and “very important” to the Risk questions then the extensible platform approach is your answer.
The Anti-Patterns I’ve Seen Fail
Over 25 years guiding digital transformation, I’ve watched certain patterns fail repeatedly:
Pattern 1: The “Let’s Try This Vendor solution” Approach
Property buys hospitality-specific chatbot from vendor. Launches on website. It can answer basic FAQs but can’t access guest data, can’t book anything, can’t personalize experiences. Results disappoint. Investment wasted. Now property is skeptical of “AI.”
Why It Failed: Closed solution without foundation (no data access, no content structure, no integration capability).
Pattern 2: The “Our Developers Can Build This” Approach
CTO decides team will build custom AI agents. Six months in, still in development. Realize they need vector databases, knowledge base architecture, integration frameworks. Project scope creeps. Timeline extends. Eventually abandoned or stuck in perpetual beta.
Why It Failed: Underestimated complexity of building and maintaining AI infrastructure.
Pattern 3: The “Tool of the Month” Approach
Property buys chatbot from Vendor A. Then adds AI-powered email from Vendor B. Then tries AI booking assistant from Vendor C. Now has three disconnected AI solutions that don’t share knowledge, don’t coordinate, and create fragmented guest experiences.
Why It Failed: No strategic AI architecture. Point solutions proliferate without cohesion.
The Strategic Pattern That Works
Here’s what success looks like:
Phase 1: Start with Foundation + Quick Win
Choose extensible platform, partner with experts who guide implementation. Deploy first use case (guest-facing chatbot or operations assistant) in 90 days. Prove value immediately.
Build the knowledge base architecture that makes AI actually work and not just today’s chatbot but future use cases.
Phase 2: Expand Strategically
With foundation in place, extend to additional use cases using the same platform. Operations, reservations, and back-office where each new deployment is faster because the foundation exists.
The Knowledge base grows, integrations deepen, AI capability matures.
Phase 3: Own Your AI Future
Platform approach gives you channel agility. When new AI-powered booking channels emerge (and they will), you can deploy there without rebuilding. When your needs evolve, you extend the platform rather than switching vendors.
You maintain strategic control. Your partner guides and builds, but YOU own the outcome.
The Composable AI Stack
This is what your AI architecture should look like if you’re building strategically:
Foundation Layer:
- Data warehouse (not PMS) as strategic center
- Content platform with structured, machine-readable content
- Knowledge base combining structured + unstructured data
AI Platform Layer:
- Extensible agentic AI platform (Voiceflow, Botpress, Rasa, or similar)
- Vector database for semantic search
- Integration framework connecting all systems
Deployment Layer:
- First-party channels (website, app, SMS, WhatsApp)
- Third-party channels (ChatGPT, Perplexity, social platforms)
- Operational interfaces (staff-facing copilots)
Governance Layer:
- Brand voice and tone consistency
- Privacy and consent management
- Quality monitoring and improvement
- Security and compliance
This stack is composable (best-of-breed tools), open (API-first integration), and strategic (you govern the architecture).
The Hard Truth About Vendor Promises
Every AI vendor will tell you they can solve your problems. Most hospitality AI vendors offer closed, proprietary solutions and claim that’s all you need.
Here’s what they won’t tell you:
“Our solution is hospitality-specific” often means “We built a chatbot that knows hotel FAQs and we’re selling it as AI transformation.”
“Quick implementation” often means “Limited to what we pre-built, can’t be customized for your unique needs.”
“Integrated with your PMS” often means “Basic API connection, not deep data access or strategic architecture.”
“AI-powered” has become meaningless marketing speak. Every vendor claims it. Ask specifically: What LLMs do you use? Can I access my knowledge base? Can I extend to other channels? What happens if I want to switch platforms?
The vendors pushing closed solutions are following the same playbook legacy hospitality vendors have used for decades: Create dependency, limit your options, keep you locked in.
Don’t fall for it with AI the same way hospitality fell for it with PMS, CMS, and CRM.
What This Means for Your Next Steps
If you’re considering agentic AI for your property, here’s my guidance:
For most luxury resorts:
Choose extensible platform approach. Partner with experts who can guide you through implementation while you maintain strategic ownership. Start with one use case, prove value, build foundation, then expand.
Look for platforms that:
- Support your composable architecture philosophy
- Provide CSM guidance and expertise
- Offer pre-built integrations but allow deep customization
- Can scale across multiple use cases and channels
- Let you maintain ownership of your AI strategy
For large hotel groups with strong technical teams:
Consider hybrid approach of an extensible platform for rapid deployment, with DIY components for truly unique requirements. Partner on architecture, selectively build custom where justified.
For small properties testing AI:
Closed solution might be acceptable as an experiment, but go in with eyes open about limitations. Budget for eventual migration to strategic platform when you’re ready to scale.
The Bottom Line
The AI agent gold rush is happening now. Properties are making implementation decisions that will shape their AI capability for years.
Most are choosing based on what gets deployed fastest or what the sales pitch sounded best.
Choose based on strategic architecture instead.
Ask yourself: Five years from now, when AI is central to how we operate, will this approach have positioned us to lead or left us locked into vendor limitations?
The answer to that question should drive your implementation choice.
Luxury hospitality deserves modern, open, composable AI architecture. Not decade-old vendor thinking wrapped in AI buzzwords.
You are the author of your AI outcome. Choose the approach that keeps it that way.
Frequently Asked Questions
What is the difference between an extensible AI platform and a closed proprietary solution?
An extensible AI platform provides a development framework with pre-built components, professional support, and the ability to customize and extend functionality across multiple use cases. You can typically integrate deeply with your systems, deploy across any channel, and build sophisticated workflows. Closed proprietary solutions offer specific pre-built functionality (like a hotel chatbot) that works out of the box but has limited customization and can't easily extend beyond the vendor's defined use cases.
Can small luxury resorts build AI agents themselves without a large technical team?
Generally no. Building AI agents from scratch requires specialized expertise in LLMs, vector databases, knowledge architecture, and complex integrations. Most small and mid-sized luxury resorts lack these resources. However, you can absolutely deploy sophisticated AI using an extensible platform with expert partnership—this gives you enterprise capabilities without needing an enterprise development team.
How does composable architecture apply to AI implementation?
Composable architecture for AI means building your AI capability from best-of-breed, independently deployable components connected through APIs. Your AI platform integrates with your data warehouse, content platform, and operational systems rather than trying to be an all-in-one solution. This approach prevents vendor lock-in, allows you to swap components as better options emerge, and gives you strategic control over your AI architecture.
What is MACH architecture and why does it matter for AI agents?
MACH stands for Microservices, API-first, Cloud-native, and Headless. For AI agents, this means building with: Microservices (independent components that can be updated separately), API-first design (everything connects through well-documented APIs), Cloud-native infrastructure (scalable, always-available), and Headless deployment (separate AI logic from presentation layer so you can deploy across any channel). This architecture enables the flexibility and channel agility modern hospitality requires.
Will an extensible platform vendor lock me in just like legacy hospitality vendors?
The key difference is architectural openness. Good extensible platforms are built on open standards, provide API access to your data and models, support standard protocols, and allow you to export your work. They make money by providing value and expertise, not by trapping you. That said, always evaluate a platform's export capabilities, data ownership policies, and integration openness before committing. True extensible platforms should support your ability to migrate if needed.
How long does it take to implement an AI agent using each approach?
DIY approach typically takes 6-12 months for initial deployment, then ongoing development cycles. Extensible platforms can deploy initial use cases in 90 days with proper foundation work, then expand to additional use cases more quickly (4-8 weeks) as foundation matures. Closed proprietary solutions can launch in 2-4 weeks but often deliver limited functionality. The key question isn't just 'how fast' but 'how strategic'—quick deployment that doesn't scale wastes time and money.
What does 'knowledge base architecture' mean and why does my AI agent need it?
Knowledge base architecture is the structured system that makes AI actually useful. It combines your structured data (from data warehouse and operational systems) with unstructured knowledge (policies, procedures, tribal wisdom) in a format AI can access and understand. Without proper knowledge base architecture, your AI agent can't answer questions accurately, can't access the context it needs, and becomes just another failed chatbot. This is why point solutions often disappoint—they skip the foundation work that makes AI effective.
Can I start with a closed solution and migrate to an extensible platform later?
Technically yes, but you'll likely rebuild rather than migrate. Closed solutions typically don't expose the underlying data, workflows, or training in formats extensible platforms can import. You'll need to recreate your knowledge base, rebuild integrations, and redesign workflows. This is why starting with strategic architecture matters—the work you do building foundation in an extensible platform compounds over time, while closed solutions often represent throw-away investment when you eventually need more capability.
