Commercial implications of Agentic AI: A practical example of the travel shopping disruption
While Agentic AI enables significant advancements in capabilities and user experience, its most transformative impact will be in unlocking new business models and revenue streams through dynamic data exchange and real-time service integration of strategic data sets.
The impact of Generative AI (GenAI) has been substantial, yet it pales in comparison to the potential of Agentic AI. Traditional GenAI assistants have mostly operated as closed systems, unable to interact with external data at scale. These technical limitations have constrained their commercial potential.
In comparison, Agentic AI breaks down these limitations by enabling seamless communication between AI systems and allowing rich API-like data exchange across different platforms and organizations. This evolution represents more than just technological progress - it creates an entirely new AI-driven ecosystem that will transform how companies interact with customers, work with partners, and monetize data and AI capabilities. Businesses will therefore be able to adapt their revenue models through advertising, data monetization, purchase commissions, or other structures that remain to be invented.
To understand the magnitude of this transformation, let's examine the evolution of GenAI systems' data handling capabilities and explore real-world examples from the travel industry, where these changes are already taking shape
The Constraints of Static Training Data in GenAI assistants
Large Language Models (LLMs) form the foundation of Generative AI. These models are trained using vast datasets provided by their respective providers. When a model is released, model users can access this content generated probabilistically using the model parameters (or "weights"). This knowledge remains unchanged until the next model update.
This approach differs significantly from traditional websites, where data is retrieved in real-time and processed synchronously in a deterministic way - a process that also generates valuable web user analytics. To address these technical differences, Generative Engine Optimization (GEO) has emerged, helping content creators optimize their visibility within GenAI assistants.
Take this typical travel question: 'Which Mediterranean destination would be best for a romantic trip in September?'
The LLM's response relies entirely on its training data, which was loaded months earlier during model development. The answer provided is usually a blend of information pieced together from different sources in its training data - for example Wikipedia entries or travel blogs and local tourism sites.
While foundational models are technologically impressive, their self-contained and static nature constrains their business applications to a limited set of use cases and severely limits data-driven monetization opportunities. Currently, the primary commercial strategy has focused on acquiring extensive licensed datasets for training purposes.
Fine-tuning & RAG: Early Steps Toward AI Data Monetization
LLM operators can enhance their AI systems' knowledge using two key methods: Fine-tuning and Retrieval-Augmented Generation (RAG). Fine-tuning updates the model through additional batch training with specific datasets, while RAG allows the model to access contextual data in real-time. RAG has proven especially valuable for enterprise chatbots, enabling them to incorporate current information and permission-restricted company data.
Here's an illustrative example of travel questions using restricted data: 'First, find Anna's Venice nightlife suggestions from our earlier conversations. Then, compare these with the recommendations in my travel guide subscriptions.'
The LLM's response is augmented using user-specific textual data such as previous private conversations or subscriptions available to the user.
These approaches give model operators greater control over data integration, opening monetization opportunities through proprietary data access. A travel guide company, for example, could charge subscribers to access an AI assistant powered by their exclusive content. However, both fine-tuning and RAG have significant limitations when handling volatile or complex datasets. As a result, commercial potential through data-related monetization has remained largely constrained.
Agentic AI: Creating a Monetizable Ecosystem Through Inter-Agent Data Exchanges
The emergence of AI Agents transforms what Generative AI can do. Using protocols like Model Context Protocol (MCP), these systems can actively and seamlessly engage with other AI agents across different platforms and organizations. This allows a single AI application to handle complex tasks by gathering live granular data from multiple sources and orchestrating actions across systems, while delivering clear and consistent results to users.
Take this complex travel query: 'Find roundtrip flights to Nice under £200 for September 1-15. If nothing is available, check October dates. Before booking, please factor in my loyalty programs, check hotel rates, and verify baggage fees'
Unlike earlier AI systems, this request requires to actively gather real-time data from multiple travel providers through API connections. Each provider controls how their data can be accessed and used, creating a dynamic data exchange over strategic industry datasets.
The ability of AI Agents to interact directly with key data sources in real-time unlocks numerous monetization opportunities. Businesses can now explore various revenue models such as:
- Serving contextual and personalized ads
- Charging for data access
- Earning commissions on transactions
- Potentially developing innovative pricing structures that remain to be invented.
The emergence of an AI super-app ecosystem in the travel industry
AI chatbots have quickly become key players in travel planning. Traffic from AI platforms to travel websites continues to surge dramatically, growing over 1700% annually . Mastering the AI channel has become crucial for travel brands:
- A rapidly growing share of travelers - now 37% - use AI for trip planning (2025 Skift report).
- While traditional search engine conversion rates continue to fall , AI channels are a critical alternative for marketers since they convert at higher rates on travel websites.
Notably, this shift in customer behavior occurred even though AI chatbots had limited shopping and booking capabilities, and content providers hadn't yet optimized for them. AI Agents are now poised to remove these barriers.
Travel booking remains a pain point for most customers, with 70% describing it as stressful . AI Agents offer a solution by enabling a more collaborative and integrated booking experience. Moreover, their ability to interact with data-rich sources - from smartphones to social media - means they can deliver increasingly personalized and relevant results that travelers have long desired.
The enhanced customer experience is just the visible part. Behind the scenes, Agentic AI also enables a fundamental transformation of the business model. This new ecosystem opens up multiple revenue streams - from advertising and sales commissions to innovative ways of monetizing customer and inventory data.
To glimpse the future, consider Google Travel's disruption of the industry during the 2010 decade. As most travelers began their trip planning with search engines, Google responded by creating an integrated travel shopping workflow that organized travel data (and sometimes enabled direct bookings) to streamline the planning process. The product and its associated revenue model, based on advertising and referral fees, proved so successful that Google Travel became the largest travel tech company within a decade .
Adapting to AI Agents: A Strategic Outlook
Two key trends - improving AI capabilities and growing consumer trust - are set to revolutionize customer expectations and revenue models. Given the potential for disruption, AI agents demand attention at the highest executive level. Preparation requires close attention to both technical developments and evolving customer behavior. Companies should prioritize enhancing their customer analytics capabilities while prototyping scalable AI agents, related tools, and connectivity (such as MCP or A2A). Building internal expertise and technical skills across the organization is equally crucial for future readiness.
These changes may arrive quickly, testing companies' technological readiness and ability to adapt. Success will depend largely on data architecture and customer-centricity: organizations that can rapidly deploy and control valuable data at scale will have a competitive advantage in this AI-driven future.
Hands-on: Testing AI Agent Workflows for Airlines
Following the AWS solution guidance , creating an example MCP server specifically designed for travel shopping can be achieved in minutes. The server provides travel inspiration and pre-shopping capabilities that MCP-compatible AI applications can access and use:
1
2
3
4
5
6
7
8
9
10
11
12
@mcp.tool()
async def get_air_destinations_and_respective_cheapest_prices(iOriginAirportCode: str) -> str:
"""
Provides an array of available destinations and respective cheapest price per roundtrip passenger from an origin airport. Insist on the fact that prices vary by date and itinerary and users need to look for detailed flight options next."
Parameters:
- iOriginAirportCode: Origin IATA airport code
Returns:
- List of destinations and their respective lowest prices. Prices are provided in GBP per adult pax.
"""
return str([{'destination_airport_code': 'VCE', 'lowest_price': 250}, etc...])
After server deployment on AWS, compatible AI assistants like Anthropic Claude Desktop, Amazon Q CLI, or assistants implemented using Amazon Bedrock and/or the Strands framework can automatically discover and seamlessly use MCP tools:
✓ mcp-sse loaded in 4.30 s.Finally, the tools are available in user queries. The latest LLMs are showing strong reasoning capabilities to use tools wisely, process real-time data, and summarize the outcome of complex tasks to users:
I can help you find good Mediterranean destinations for September! Let me check what flight options and prices are
available from London Heathrow (LHR) to various Mediterranean destinations.A growing list of MCP-compatible tools is maintained on the AWS samples repository.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article