The Memory Advantage: Why Rental Car Brands Must Act Now on Agentic Commerce
AI agents are becoming a new distribution channel for travel. Bob Kwik, Dean Smith and Vinicius Pedroni explore how rental car brands can use agentic commerce to win new customers, deepen customer relationships with every interaction, and streamline franchise operations.
A traveler's AI just booked a car. It never visited your website.
A business traveler lands in Denver on Tuesday evening. They didn't search for rental cars. They didn't compare rates on a metasearch site. They didn't download an app. Their AI assistant confirmed the flight, checked their calendar, and booked a midsize SUV at the airport. Loyalty rate applied. GPS pre-loaded with the hotel address. Insurance declined (as always). Counter bypassed.
The rental car brand that won this booking wasn't the one with the best search ranking, the slickest app, or the most compelling TV campaign. It was the one whose data was machine-readable, whose agent could transact reliably, and whose system remembered what this traveler prefers.
Welcome to agentic commerce. And for the rental car industry, the opportunity to lead is wide open.
The distribution landscape is fracturing; a new channel is emerging
Rental car distribution has always been multi-channel. Direct bookings through brand websites and call centers. Global Distribution Systems (GDS) powering corporate travel through TMCs like Amex GBT and BCD Travel. OTAs and metasearch engines driving leisure comparison shopping. Airline ancillary platforms bundling car hire with flight bookings through white-label intermediaries. Aggregators and brokers reselling inventory wholesale.
But a sixth channel is now emerging; one that nobody fully controls.
Consumer AI assistants are becoming the starting point for travel planning and booking. In just two years, generative AI platforms have gone from niche to being as widely used as general search engines for travel research : AI usage for travel research has increased 5x to 33% of travelers, almost matching general search engines at 35% (down from 51%). Major technology platforms have built agentic booking directly into their AI-powered search experiences, naming OTAs and hotel chains as transaction partners. Consumer AI assistants now offer in-chat purchasing with integrated payments. Others explored direct booking before pivoting to route transactions through established travel platforms.
For rental cars, this shift creates a strategic choice. In the airline bundling channel, B2B intermediary platforms already aggregate and normalize rental car supply for over 70 airlines and 250 travel companies across 176 countries. These platforms have invested heavily in structuring vehicle and rate data for machine consumption. When a consumer AI assistant needs to book a car, it queries whatever structured data source it can access. If a brand relies solely on intermediary channels for agent-accessible data, it gains distribution but may miss the opportunity to build a direct relationship with the traveler.
These intermediaries are naturally positioned as the structured-data layer between rental brands and the agentic distribution environment. They play a valuable role in extending reach, but the question for rental car CEOs is whether their brand will also be directly accessible to agents, or whether they rely entirely on third parties for this emerging channel.
Why rental cars are uniquely positioned
Every travel vertical faces the agentic commerce shift and rental cars have structural characteristics that make the opportunity potentially larger than in airlines or hotels.
Unclaimed territory: the post-booking experience. Airlines have already filled their in-journey digital experience with apps for boarding passes, gate changes, rebooking, and in-flight entertainment. Hotels have apps for mobile check-in, room keys, concierge, and checkout. These verticals have less room for agentic AI to add transformative value; they're optimizing an experience that already exists digitally. Rental cars are different. Between the confirmation email and the counter, the digital experience is largely a blank canvas. Some operators have invested in AI-powered contact centers (handling, for example, roadside assistance and reservations), but the industry largely lacks a proactive, autonomous digital experience between booking and return. The brands that fill this space with an agentic layer won't just incrementally improve an existing digital experience; they will create an entirely new one. The opportunity for transformation is greater here than in verticals where digital self-service is already mature.
High-frequency, relationship-driven usage. Corporate business travelers may rent 20-50 times per year across multiple cities, each with different vehicle needs, rate codes, and pickup logistics. This frequency is the ideal foundation for building compounding memory. Airlines see a customer maybe 10-15 times a year; hotels perhaps 10-20 nights. A frequent business renter generates more interaction data, more preference signals, and more opportunities for anticipatory service; each rental is a moment where the agent learns something new about how that traveler prefers to move.
Multi-brand orchestration as advantage. Major rental car brands manage portfolios spanning premium, business, and value brands. Today, this creates coordination complexity across systems and customer data. With an agentic orchestration layer, a multi-brand portfolio becomes a strategic asset: a single agent routing a traveler to the right brand based on context (premium for corporate, value for leisure, carshare for urban) with unified memory spanning all brands. Operators who solve this orchestration challenge gain more leverage from the investment than a single-brand competitor ever could.
Building context that intermediaries can't. Intermediary travel platforms play an important role in distribution: they see the customer's full consideration set, cross-brand shopping behavior, and trip context that no single rental brand can access through its own channels alone. They know which customers compared your brand against competitors, what price differential caused a switch, and what the broader trip looks like. When an AI agent assembles a travel itinerary, the intermediary's aggregated view of the customer can be more complete than any single brand's view of that same customer. This is precisely why building a direct agentic relationship matters. The brands that earn direct permission and build proprietary customer memory through repeated interactions, the nuanced behavioral understanding that only comes from being the chosen provider, interaction after interaction,create context of the entire journey and rental experience that intermediaries cannot replicate.
The protocol landscape is moving fast
The infrastructure for agentic commerce in travel is already being built and deployed.
The Model Context Protocol (MCP) , open-sourced by Anthropic in November 2024 and donated to the Linux Foundation in December 2025, has emerged as the standard for connecting AI agents to external systems. It defines how agents discover what a service can do, query its data, and execute transactions through a standardized interface.
The travel industry has moved fast:
- Sabre launched MCP-based agentic APIs in September 2025: a proprietary MCP server that acts as a "universal translator" making travel technology understandable to any AI agent
- Travelport partnered with Anthropic and Cognizant to build an MCP layer on top of its Trip Services platform, targeting automated exchanges, rebookings, and disruption intelligence
- BCD Travel, one of the world's largest TMCs, deployed MCP across its Tripsource platform in May 2026, enabling AI agents to connect across booking, policy, and data systems through natural language
- The Universal Commerce Protocol (UCP) launched in January 2026: an open standard enabling AI agents to handle the full commerce journey from discovery to checkout, with over 20 companies endorsing at launch
- Amazon launched the Agentic Shopping Assistant (ASA) on AWS in May 2026: built on Amazon Bedrock AgentCore and inspired by learnings from Amazon's own shopping AI (which drove nearly $12 billion in incremental sales), ASA packages the architecture, starter code, and expert guidance for any retailer to build conversational, agentic commerce experiences
Hotels, airlines, and corporate travel platforms are already becoming agent-accessible through MCP. Rental cars are the natural next step in this journey; and the sooner a brand leans in, the more it stands to gain.
The franchise challenge; and why agentic architecture solves it
The conventional response to franchise fragmentation is a multi-year data transformation program: standardize systems, migrate platforms, build integrations. These programs are expensive, slow, and frequently stall.
Agentic architecture offers a fundamentally different approach; one that works with heterogeneity rather than requiring its elimination.
The two-way franchise challenge today:
Corporate headquarters builds digital capabilities (dynamic pricing, mobile check-in, connected vehicle features, loyalty integration) but deploying them across franchise locations with different tech stacks requires custom integration for each system variant. Result: corporate locations get new capabilities in months; franchise locations wait years or never receive them.
In the other direction, corporate needs real-time fleet availability and vehicle data from franchise locations to present a unified brand to the market. But each franchise system speaks a different language. Building point-to-point integrations for every variant is an N×M problem that doesn't scale.
How MCP and agentic architecture change this:
MCP was designed to solve exactly this class of problem: connecting AI systems to heterogeneous data sources through a standardized interface. The same principle applies internally.
Corporate publishes MCP servers for its digital services (pricing engine, loyalty system, fleet management). Any franchise location, regardless of its underlying technology, connects through a standard protocol rather than requiring custom integration per system variant.
Franchise locations expose MCP servers for their local data (fleet availability, vehicle condition, location hours). Amazon Bedrock AgentCore Gateway transforms existing APIs and functions into MCP-compatible tools with zero code changes to the underlying system. A franchise doesn't need to rebuild its rental management system. It needs a thin MCP layer in front of it.
The brand's central AI agent then orchestrates across all locations, corporate and franchise, through this unified protocol. And here's the strategic insight: the same interface that lets the brand's agent manage franchise operations also lets external AI agents book across the full network. One architecture, two problems solved. The franchise network transforms from a liability (fragmented, inconsistent, slow to integrate) into an asset (broader coverage, more locations, all machine-accessible).
An AI agent also handles the messiness that traditional integration cannot. If a franchise system returns data in a slightly different format, the agent interprets it. If a system is slow to respond, the agent works with cached data and flags confidence levels. If a location lacks certain capabilities, the agent gracefully degrades rather than failing. This is fundamentally different from brittle point-to-point integration; it's adaptive by design.
What "agentic-ready" means for a rental car brand
The end-state isn't a single system. It's three capabilities working together:
1. Be findable by agents. Your fleet, rates, policies, and availability are exposed through standardized protocols so that any AI agent (whether it's a consumer AI assistant, a TMC's system, or a traveler's personal agent) can discover and transact with you directly. This is the new discipline of AI brand visibility: how often, and how favourably, your brand appears when AI systems recommend options. In the agentic era, visibility goes further; it's not enough to be mentioned in an answer. Your brand must be selectable and transactable when an agent is ready to act. (See the Brand Visibility Analysis Tool .)
2. Have your own agent. A brand-owned AI agent, built on Amazon Bedrock and Amazon Bedrock AgentCore , that reasons over your data and acts on behalf of the brand. It answers availability queries from external agents, executes bookings, handles modifications, applies loyalty and corporate rates, connects to telematics for vehicle intelligence, and escalates to humans for edge cases. While the foundational AI technology is proven and readily available, the knowledge, business logic, and operational data your agent operates on is the brand's competitive advantage.
3. Earn permission to act. The customer grants your agent authority to act autonomously: extending rentals when flights are delayed, swapping vehicles when plans change, proactively booking when travel is confirmed. Permission transforms the relationship from transactional to anticipatory. And permission, once granted, creates the conditions for the most powerful competitive advantage in the agentic era.
Acquisition and retention in an agentic world
In the agentic era, the most valuable asset a rental car brand owns isn't its fleet. Fleets are increasingly commoditized: similar vehicles, similar locations, similar base rates. The new differentiator is the accumulated understanding of each customer built interaction by interaction, rental by rental, disruption by disruption.
This is the agent memory advantage.
Traditional loyalty programs store transactions: tier status, points balance, rental count. An agentic memory system stores context; the nuanced, behavioral intelligence that compounds over time and becomes impossible for a competitor to replicate.
What a brand's agent memory accumulates:
- Vehicle preferences by trip type: SUV for family holidays, compact for solo business, EV when charging is available at destination
- Pickup behaviors: always arrives 30 minutes after landing, prefers to skip the counter, never takes the insurance upsell
- Disruption responses: wants proactive rebooking, not just a notification; will flex on vehicle class for availability but not on location
- Corporate context: which rate codes apply, which approval workflows matter, which expense policies constrain choices
- Seasonal patterns: always needs a car in Munich the third week of September; books 3 weeks ahead for business, 3 months ahead for family
- Tolerance thresholds: will accept a 15-minute wait but not 30; will pay more for guaranteed vehicle class
These insights go far beyond a traditional customer loyalty database and they emerge only from repeated agentic interactions where the agent observes, learns, and adapts.
Why agent memory is a genuine competitive advantage:
It compounds. The 1st rental teaches the agent a vehicle class preference. The 5th teaches pickup behavior. The 20th teaches disruption tolerance. The 50th means the agent knows the traveler better than any human counter agent ever could. A competitor starting from zero cannot replicate 50 rentals of learned context.
It creates switching costs that aren't about points. Traditional loyalty switching costs are rational: "I have 50,000 points I don't want to lose." Memory switching costs are experiential: "My agent just knows what I need. Starting over means re-explaining everything, getting wrong recommendations, and losing the frictionless experience." Research calls this "behavioral lock-in" : the accumulated learning isn't exportable.
It enables anticipatory service that non-memory systems cannot match. Without memory, an agent is reactive: "Find me a car in Denver." With memory, the agent is proactive: "Your Denver flight is confirmed for Tuesday. I've reserved the same midsize SUV you prefer at that location; loyalty rate applied, GPS pre-loaded with your hotel. Shall I confirm?" The customer didn't ask. The agent anticipated. That's only possible with deep, persistent memory.
It creates a flywheel. Anticipatory service earns more permission. More permission generates more interactions. More interactions build deeper memory. Deeper memory enables better anticipation. The cycle accelerates; and the gap between the incumbent and any challenger widens with every turn.
The operational memory advantage:
Beyond individual customer memory, rental car brands possess institutional intelligence that no external agent can replicate: telematics data from hundreds of thousands of connected vehicles, contact center interactions revealing common pain points, fleet utilization patterns across locations and seasons, vehicle condition histories informing assignment decisions.
A traveler's personal AI assistant might know the traveler's preferences. But only the rental brand's agent knows that the Denver airport location runs low on midsize SUVs on Tuesday evenings, that the specific vehicle assigned has a maintenance flag, or that a better option is available at the off-airport location 10 minutes away. The combination of customer memory and operational memory is what makes the brand's agent genuinely valuable: not just a booking interface, but an intelligent partner that understands both the traveler and the fleet.
Amazon Bedrock AgentCore Memory provides the infrastructure for this: both short-term memory (session context for the immediate interaction) and long-term memory (persistent knowledge that spans sessions, channels, and time). The memory is scoped per customer, accumulates automatically as the agent operates, and uses configurable strategies to extract actionable insights from raw interactions.
This is where the compounding starts.
Agentic commerce also changes how brands acquire new customers. Like comparison sites today, third-party AI agents acting on behalf of travelers will compare options instantly. But unlike a static rate listing, a brand's own agent can actively shape how it's evaluated: communicating total value in machine-readable terms (included mileage, cancellation flexibility, loyalty benefits), offering personalised bundles based on the traveler's context, and providing confidence signals (verified availability, guaranteed vehicle class) that make the third-party agent more likely to select them. The brand that's easiest for agents to evaluate and trust wins the first booking. From there, the memory advantage takes over.
Memory accumulates once you start. A business traveler renting 30 times per year generates 30 interactions of behavioral intelligence annually. After 12 months with one brand's agent, that brand holds context that would take any new entrant a year to build. The traveler might try an alternative offering better rates; but the alternative's agent doesn't know their pickup preferences, doesn't anticipate their Munich pattern, doesn't proactively adjust when flights are delayed. The traveler returns because the experience is better, even when the price isn't.
Every month of agentic commerce means more behavioral intelligence accumulated, more permission earned, and a stronger foundation for anticipatory service.
What to build today
The path forward is not a multi-year data transformation program. It's a series of concrete actions that deliver value immediately while building toward the memory advantage.
Start now: first 90 days
Launch an MCP server for your brand. Even if it initially covers only your top 50 airport locations. Expose availability, rates, vehicle classes, and policies as structured, machine-queryable endpoints. Amazon Bedrock AgentCore Gateway transforms your existing reservation APIs into MCP-compatible tools without touching backend code. Sabre, Travelport, and BCD Travel have all shipped MCP implementations in the last 9 months. The protocol is production-ready and the environment is forming now.
Build a brand sub-agent on Amazon Bedrock AgentCore. Start narrow: answer availability questions, quote rates, execute simple bookings, handle modifications. This agent represents your brand in machine-to-machine interactions. It doesn't need to handle every edge case on day one; it needs to handle the high-volume, straightforward interactions that currently flow through intermediaries.
Measure your AI brand visibility. When consumer AI assistants recommend rental cars, how often is your brand selected? What data source are they pulling from? If the answer is solely an intermediary's structured feed rather than your own direct presence, there's an opportunity to strengthen your position in this channel. This is your AI brand visibility baseline: the metric that complements SEO rankings in the agentic era.
Deploy one agentic use case that fills the post-booking void. Disruption recovery is the obvious candidate. When a flight is delayed, your agent proactively adjusts the reservation, notifies the customer, and confirms the vehicle is still available. No phone call. No hold queue. This demonstrates immediate value, earns permission for future autonomous actions, and starts building the memory that compounds over time.
Build momentum: 6 to 18 months
Extend MCP coverage across your network, including franchise locations. Use AgentCore Gateway to put a thin agentic layer in front of franchise systems without requiring system replacement. Start with the 80% of volume that runs on corporate systems, then progressively onboard franchise locations.
Connect telematics to the agent. Vehicle condition, fuel level, and mileage data from connected vehicles enables automated returns (no inspection queue), proactive service alerts, and verified handoffs. This is operational data your brand uniquely owns; and it makes your agent smarter than any generic AI assistant.
Earn permission from high-frequency travelers. Corporate travelers renting 20-50 times per year are the beachhead. Offer them autonomous management: automatic extensions, proactive swaps, seamless loyalty application. Each permitted interaction compounds the memory advantage.
Deploy the agent across channels. The same brand agent that serves a traveler's personal AI assistant also powers your contact center (agent-assisted calls), your corporate TMC integrations, and your own digital surfaces. One intelligence layer, multiple channels, unified memory.
Win the default position: 18+ months
From reactive to anticipatory. The agent doesn't wait for the customer to ask. It books when flights are confirmed, adjusts when meetings move, recommends based on learned preferences. This is only possible with the deep memory accumulated in earlier phases.
Cross-portfolio orchestration. A single agent managing the full mobility portfolio (premium brand for business, value brand for leisure, carshare for urban) routing based on context. The multi-brand portfolio becomes a strategic asset when a single memory layer spans all brands.
Scale the memory advantage. Every interaction enriches the customer profile. Every disruption handled autonomously deepens trust. Every proactive recommendation that lands correctly earns more permission. The flywheel accelerates; and the gap between your brand and any challenger widens with every turn.
The opportunity is now
The infrastructure for agentic commerce in travel is production-ready. GDS providers have shipped MCP servers. TMCs have deployed agentic frameworks. Consumer AI platforms are integrating travel booking. Open commerce protocols are onboarding travel verticals. Amazon has packaged its own agentic shopping technology for any retailer to deploy. Hotels and airlines are already connected.
Rental cars are the natural next step. The question is not whether this shift will happen; it's how quickly your brand starts building on it.
The memory advantage compounds. Every month of agentic commerce means deeper customer understanding, stronger permission relationships, and richer context that drives anticipatory service. Brands that start sooner accumulate more; brands that wait longer have more ground to cover.
This same compounding dynamic applies across travel and hospitality. Any brand with high-frequency customer interactions, complex operational data, and multi-channel distribution stands to gain: airlines and rail operators with frequent flyers, hotel groups with loyal guests, cruise lines with repeat passengers. For the broader framework on how agentic commerce will reshape discovery, booking, and the customer relationship across all of Travel & Hospitality, read The Agentic Commerce Shift: How Travel & Hospitality Will Win the AI Era .
About the Authors
Bob Kwik is Worldwide Head of Solutions and Go-to-Market for Airports & Ground Transportation at AWS, based in the UK. He works with airport operators, ground transportation brands, and technology partners to accelerate cloud adoption and AI innovation across the travel ecosystem. Bob has been with AWS since 2021.
Dean Smith is a Senior Solutions Architect at AWS, based in Texas. He works with major rental car and ground transportation customers in North America, helping them architect cloud-native solutions and adopt AI and machine learning capabilities.
Vinicius Pedroni is a Senior Solutions Architect at AWS, based in Florida. With over 9 years at AWS, Vinicius works with leading rental car and mobility companies, helping them modernize their technology platforms and adopt generative AI to transform customer experiences and operational efficiency.
To get started, connect with your AWS account team or visit aws.amazon.com/contact-us to speak with AWS Solution Architects about Amazon Bedrock and Amazon Bedrock AgentCore .
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article