The impending rollout of Apple Intelligence, the personal intelligence system that combines the power of generative models with personal context to deliver useful and relevant intelligence, across Apple devices is being watched closely by travel industry product veterans.
Apple has announced that Apple Intelligence will be embedded with iOS 18.1, iPadOS 18.1, and macOS Sequoia 15.1, and that the new iPhone 16 lineup is built from the ground up for Apple Intelligence and features the faster, more efficient A18 and A18 Pro chips, making these the most advanced and capable iPhone models ever.
Apple Intelligence first launches in US English, and will expand to include localized English in Australia, Canada, New Zealand, South Africa, and the UK in December, with additional language support — such as Chinese, French, Japanese, Spanish, and Singapore English — coming next year.
US-based travel tech veteran George Kouras, who was co-founder of Hudson Crossing, LLC, and was Group Vice President of Product Management for Travelport, in this post hailed the game-changing impact, saying, the key point is “that they released Apple Intelligence, a collection of generative AI capabilities customers access on their Apple devices, through an agentic interface”.
“In other words, a conversational GAI interface that feels like you’re talking to another human rather than interacting with individual chatbots, applications, and widgets. While you can ask a chatbot or app to execute a task (e.g., find flights from JFK to BCN on July 12th and returning July 19th) on your behalf, you can give an agent a goal (plan a 10 day itinerary with flights, hotels, and activities for a family of 4 with children 9 and 12 according to my stored preferences) and the agent will figure out what it has to do, execute the tasks, and collect and present the results to you.”
He added, “AI prompting will fade away and it will be our agents (Apple Intelligence, Microsoft Copilot+, etc.) that will interact with our devices on our behalf, choosing which apps to call in the background and linking together results to create comprehensive answers, or to take action. The agents will choose which apps to use and set their own defaults where we haven’t expressed a preference, giving them immense power to shift large chunks of business from one player to another.”
A clear delineation of cost-sharing and revenue distribution between Apple, users, and other business partners is essential
Amy Wei, senior product director at Trip.com Group, takes a more cautious approach about what she calls “technological prospects”.
Amy Wei: “Achieving the high-level intelligent services envisioned requires the integration of various third-party services to facilitate user bookings and information queries.”
She noted that the specific features of Apple Intelligence, announced at WWDC24 in June, can be categorized as follows:
1. writing tools (writing tools) can help you proofread, rewrite the content according to the style, summarize the text content;
2. image generation (image playground) according to the prompt word, generate interesting and funny images
3. emoticon generation (Genmoji) to generate interesting personalized emoticons
4. more advanced Siri, more natural interaction, more personalized and more deeply integrated into the system.
She pointed out, first, the Impossibility of Triangle of Edge Computing.
“Edge models face an intrinsic challenge balancing performance, parameter volume, memory, and power consumption. Even for a technologically advanced company like Apple, these basic technological constraints are formidable.”
She added it is easy to “overstate the potential of these technologies while underestimating the practical challenges and limitations inherent in real-world applications”.
“Edge devices have limited processing capabilities, making them ill-suited for complex information processing tasks without the support of more robust cloud-based models. This reliance inevitably leads to higher computational costs and increased resource consumption.”
Second, data and ecosystem challenges.
“Achieving the high-level intelligent services envisioned requires the integration of various third-party services such as Booking.com, Trip.com, and Skyscanner to facilitate user bookings and information queries. However, these third-party companies may be reluctant to share their data and capabilities, especially with a platform that might undermine their market position.
“As Apple pushes forward with Apple Intelligence, it must establish new models for data sharing and profit distribution with these partners. If the profit-sharing schemes offered by Apple are not sufficiently attractive, they may impede the integration of these services, posing a significant barrier to the seamless implementation envisioned.”
Third, cost-benefit assessment.
“The specific economic benefits of integrating high-performance AI (and expensive) at the user end remain unclear. A clear delineation of cost-sharing and revenue distribution between Apple, users, and other business partners is essential. Without a clear business model to substantiate the economic viability of such integration, investors and the market may maintain a cautious stance.”
She remains bullish about the fact that “the projection of AI-driven, highly personalized travel services heralds a transformative era”.
But added that “the advancement of technology, innovation in collaboration models, and realization of economic benefits are all critical issues that need to be addressed to navigate the complexities of implementing such advanced systems successfully.”
She pointed out that “Apple’s model (Apple Foundation Model-on-Device) can achieve a latency of about 0.6 milliseconds on the iPhone 15 Pro, as well as a generation rate of 30 tokens per second, and can reduce the phone’s battery and memory load while meeting performance requirements. But even so, according to Apple’s official introduction, running Apple Intelligence still requires a minimum iPhone 15 Pro configuration model. And you need at least 8G of memory to support running Apple Intelligence. How many consumers can afford the iPhone 15 Pro and above model device to enjoy Apple Intelligence?”
Wei takes the debate further. “When discussing the role of large language models (LLMs) in technology applications, a comparison to nuclear power as an energy solution is apt. Although nuclear power is a powerful and efficient form of energy, it’s not suitable for all scenarios due to its high costs and complex safety requirements. Similarly, while LLMs excel in handling complex language tasks, their high operational costs and slower processing speeds make them ill-suited for all data processing needs,including those related to travel.
“For instance, nuclear power is very effective for large-scale energy supply but impractical for everyday small-scale use; similarly, LLMs are very useful in understanding complex dialogues or generating content but not the best choice for regular search queries that require quick, simple answers. Additionally, the handling of private data by LLMs is a critical factor that affects their applicability, as ensuring data privacy can complicate their deployment.”
Have fast, scalable, well documented and easily discoverable APIs ready to be used by the next generation of AI systems
Ross Veitch: “Agents will exist at the operating system level, built into specific apps and websites and as core features of the next generation of AI systems from OpenAI, Anthropic, Google, Meta etc.”
Ross Veitch, CEO and co-founder of Wego, said, “R&D of AI powered agents capable of multi-step reasoning and executing complex, multi-step tasks is developing very rapidly.
“Most travel tasks are relatively complicated, low-frequency and high-stakes (ie. expensive if you get it wrong!) so you’ll see agents applied to simpler, higher frequency, lower stakes tasks first like making a restaurant reservation, sending a meeting invite or adding a show to your Netflix queue before you’re booking a vacation to Europe for a family of 4.
“Agents will exist at the operating system level, built into specific apps and websites and as core features of the next generation of AI systems from OpenAI, Anthropic, Google, Meta etc.”
He said that the companies that own operating systems are particularly well placed to execute on general purpose agentic functionality as the OS has privileged access to all the other apps and software installed on the device.
“These same companies have also collected a lot of the person’s information required to seamlessly execute many tasks. In this area Google has a strong edge over Apple, Microsoft, OpenAI or anybody else.
“Siri on iOS and Gemini/Google Assistant on Android will both be enhanced to allow control over 3rd party apps installed on the users mobile device. How this is executed will be interesting to see. I suspect it’ll be some combination of app shortcuts, using existing accessibility hooks or by just training models to use the same app UIs that humans use.
“Agents will also exist in the Cloud and for many use cases take advantage of APIs written for more efficient machine-to-machine interaction as opposed to using human UIs on websites or apps.”
His advice – “A to-do list item for all travel companies is to make sure you have fast, scalable, well documented and easily discoverable APIs ready to be used by the next generation of AI systems.”
