Quick note – Mindful this is quite tech focused. I’ve added some definitions at the end. Really keen to hear your feedback. Does my argument make sense? Do you agree? Let me know in the comments!
The Trainline recently launched an AI assistant designed to help customers book UK train tickets and generally interact with their service.
It sounds great on paper. An AI-powered customer interface to guide users through journey planning, ticket purchase, refunds etc.
But to me, it reveals something much deeper: how many digital service companies are still missing the true existential threat of the next paradigm in AI.
The Trainline’s model assumes that in the future, customers will go to their AI assistant to get train travel sorted. But that presumes customers will go to Trainline at all. The real shift happening is not that services are adding AI, but that AI is becoming the service.
Specifically, one of the many impressive general-purpose AI agents like
OpenAI’s ChatGPT
Google’s Gemini
Microsoft’s Copilot
Anthropic’s Claude 3
Meta (Facebook)’s Meta AI
xAI’s Grok
Imagine the future scenario – you’re looking at your calendar scheduling the week ahead. You ask your AI agent to sort out your travel options. Your AI already has access to your calendar, preferences, location, constraints, as well as the required open bus / taxi /railway timetables, fare data, weather forecasts, any delays or planned engineering works etc. Within seconds it is able to tell you the different options, talking to you in natural language. When you confirm your route, it buys your tickets for you and saves them directly to your mobile phone wallet. Your AI agent won’t need to involve The Trainline at all. The assistant is the travel agent. The Trainline becomes just one more source in the background, if it’s used at all.
This is the structural disruption AI brings. Not because AI is “clever” in the sci-fi sense, but because it rewires the value chain. Services built on wrapping user interfaces around open data are especially exposed. Their “value-add” layer is replaceable by agents working on the user’s side.
Disruption isn’t new
We’ve already seen similar shifts. For years, OpenAI said it wasn’t in the business of internet search. Until it was.
ChatGPT now returns real time internet search responses. But it does more than search, it contextualises, summarises, remembers your history, and returns actual answers in your preferred tone. It doesn’t point you to a site; it delivers the insight, tailored to you.
That’s a fundamentally different experience to classic search engines, and a real threat to business models that rely on capturing user attention – both in terms of search AND in terms of the websites being searched.
This kind of disruption is going to impact every aspect of our lives. I’m not convinced that Artificial General Intelligence (AGI) will happen. I’m more cynical in that AGI feels to me more like hype to convince venture capitalists to fund AI companies. However, I don’t think AGI is required to disrupt whole value chains. This isn’t limited to travel agents – The Trainline just provides an obvious example (in an industry I’m familiar with).
And to be clear – there are some bigs positives. The journey booking process will deliver better outcomes for users and more quickly. Strategies like dynamic pricing and congestion avoidance will be more effective with AI-agents making recommendations on all available information. Ultimately, travel agents extract value by facilitating trades. AI agents are by nature more efficient.
One other thing to reflect on about The Trainline. They currently put effort (and money) into marketing through SEO and search advertising to insert themselves into your journey planning. Any google search for “train from X to Y” in the UK will return The Trainline’s website as the first (promoted) result. But in a world of personal AI agents that retrieve, compare, and transact on your behalf, there’s no “search results page” to advertise on.
Which is very revealing: AI tools like ChatGPT don’t yet include advertising, but if they start to dominate these kinds of user journeys, it is unlikely that they will “leave money on the table”. They would be perfect for embedded advertising, just as Google did with search, and just like meta do with Facebook. I know it’s a bit reductive, but if the service is free, then you are the product.
If there is a market, there will be value from marketing. It will be interesting to see if the future will include The Trainline selling tickets through native AI-agents, or if their business model will indeed be eaten by those AI-Agents…
My main take aways are:
Any client organisation being offered an AI service should reflect on their own value chain, and whether the service offered makes sense in a future with general AI=agents.
Any business whose value depends on owning the user journey, rather than owning the data, infrastructure, or product, should be rethinking fast. AI isn’t just an assistant. It’s a new interface to the internet.
In this current AI growth phases, AI agents work for the user as they try to capture market share. We are probably living in a golden age of AI!
Once the pace of disruption plateaus I would expert service to be compromised by commercial pressures. Like with all markets - once the new paradigm lands and the competition has been eaten, expect services to get polluted by monetisation, probably through embedded adverts, and probably this will feel more like product placement than a flashing sidebar advert.
Definitions
The Trainline
Trainline is a UK-based digital rail and coach booking platform that aggregates ticketing information from over 270 rail and coach operators across 45 countries. While it offers a user-friendly interface and features like split ticketing, its data is primarily drawn from public rail schedules and open datasets.
Trainline’s AI Assistant Launch (Press Coverage)
Trainline has introduced what it describes as the UK’s first AI assistant for rail travel. The system can answer queries on ticket conditions, real-time information (like platform numbers and delays), and even guide users through the booking process. It integrates conversational AI technology to help with journey planning and FAQs in a chat‑style interface
AI Agent
An AI agent is a software system that acts on behalf of the user, using context, memory, and goals to perform tasks. Unlike traditional software, it doesn’t just respond to inputs—it proactively helps users achieve outcomes, often drawing from multiple datasets and interfaces behind the scenes.
Open Dataset / Open Rail Data
UK rail data—including timetables, ticket prices, and real-time train running info—is made publicly available through the National Rail Enquiries API and other sources. This enables third parties to build travel apps and services without owning the core infrastructure.
Disintermediation
Disintermediation is the process of cutting out middle layers—like brokers, platforms, or booking sites—when a user or technology can go directly to the source.
Search vs. AI Retrieval
Traditional search engines return links and snippets from web pages. AI retrieval systems like ChatGPT’s browsing tool synthesise multiple sources and generate an actual answer—tailored to the user—without needing to visit external sites.
ChatGPT Search Feature and Google Comparison
In December 2024, OpenAI rolled out its web browsing feature—previously limited to Pro users—to all ChatGPT users. This marked a major step towards turning ChatGPT into a full search engine alternative. Unlike traditional search engines that return links, ChatGPT’s browsing tool delivers synthesized, conversational responses, positioning it as a direct competitor to Google’s search dominance.
SEO
Search Engine Optimization (SEO) is the practice of improving web content to appear higher in search results, increasing visibility and traffic.
How Google’s ad model works:
Google makes most of its revenue from advertising, especially through “sponsored” search results and paid rankings on pages like Google Maps and Google Shopping.
Thanks for reading
If you enjoyed this article- the best compliment you could give me would be to share it with someone else who you think would enjoy it.
And if you liked or disliked this article, please let me know why in the comments below.
If you’d like to know when I’ve posted you can sign up to my blog here.