86% of Singaporean travellers are keen on hotels that use artificial intelligence (AI) to personalise their services and offerings

Missed last weeks Newsletter? No worries - here is an audio summary produced with AI! These summaries will now be available live at the same time as the Newsletter is sent. The audio version will be available those who Subscribe on LinkedIn (it’s a different subscription to the one you signed up for to get the written newsletter - sorry - but you can now choose your preference of content delivery)

“Travel is a trillion-dollar industry stuck in the past.”

These were the opening lines of an Andreessen Horowitz (a16z) report called “Adventure is Out There” published earlier this year. This is another of the great early pieces to dig into if you are new to the subject of AI and its conext in travel.

The analysts authoring the report, Bryan Kim and Justine Moore, take us through some of the opportunities that they saw from their 10,000 ft view of the industry. They go through the phases of travel from inspiration to planning, booking and travelling and outline some of the early efforts already emerging in each of those areas.

The one prediction they made that did come 100% true was “In the coming months, we expect to see a wave of startups leveraging AI to make the travel experience more personalized, efficient, and enjoyable”. Boom - nailed it. That’s some crystal ball they’ve got there at a16z.

Show me the money! Travelshift raises $10M for their AI powered travel booking platform

What we have seen very little of thus far is investment from institutional investors in any of the newly minted startups looking to build on top of the latest generative AI innovations. I suspect the word “moat” is being used a lot in the rejections.

Iceland’s Travelshift however have bucked that trend by raising a fresh $10 million from their existing investor General Electric. 

Travelshift has actually been around since 2012 but the raise comes on the back of hiring a new CEO, Dave Stewart, who was previously “Global Head of Private Equity at State Street Global Advisors, the world’s fourth-largest asset management company” & who already sat on the board of the company.

Digging in to Travelshift we couldn’t help notice the reference to a supplier extranet which in an API & now AI driven world seemed a bit out of place. And as someone who worked a decade on the supply side, it gave me chills. A mix of old and new it seems.

Although not officially part of the trial of 52 Travel Planning tools in 52 weeks (see below for more), I did take it for a spin on the B2C facing https://guidetoeurope.com site. I asked the AI for a 5 day itinerary for Amsterdam and a 5 day itinerary for Manchester as this is a trip I’m doing later in the year. It came back with a 10 day Amsterdam and 8 day Manchester. When it pointed out the error it relented and gave me 5 days for each. 😃 Nice upsell attempt! The itineraries are probably what you’d expect but the “Experiences” cards failed to load. It did seem to dynamically price the whole itinerary which was impressive.

86% of Singaporean travellers are keen on hotels that use artificial intelligence (AI) to personalise their services and offerings

In an article penned this week by Traveloka President, Caesar Indra, he talked about the building customer expectations around AI and service levels generally.

Other stats that jumped out were “At least 32% of travellers in the US are likely to use ChatGPT to plan their trips.” Caesar says “Travel planning is its next big remit.” however he goes on to say it isn’t going to be easy because training a LLM “costs more than US$4 million (RM13.7 million)”.

Who is working with whom? It might matter a lot!

Most clicked in week 1 was the story about Sabre using AI to improve attachment rates in its GDS software. What really caught our eye however was that this was being done as part of their partnership with Google. This is part of a 10 year tie up announced back in 2020 between the two organisations with a mission “to help digitally transform the traveler’s experience, and co-create the future of travel.” eDreams & ODIGEO are also on the Google train

We have seen Expedia & Kayak both invited to be initial launch partners with OpenAI’s plugins. Plugins were an initial solution to help connect ChatGPT, which was trained only on data from prior to September 2021, to live data available in the present. (Chritan Watts covered this subject brilliantly in a video post at that time). I haven’t really heard anyone mention plugins for a few months now?

Widely reported last month was that Lonely Planet and Ryanair would be launch partners in Amazon’s new Generative AI Innovation Centre (that’s not a centre at all, but a program). Digging into that little deeper, Amazon explains “to fine-tune generative AI models, businesses need quality datasets, and they need to understand the business potential that generative AI can unlock from their data.”

Skift reported at the time that what that means for Lonely Planet is more personalised Trip Planning according to a statement from Chris Whyde, senior vice president of engineering and data science at Lonely Planet.

With all the major tech players plus OpenAI all competing for a winner takes most prize, who you partner with could be crucial.

Open AI releases Enterprise version to allow companies to fine tune whilst expert testimony to congress says this isn’t possible!

This week Open AI announced their biggest and best yet, Enterprise version of ChatGPT. According to CNBC:

“One key differentiator between ChatGPT Enterprise and the consumer-facing version: ChatGPT Enterprise will allow clients to input company data to train and customize ChatGPT for their own industries and use cases, although some of those features aren't yet available in Monday's debut.”

This is completely contrary to the information about the capabilities of generative AI LLM’s in this post by technology consultant, Dr Jeffery Funk. Funk analyses a story from the Financial Times which covered testimony to the US Congress by Gary Marcus.

Marcus, co-founder of the Center for the Advancement of Trustworthy AI, says that “hullucinations” (a situation where LLM’s just make things up when there is a dearth of other data available and then says them so confidently, the user feels they are statements of truth) are “a feature, not a bug” and “there is a fantasy that if you add more data it will work. But you cannot succeed in crushing the problem with data.”

Which is exactly what the likes of nearly every company with data, is trying to do in the AI space. It should be noted that the post says Marcus “has been a longtime sceptic of AI, neural networks.”

The answer is probably somewhere in the middle. Given the sparse detail given by Open AI about the enterprise version, I’m thinking what they mean is the ability to upload and interrogate your own documents as an employee but keep the contents of those documents safe from them being used to train the base LLM model. It isn’t going to be an all knowing oracle based on all the information a business has in its data where you can just type any question and get the correct and incisive answer. Training LLM’s is hard.

Last week’s poll

The winning answer from last week’s poll was that 68% thought Tour Guiding was safe. It should be mentioned that my Linked In is highly indexed with people from the Tour Guiding world which may be reflected in the results. In that context I hope for their sake, that this isn’t a massive blind spot!

If I were wanting to look on the opportunity side, I would be having discussions around what would our tours look like if we didn’t have to spend a good portion of them regurgitating historical text? If that was removed, what (value add) would include instead?

Got an idea you want to share? Hit reply to this NL and let me know.

Enjoying the content so far and know someone else who might benefit from it? Forward it on to them and they can click this button below to join.

Our first test of the multitude of AI Travel Planning solutions out there. We’ll test 1 a week for the next 52 weeks! This week it is….. journeai.com

I started our 52 week exploration with journeai.com by putting in my standard request which is for around a two week holiday in Borneo. I travelled to Borneo in July with my partner and daughter so we know the lay of the land pretty well having done our own research prior to that trip. I also like to use Borneo because it isn’t a country (there are 3 countries sharing the island) and that gives an added layer of complexity.

So how did Journeai do? It was adequate would probably be the best description. It was pretty clear how to get started and it broke the trip down into chunks that you revealed by scrolling through the days. It kept the entire itinerary in Sabah (many others haven’t) and from a geography POV it flowed OK. The one glaring error was the suggestion to fly from Kinabatangan to Sandakan. It’s about a 2 hour drive and no airport I know of around there. It never mentioned how we’d get to/from any of the other places, one of which would’ve been a 7-8 hour drive. It did leave enough time however to do the things it suggested even taking into account the travel times. Not sure if it just got lucky on that or is tuned that way.

There was no way to book anything except flights which just kicked you out to Skyscanner, presumably on an affiliate link. There was a map that had some but not all spots marked on it. If just looking for some ideas it was fine but GPT would probably just give you the same ideas. I didn’t see a lot else going on.

What would BourdAIn do?

I think this is a question many of us ask ourselves when travelling. In one of the best uses of AI I’ve seen in marketing to date, this is exactly what creator Michael Motamedi asks, as he travels around the world and gets the AI version of Tony to tell him what he would do next. He then goes and does it, whatever it is.

Motamedi is using the Guide Geek app build by Matador as his tool and then filming the results and showing them to his audience on TikTok as well as via the Matador and Guide Geek Instagram accounts.

You can hear Michael talking all about it on the Get Lost podcast here including what he thinks Tony might make of it all!

Tip! The best way to embed knowledge is to talk about what you’ve learned with someone else. Share this email with a curious friend with whom you can brainstorm the ideas.

Most clicked last week was my bio 😅 Hate the game, not the player people! But seriously - Hi 👋- feel free to actually connect and shoot me a message.

That’s it - you’ve made it to the end of this edition. I’ll be putting the result of the most clicked post in next week’s edition so you can see where others are focussing. If I’ve missed something, you’ve got a tip or any feedback at all - you can simply reply to this email and it will come straight to me. I’m doing this for You so please don’t be shy to tell me what you think