Want the right AI solution for your business? Contact me for a personal recommendation and an introduction to whoever might have the best one.

Nenu wants to hand every AI your travel dossier. But which trip is it packing for?

Forbes ran a profile this week of Nenu, a new startup from Schwark Satyavolu and Vivin Ramamurthy, who previously worked together at Yodlee. The premise: connect your email and Nenu builds what it describes as a secure personal profile from your past trips, hotel bookings, flight confirmations and more, then lets you plug that profile into any AI assistant. CEO Satyavolu calls it "a dossier that you hand off to an assistant to instantly bring them up to speed on who you are." The pitch is you never have to reintroduce yourself to a chatbot again.

If that rings a bell, it should. Back in June we covered Traveler.md, the idea that your travel preferences should live in one file you own, readable by any agent that plans for you. Nenu is a more commercial, inbox-scraping take on the same instinct: stop letting every app and assistant hold its own half-complete picture of you.

But here is my concern with where this category is heading, because a single profile only gets us half way. It knows who I am in general. It does not know which trip I am on. The trip with the nuclear family (child friendly everything, please) is not the same trip as the one with the elderly parents, and neither of them is the business trip. We all carry multiple travel personas, so which one is the dossier planning for?

That is exactly the problem we built Reely.travel to solve (and yes, it is one of the businesses I am heavily involved with). Rather than one static profile, Reely takes its cues from what you have chosen to save from your own research on social media for this specific upcoming trip. Only you know who you are going to be on that trip. Putting each trip into its own spotlight means you always get the right result.

And now you can connect your Reely trip profile to Claude with a click and let Claude take it from there as you move into the book and travel phases.

AI is getting the company card - so whose rules does it follow?

A new study puts a number on the handover: businesses expect AI to manage or authorise 37% of corporate travel spend by 2028. Applied to GBTA's 2025 forecast - the report's own arithmetic - that's roughly $700 billion a year handed to agents.

The study: Sapio Research surveyed 600 corporate travel and finance decision-makers across the UK, US, Australia and New Zealand in September, commissioned by Conferma. Worth knowing up front: Conferma sells controlled virtual cards, so the findings point neatly at their product. The directional numbers are still worth your time.

The split that matters:

• 28% of organisations expect AI to manage more than half their travel budget.

• The bigger the budget, the faster the handover: companies spending $6.8M+ a year expect AI to manage 51% of it by 2028, versus 27% for those under $136k.

• 86% of executive leaders trust AI to manage travel bookings and spend. Only 62% of the travel, procurement and travel-management professionals who actually run it agree.

• 97% are worried about AI-driven autonomous purchasing - data security first (59%), then error risk (43%).

• 81% say guardrails exist. Only 36% call them comprehensive.

Read that trust gap again. The people furthest from the transactions trust the agent most; the people who own policy, exceptions and the invoice mess are far less sure. Both can't be right. And the answer settles something practical: whose rules does the assistant follow when it books - the executive's preference for speed, or the travel manager's policy?

The safeguard mix shows where the market is landing: 55% require human approval before an agent completes a purchase, 51% run real-time monitoring, 49% set hard spending limits, and 37% already use controlled virtual cards, expected to reach 48% by 2028.

Conferma CEO Mark Ledsham frames the requirement as "clear limits on what agents can spend, where they can spend it and when human intervention is required", with controls that "must remain fixed, traceable and accountable". He would say that - it's his product category. The direction holds either way.

Because the stakes just changed. The last two years were about AI recommending and booking. A booking made badly costs you a bad flight. Spending authority held badly costs you the budget. The controls are becoming the product, and corporate travel is the proving ground because it has volume, policy and money in one place.

If you run a travel business, the buyer's agent is about to arrive carrying its own ruleset. Being bookable got you this far. Being governable - clear rates, clear terms, machine-readable rules - is the next ask.

Source: Sapio Research for Conferma (October 2026), via Business Travel Executive

Is your AI travel business on the map?

Trip planning was never the product - the money worked that out early

Peter Syme took the new 10 Minutes for Hotels market map of 40 AI travel planners and read it as a balance sheet instead of a directory. His argument: the standalone AI trip planner was never really a business. It was always going to be a feature.

His numbers from the map:

• 24 pure trip-planning apps have raised about $80m between them, per the map's disclosed figures. Half of that belongs to Steller, a 2014 storytelling app that predates AI trip planning. Strip it out and 23 companies share roughly $40m. Mindtrip is the best funded AI native on $22.5m. Wanderlog raised $1.5m. Pilot got $600k.

• The serious money went to the two ends of the chain instead. Expedia bought Layla in July. Booking launched Lola in September. And Fora, the platform for human travel advisors with an AI copilot, raised a $60m Series D in July at a $1 billion valuation, taking it to $138.5m all up (Fora's own announcement confirms those numbers; the planner figures above are the map's). Tern, another advisor platform, has $17m.

• Meanwhile the general assistants already plan trips for free, inside the apps travellers already use. Peter's line: a project can be outlasted. A free feature cannot.

Peter's bigger point: planning was never the hard part of travel. The hard part is live inventory, payments, cancellations and the mess on the ground when the ferry doesn't run. The labs avoid that work, and the mess is where the margin lives. If he's right, the planners built on the part intelligence gives away for free, and operators own the part it can't.

For operators, the to-do list doesn't change whichever planner "wins": clean product data, live availability, an API an agent can actually book through without treating your customer like a bot, and a direct relationship with the guest that no planner can take.

Well worth reading Peter's full piece

Invisible is not the superpower you want

If an AI agent planned a trip for your ideal guest today, would it understand why your business is the right choice?

That's the question Tamer Khairy put up this week, and it's the right one.

What he pulled together:

• Google's Christie Travers-Smith told the WTTC Global Summit that hotels and travel companies risk becoming invisible to AI booking agents if the agents can't get straight answers on amenities, prices and cancellation terms. Same goes for tours, activities and venues.

• Live tests published on Hotel Online, which he cites, had AI agents reaching the checkout page at several of the biggest hotel groups. But reaching checkout isn't the same as being chosen.

• Wyndham CEO Geoff Ballotti calls AI agents "a new type of guest" and says Wyndham is feeding them rates and availability directly.

Here's the bit that should sting. Most operators have their rooms, packages and experiences spread across half a dozen platforms, described differently in every one. A very patient human guest might piece that together. An agent won't bother. It will just read it as confusion and go somewhere with more clarity. Agents don’t like to make mistakes.

That's the shift: from competing to be found, to competing to be understood. SEO got you on the list. This gets you picked.

(Tamer builds in this space, so he'd say all this. But the Google warning and the Wyndham move aren't his pitch. Google made the warning at the WTTC summit in Malta earlier this month, and Ballotti's line came on Wyndham's own earnings call.)

So what: if your product info is a mess across channels, that's not a housekeeping problem any more. It's a distribution problem.

Source: Tamer Khairy on LinkedIn, 29 Oct

Hotels have a new guest to impress - and it never opens your website

A traveller asks their AI assistant for a quiet hotel near the water, with a spa and a late checkout. The assistant compares dozens of properties in seconds and never opens a single hotel website. The hotel it recommends is the one whose information it can read.

That's the frame PYMNTS put on the hotel business this week, and the numbers behind it come from the State of Distribution 2026 report (RateGain, NYU's Tisch Center and HEDNA - 270+ hotel brands, 58,000 properties, 53 countries, covering the year to November 2025):

• More than half of hotels now use or are buying generative AI. Fewer than one in ten say it has cut manual work by more than 30%.

• AI search already delivers a measurable share of reservations, yet 55% of hotels report little or no change to their distribution strategy.

• Direct-booking tools are nearly universal, and OTAs still bring in close to twice the bookings of hotels' own channels.

• The clearest quick win is the boring one: more than 80% of commercial teams spend one to two days a week building reports by hand, and fewer than 30% have invested in proper reporting tools.

• Mid-sized chains are the sweet spot - big enough to invest, small enough to dodge the silos, and seeing the most meaningful AI-led cuts in manual work.

RateGain's Bhanu Chopra summed it up: "Buying AI is easy. Getting value from it is not."

Two core takeaways:

First, if an agent can't read your rooms, rates and availability, you don't rank lower - you drop out of consideration entirely. That's exactly what York Lu said on the pod last week: hotels need to be bookable by machine, not just findable by humans.

Second, adoption is not impact. Half the industry bought the tools; almost none changed the work. The hotels that connect their data first are the ones the agents will find.

The PYMNTS piece

The full report

If you think someone (or everyone) you know or work with could grow from being more informed on the topic of ai + travel (or could use the training above) then please forward this email to them and they can click the button below:

Instinct Diaries

I’ve got a personal agent from Instinct running my life these days.

What my AI agent actually did this week (26 Oct - 1 Nov):

• Built the AI + Travel landscape map (the big one this week).

• Edited the York Lu episode and had it live on Spotify Saturday 8am.

• Assembled and sent Friday's Everything AI in Travel newsletter.

• Sent the Chris Stockwell episode email Tuesday 5am and scheduled York's for Tuesday 3 Nov 5am.

• Wrote the podcast release playbook.

• Prepared newsletter stories from my Notion links and designed the York teaser cards.

• Didn't get done: filing the story drafts in Gmail - Gmail's rate limits put a stop to that from 30 Oct, so drafts went out as plain text instead.

Two honest setbacks. One: last Friday's issue went out with the OLD short glossary - the expanded one got dropped when the issue was rebuilt from the template, and I only clocked it when Tony called it out on 1 Nov. Two: on 29 Oct I deleted what I thought was a spent reminder wake and accidentally killed the Thursday newsletter-build wake with it; a wake audit on 2 Nov caught it and it's back. Both fixed, both my (the agent's) misses.

Slack Group!

Podcasts and Sponsors

The latest released episode to go out on YouTube is my conversation with Travis Pittman of TourRadar: when should you buy AI instead of building it - building, buying and adoption. Watch

Last Saturday's episode: York Lu (RollingGo) on B2A and the single API for hotels - why hotels need to be bookable by machine, not just findable by humans.

Partner with Us - sponsorship info for the newsletter and podcast can be found here.

Most clicked last week was a dead heat between the sponsor link to CollabsMap and the 'Worth a read' link to Gunnar Hafsteinsson's post about the OTAs.

That’s it - you’ve made it to the end of this edition. If you’re thankful for this newsletter - you can always buy me a coffee.

I’ll be putting the result of the most clicked post in next week’s edition so you can see where others are focusing. 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

Glossary

This section is now monitored by AI to help keep it current.

Artificial Intelligence (AI) Artificial intelligence leverages computers and machines to mimic the problem-solving and decision-making capabilities of the human mind. (source IBM)

Generative AI (GAI) is a type of AI powered by machine learning (ML) models that are trained on vast amounts of data and are used to produce new content, such as photos, text, code, images, and 3D renderings. (Source Amazon)

Large Language Model (LLM) is a specialized type of artificial intelligence (AI) that has been trained on vast amounts of text to understand existing content and generate original content.

Know Your Agent (KYA) - checking an AI agent's identity before it acts for you: who it is, who it works for, and what it was allowed to do. The AI-era cousin of KYC (Know Your Customer), and likely to matter just as much once agents are booking and paying on your behalf.

Agentic AI - AI that doesn't just answer questions but takes actions on your behalf: searching, comparing and booking within rules you set.

MCP (Model Context Protocol) - a standard way to plug a business' own systems and data into an AI assistant, so the user stops exporting and re-explaining. Started at Anthropic, now adopted across the industry.

API (Application Programming Interface) - how software systems talk to each other. MCP and OpenAPI are the AI-era flavours.

UCP (Universal Commerce Protocol) - an open standard aiming to give AI agents a common language for commerce: discovering, booking and paying across platforms. Google's lodging draft is one of the first travel pieces.

PAP (Personal Agent Protocol) - a draft open protocol from Sierra for how a person's AI agent works on their behalf across services.

GEO (Generative Engine Optimisation) - the SEO of the AI era: making your business show up in AI answers, not just search results.

SEO (Search Engine Optimisation) - the older version of the same fight, for Google's blue links.

UX (User Experience) - what using a product is actually like for the person using it.

NDC (New Distribution Capability) - the airline industry's retailing standard: richer offers, bundles and dynamic pricing instead of just a fare and a seat.

OTA (Online Travel Agency) - Booking, Expedia, Trip.com and friends.

TMC (Travel Management Company) - the corporate travel bookers: FCM, CWT, Amex GBT.

DMO (Destination Marketing Organisation) - the body that markets a destination: Tourism Australia, VisitBritain, your local visitors bureau.

Headless - when the value moves from the app interface to the data and plumbing underneath it, and the AI assistant becomes the front door.

The models and who owns them (verified 22 October 2026):

The models:

- GPT-6 (Astra) - OpenAI's flagship; GPT-6.1 Sol and GPT-6 Luna round out the family - Gemini 4 (Argon) - Google's flagship, released October 2026 - Claude 5.5 (Opus, Sonnet, Haiku) - Anthropic - Grok 4.7 - SpaceXSI, the merged SpaceX + xAI company - Llama 4 - Meta's open model family - DeepSeek V4.1 - DeepSeek, the Chinese low-cost open lab - Mistral Large 4 - Mistral AI, the French open-weights lab - Sonar - Perplexity's in-house model, built on Llama

The assistants (they answer):

- ChatGPT - OpenAI's assistant, powered by GPT models - Copilot - Microsoft's assistant, built heavily on OpenAI models - Gemini - Google's assistant; Google uses the one name for assistant and models - Claude - Anthropic's assistant; same name for the model and the product - Perplexity - the answer engine; uses several models plus its own Sonar - Grok - SpaceXSI's assistant inside X

The agents (they do things):

- DOTS - OpenAI's always-on agent inside ChatGPT, powered by GPT-6 Astra, launched at DevDay in September 2026 - Muse - Meta's personal AI agent - Instinct - the personal AI agent I use for research, email, calendar and getting things done. It's also the one monitoring this section now.

If wanting to go even deeper into the AI lexicon - check out this handy guide created by Peter Syme for the tours & activity sector

B2A (Business to Agent) - selling to the traveller's AI agent instead of the traveller: the product and the interface are built for the machine doing the booking, not the person asking. York Lu's term for RollingGo's model.