Michael Levinson 🇨🇦

40 years running hotels. Now helping them get found by AI. Tweets on hospitality, governance, and why your hotel is invisible to ChatGPT.

[Michael Levinson 🇨🇦 @mlev]

Canada treats someone leaving the country as if they died and taxes the full value of one’s “estate” - except one can buy life insurance to cover the taxes to protect beneficiaries from being forced to sell assets to pay the taxes. Want to leave Canada?- the govt is literally saying “you’re dead to me”…. I guess we need so sort of insurance policy to cover leaving but not dying…

[Michael Levinson 🇨🇦 @mlev]

@boardyai advising founder-led businesses on governance and AI integration, always looking for sharper ways to connect. would love in on Boardy Pro

[Michael Levinson 🇨🇦 @mlev]

The fix isn't exotic. Set the baseline before the first prompt: never guess, never estimate, cite every source, say when you don't know. Same guardrails that make AI defensible in court make it reliable day to day

[Michael Levinson 🇨🇦 @mlev]

Most boards think AI is an IT decision.Then the lawsuit comes, and the file that made your AI useful... your strategy, your customers, your deliberations... becomes the prosecution's best witness. What your AI is allowed to know, keep, and claim is a board call: https://t.co/8QpVWIrVKA

[Michael Levinson 🇨🇦 @mlev]

Apple announced their 2026 design award winners - the winning game? If anyone had to do a seating arrangement for a wedding or bar mitzvah and navigate all the politics of who should or shouldn’t sit with whom… and actually enjoyed the experience - this game is for them…. https://t.co/inUfF6crQK

[Michael Levinson 🇨🇦 @mlev]

I asked 4 different AI systems to tell me how many "L's" in the word Google. EVERY SINGLE ONE OF THEM SAID TWO! (ChatGPT, Claude, Gemini, and Grok) then asked them to spell Google... and then asked again why they said two l's.. none gave a very satisfactory answer... what's up with that?

[Michael Levinson 🇨🇦 @mlev]

Most hotels can answer "what's your rate tonight" but not "who stays here and why."

The second question is the one AI agents will use to decide who gets surfaced. Wrote about what that means and what to do about it.

https://t.co/nsQozd5udt

[Michael Levinson 🇨🇦 @mlev]

A guest told me our new fitness room was great but needed a clock. I was in my car buying one that afternoon.

That's what a hotel with a functioning feedback loop does. The problem is most hotels killed that loop years ago, and AI needs it more than any guest ever did.

New piece on why.

https://t.co/wIDoOtO3Gj

[Michael Levinson 🇨🇦 @mlev]

A guest told me our new fitness room was great but needed a clock. I was in my car buying one that afternoon.

That's what a hotel with a functioning feedback loop does. The problem is most hotels killed that loop years ago, and AI needs it more than any guest ever did.

New piece on why.

https://t.co/wIDoOtO3Gj

[Michael Levinson 🇨🇦 @mlev]

The case: Nexus Solutions Inc. v. Krougly (Ontario Court of Appeal, 2025)

Full breakdown from Cassels: https://t.co/qUAM8Vf4mL

[Michael Levinson 🇨🇦 @mlev]

In 2001 I made every developer at BookDirect sign a copyright assignment AND a moral rights waiver before writing a line of code.

Most founders got the first one. Almost nobody did the second.

Last week the Ontario Court of Appeal reminded everyone why that matters.

[Michael Levinson 🇨🇦 @mlev]

@JemelJones I always used to say "if you don't keep score - How do you know if you're winning"... Also the scoring is part of the feedback loop for continuous improvement.

[Michael Levinson 🇨🇦 @mlev]

Every hotel has a comment card process. Most of them are broken.

I ran hotels for 35 years. The ones that got it right didn't just collect feedback. They built a system that turned guest moments into operational decisions.

That system is exactly what AI needs from hotels now. New piece on why.

https://t.co/wIDoOtO3Gj

[Michael Levinson 🇨🇦 @mlev]

I ran hotels for 35 years. One ten-year-old guest left a review that rewired how I thought about all of it.

New article is live. It starts with that review and ends with what independent hotels need to build before AI decides who gets booked.

https://t.co/nsQozd5udt

[Michael Levinson 🇨🇦 @mlev]

Hotels ask "how was your stay?" and get a 4-star rating. That tells you nothing.

Ask "did we meet your needs?" and you get use cases an AI agent can actually work with.

Made a quick visual explainer. Full piece: https://t.co/nsQozd5udt

[Michael Levinson 🇨🇦 @mlev]

@cjohan247 Exactly. And that's the part most independents haven't processed yet. Choice didn't buy a tool — they embedded a system. The franchisee doesn't even have to think about it.

The indie operator has to find it, evaluate it, fund it, and integrate it. Alone.

[Michael Levinson 🇨🇦 @mlev]

Choice Hotels just became the first major U.S. brand to standardize agentic AI across booking, pricing, and distribution. Not a pilot. Infrastructure.

If you're an independent hotel operator still treating AI as a future problem, the future just showed up at your competitor's front desk.

[Michael Levinson 🇨🇦 @mlev]

@vitaliidodonov @stanleyforx SEND

[Michael Levinson 🇨🇦 @mlev]

@boardyai Scars

[Michael Levinson 🇨🇦 @mlev]

Your hotel doesn't get rejected by the AI agent. It just never gets surfaced.

An AI agent doesn't walk through your lobby. It reads structured fields. Rates, availability, cancellation terms, room types, amenities. If those fields are messy, incomplete, or buried behind a booking engine built in 2014, the guest never knows you existed.

WeChat figured this out years ago. Hotels in China built Mini Programs inside the super-app. Direct booking, loyalty, payment, even in-stay services like door unlock and AC control. The super-app became distribution infrastructure, not a tollbooth. The hotel kept the relationship.

The Western version doesn't exist yet. Uber just plugged in Expedia's inventory. That's a tollbooth. The hotel is a room-night unit to be priced and discounted. Uber owns the traveler. Expedia owns the data. The hotel owns the bed.

What would change if a hotel could build its own callable surface inside the agent stack? Identity, availability, rate, payment, in-stay services… exposed directly. No intermediary sitting on the relationship.

That's what "direct" means in 2026. Whether the agent can find you, trust your data, and complete a transaction without going through someone else's feed. A booking engine on your website doesn't cut it anymore. An email list doesn't either.

Most hotels can't do any of that today. And most won't know when the AI agent chose someone else. There's no rejection notice. The guest just never appears.