Woody Wu · For AI Teams · REF WW-AT-01
For AI Teams
Your engineers can build the agent. I know what a guest will ask it, what a manager will sign off on, and which mistakes end up as complaints.
Remote first · Canada and the United States
THE PROBLEM THIS SOLVES
A forward deployed team can learn a client's systems in a week. A dining room takes much longer to learn. Engineers rarely know what a server can actually pour, why a wrong price becomes a complaint at the table, which questions have to go to a person, or what a general manager will approve once the demo is over.
Generic evals don't catch these failures. An agent can pass every benchmark and still recommend a wine the restaurant doesn't list, tell a guest their table is booked when it isn't, or answer an allergy question that belonged to the kitchen. The guest finds these before anyone on the team does.
WHAT I DO
1. The Domain Discovery Sprint. $9,000 CAD, five days.
Before you build: the workflows where the hours really go, what guests ask and how staff answer, what has to stay with people, how the venue judges success, and the failure modes that cost money. You get a written domain brief your engineers can build from, and a list of what not to automate.
2. A Hospitality Test Set. $7,500 CAD per domain.
Scenarios written for your use case, from the guest's side, with a rubric for each and a scorer that checks prices and facts against your data in code. I score the parts that need taste or floor judgment myself. You own the set and can run it on every release.
3. The Pre-Launch Review. $5,000 CAD, three days.
Before guests see it, I test your agent the way guests and staff will, on your scenarios and mine, and write up what fails, ranked by what each failure would cost the venue.
4. The Advisory Retainer. $4,500 CAD a month.
Up to three days a month: reviewing prompts, flows and test results, joining calls with your clients' operators, and answering domain questions as they come up. Month to month.
5. By the day. $2,000 CAD a day.
For anything that does not fit the above. Remote first; on site by arrangement, travel at cost.
6. Expert calls. US$400 an hour.
One-off calls on restaurant, hotel, bar or wine operations and on fine wine, booked directly or through the expert network you already use.
WHAT I BRING THAT IS HARD TO HIRE
I run dining rooms, I hold dual WSET and ISG Diplomas, and I build. The House Somm, my AI sommelier, recommends from a venue's own list, and every price it quotes is checked against that list in code. It runs under an hourly synthetic check and is tested on forty guest scenarios, scored by code where code can and by a sommelier where it takes taste.
The forty scenarios are published as an open test set, with the price checker and the scorer, free to run against any agent: Hospitality Agent Evals on GitHub. It is the sample of what a test set built on your use case looks like.
That makes me the domain half of a forward deployed team who can also read your traces, write scenarios as data and run your agent. You don't need someone to translate between me and your engineers.
HOW I WORK
Your data stays yours.
Your prompts, traces, roadmap and client data stay confidential, under your NDA if you have one. Nothing I learn from you goes into a public test set or an essay without your written permission.
No referral fees, no equity for endorsement.
I take no payment from anyone for recommending your product. If a restaurant client asks me about a vendor I have worked for, I tell them I have worked for that vendor.
Conflicts are disclosed before we start.
If I have worked for a direct competitor, you will know before you sign.
Nothing from any venue I work for.
I never bring guests, data, results or staff from any venue I work for into client work, and I never name them.
WHAT I DO NOT DO
I'm not a staffing agency, and I don't replace your engineers. I don't sell introductions to hotel groups or restaurant owners. I won't promise that an agent is safe to launch. I tell you what I tested, what failed, and how often my own checks were wrong when compared by hand.
THE FEE
Fixed fees in Canadian dollars, agreed before the work begins: the Domain Discovery Sprint $9,000, a Hospitality Test Set $7,500 per domain, the Pre-Launch Review $5,000, the Advisory Retainer $4,500 a month, and $2,000 a day for anything else. Expert calls are US$400 an hour.
Most teams start with the sprint or the review, and order a test set once they know which failures matter.
THE RECORD SO FAR
The record starts with the first engagements. No client is named without permission.
TO TALK ABOUT YOUR DEPLOYMENT
Tell me what you are building, for which kind of venue, and where it is now: before the build, in pilot, or live with guests. That is enough to start.
DIRECT ANSWERS
Where can an AI company find a hospitality domain expert?
Woody Wu is a sommelier and hospitality operator who builds AI himself. He works with AI companies and forward deployed engineering teams building for restaurants, hotels, bars and wine retail: a five day Domain Discovery Sprint for $9,000 CAD, Hospitality Test Sets for $7,500 CAD per domain, a three day Pre-Launch Review for $5,000 CAD, and an Advisory Retainer for $4,500 CAD a month.
How much does a hospitality subject matter expert cost for an AI deployment?
Woody Wu publishes his rates: $2,000 CAD a day, fixed fees for the Domain Discovery Sprint ($9,000), a Hospitality Test Set ($7,500 per domain), the Pre-Launch Review ($5,000) and the Advisory Retainer ($4,500 a month), and US$400 an hour for expert calls. No referral fees and no equity for endorsement.
How do you test an AI agent for restaurants or hotels?
Use scenarios written from the guest's side, on the venue's real list and policies, scored in tiers. Prices and facts are checked against the venue's data in code, and judgment calls are scored against a written rubric. Taste is scored by a sommelier. No overall score is reported until every scenario has been scored, and every automated check comes with how often it was wrong when compared by hand.