AI drafts. You approve.
- A person checks every output. Each prompt marks its result DRAFT - PENDING OWNER REVIEW (or STAFF REVIEW). Read it against your own facts before it goes anywhere. An AI checking its own work is not verification. Recommendation
- Use approved facts only. Your own fact sheet, prices and policies. Where a fact is missing, the prompts tell the AI to leave a marked gap, such as [FILL], instead of guessing. Recommendation
- Never paste guest personal or payment data. No names, emails, phone numbers, card numbers, IDs or medical details. Dates, party size and ages are enough. Recommendation Card data handling is also set by your card processor or merchant agreement. Contractual
- Safety calls stay with a qualified person. Weather go/no-go, water, glacier, wildlife, aviation and first aid: AI can draft the wording, a qualified person owns the decision. Recommendation Guide licensing is not delegable: AS 08.54 for big game guides, and annual ADF&G registration for sport-fishing guides and businesses. Legal mandate
- The examples are fictional. Every tool uses Bay Light Day Cruises, Seward, and its partners Mile 12 Kennel and Harbor Kayak Co. All three are invented teaching businesses.
See the fictional teaching facts
FICTIONAL TEACHING BUSINESS: Bay Light Day Cruises, Seward Offer: 4-hour wildlife and glacier cruise. Departs 9:00 and 14:00, May 15 to September 15. Capacity 40. Minimum age 4. Heated enclosed cabin, two restrooms, hot drinks included, lunch not included. Meeting point: small-boat harbor, a 12-minute walk from the cruise ship dock. Runs in rain. Cancels only for a marine warning, with a full refund. Free cancellation until 48 hours before departure; inside 48 hours, no refund. Channels: own website, Viator, Google Business Profile, two hotel desks. Not supplied (ask the owner): group discounts, private charters, pet policy, accessibility of the vessel. Partner A: Mile 12 Kennel. Summer sled-dog cart ride, 90 minutes, 15 minutes by car from the harbor, ages 2 and up, runs in rain. Referral code BAYLIGHT-K. Partner B: Harbor Kayak Co. 3-hour guided paddle, departs 8:00 and 13:00, minimum age 12, cancels in wind over 15 knots. Referral code BAYLIGHT-P.
Start here tonight
Run the audit first so you have a starting score. Then the three basic rungs: an afternoon each, free tier, no code.
- First · AI visibility audit What do AI tools say about you? Ask three AI tools what a real traveler would ask. Score whether you show up, described correctly. You get a score out of 40 per tool and a fix list. Tonight: run prompt 1 in one AI tool. Ten minutes. About 15 minutes per AI tool · free tiers work · feeds Number 1
- Then · Pain 1 · 1.a Rewrite for the short list Generic adjectives out, checkable facts in. The same facts on every channel. About 1 hour · free tier
- Then · Pain 2 · 2.a The honest picture Your reviews sorted into what guests wish they had known, plus reply drafts a person posts. About 2 hours · free tier
- Then · Pain 3 · 3.a The AI trip kit A one-page plan per booking with partner options and a weather plan B, plus post-booking messages. 1 hour to build, then minutes per booking · free tier
Looking for the four Monday tools from the slides? Tool 1 is the AI visibility audit. Tool 2, the one-listing test, is in 1.a. Tool 3, the restart-tax checklist, is in 2.b. Tool 4, the post-booking calendar, is in 3.a.
The map: three pains, nine tools
Rows are the three planning moments travelers most often call frustrating. Columns are rungs of difficulty. Each rung feeds the next, so start with the basic rung on all three and climb only what you can measure. Tap a tool to open it.
45% Pain 1 Comparison overload
Traveler: too many options and no way to tell them apart. 63% would pick ten curated options over 500 with filters.
What it costs you (my read): not on the short list means not in the decision, and AI increasingly writes the short list.
- Basic · 1.a Rewrite for the short list Generic adjectives out, checkable facts in, published the same on every channel. About 1 hour · free tier
- Intermediate · 1.b Your own curated ten A "Which trip fits you?" page: three picks per traveler type, each with its reason. A week of evenings · paid tier, about $20 a month
- Advanced · 1.c AI answer monitoring + structured data A weekly prompt panel across AI tools, plus schema.org markup so machines can read your facts. 2 to 4 weeks · developer or vendor · model API fees
43% Pain 2 Second-guessing
Traveler: the booking never feels final. 52% feel anxious about booking the wrong option.
What it costs you (my read): doubt shows up twice, as abandoned inquiries before the booking and as changes and cancellations after it.
- Basic · 2.a The honest picture Reviews mined for what guests wish they knew, a "What to expect, honestly" section, and reply drafts. About 2 hours · free tier · a person posts replies
- Intermediate · 2.b Grounded answer assistant A staff assistant that answers only from your facts, cites them, and otherwise says "Not in our facts. Ask the office." 1 day to set up, then a monthly check · paid tier
- Advanced · 2.c Post-booking retention engine Exposure scoring, drafted offers in an approval queue, and a holdout group to test the gain. 4 to 8 weeks · developer plus booking-system API
39% Pain 3 Coordinating the trip
Traveler: nobody holds the whole trip together. Nine research channels are each used by roughly a third or more of travelers.
What it costs you (my read): an Alaska trip spans several businesses, and every handoff brings repeat questions and missed add-ons.
- Basic · 3.a The AI trip kit A one-page plan per booking: partner options, what to wear, a weather plan B, and post-booking messages. 1 hour to build, then minutes per booking · free tier
- Intermediate · 3.b Partner circle with shared facts One fact sheet shared by three to five partners, and a kit drafted automatically a month out. About a week · no-code automation, roughly $20 to $30 a month plus cents per kit
- Advanced · 3.c Agent-ready trip record A structured trip record and reversible holds an AI agent can use. For most operators, it starts as one question to your booking platform. 15 minutes to ask · months to build, with a developer or vendor, or ATIA-wide
Percentages: share of US travelers naming each planning moment as frustrating. Source: Skift Research + McKinsey, Winning Hearts in an Age of Infinite Travel Choices, Sept 2026; survey of n = 1,015 US travelers, data as of June 2026. US travelers, not Alaska visitors. "My read" lines are the presenter's interpretation. Time and cost are estimates.
For your team
The same approach, pointed behind the counter: staff procedures for the jobs that live in one person's head, guest replies drafted only from approved facts, and seasonal campaigns. AI drafts. You approve.
The two ready-made assistants are draft instructions, not yet tested in a live account. Try them with the fictional example first.
Your first 30 days
One sitting a week. Each week ends with something written down.
- Week 1: run the AI visibility audit and write down your score per AI tool. Then 1.a Rewrite for the short list.
- Week 2: 2.a The honest picture. Publish the "What to expect, honestly" section. A person edits and posts the review replies.
- Week 3: 3.a The AI trip kit and the post-booking messages (in 3.a). Decide what you do when a guest finds a lower price inside the cancellation window. Policy choice
- Week 4: count the four numbers below and write the baselines. Then choose one intermediate rung from the map for next quarter. Bring your scores to a no-cost Alaska SBDC advising session at aksbdc.org. Recommendation
The four numbers
Count them the same way every time. The last column shows which tools move each number.
| Number | How to count it | How often | You want it | Moved by |
|---|---|---|---|---|
| 1. AI visibility score | The audit: 5 prompts × 4 checks, each scored 0 to 2, per AI tool (max 40). | Quarterly | Up | 1.a, 1.b, 1.c |
| 2. Direct share of bookings | Bookings on your own site or phone, divided by all bookings. Use your booking system's report by channel. | Monthly | Up | 1.a (the one-listing test), 1.b, 2.a |
| 3. Repeat questions per booking | Guest questions whose answer is already on your four surfaces (Google, OTA listings, your site, the confirmation), divided by bookings. | Monthly | Down | 2.b (the restart-tax checklist), 3.a, 3.b |
| 4. In-window changes and cancellations | Bookings changed or canceled inside the free-cancellation window, divided by bookings. Count flexible and non-refundable bookings separately. | Monthly in season | Down | 2.a, 2.c, 3.a (the post-booking calendar) |
No targets on purpose: set them after a season of your own data. At a few hundred bookings a season, a few points of monthly change is noise, so compare season to season.
The 90-day frame
- Days 1 to 30Establish the truth
One approved fact sheet, the basic rungs running, and a baseline for each of the four numbers.
- Days 31 to 60Test with review
Climb one intermediate rung. A person reads every output, and you log every correction.
- Days 61 to 90Keep, change or stop
Compare the old effort with AI plus review and upkeep. Decide in writing, then repeat.
Prompt: decide whether to keep a rung
Paste counts and times only. Leave out guest names, emails, phone numbers and payment details.
Your prompt
Using only my results below, compare the old effort with the AI effort plus review and correction. Subtract upkeep and show setup time separately. List errors and usefulness as well as time. Do not turn saved minutes into revenue or profit without evidence. Say whether to keep, change or stop this rung, and what remains uncertain. Mark the result DRAFT - PENDING OWNER REVIEW. [PASTE DE-IDENTIFIED RESULTS: the rung, weeks used, outputs drafted, corrections needed, minutes before and after, and the four numbers]