A framework · not a pitch

AI for your travel business — a way to think about it.

There is no bad time to adopt AI. There are a lot of bad ways.

What follows is how I'd think it through if I were you — regardless of whether my product ends up being the fit. Skim the map, then read what interests you.

Full disclosure

I built Orbis — an AI back-of-house for companies that run on relationships. Not just software: a small senior team behind the product that actually shapes and implements it for each business. That's a real conflict of interest in a document like this.

So the deal: the first six sections don't mention Orbis. They're the questions I'd want a friend to ask before spending a dollar on any AI tool, including mine. The seventh section is where I lay out honestly where Orbis fits and where it doesn't. Then you decide.

The seven considerations

How to think about AI in a travel business

THE QUESTION Is AI right for your business? (and in what form) 01 Start with the problem, not the technology. 02 What AI does well — and what it doesn't. 03 The three paths: generic · vertical · partnership 04 Questions to ask any AI vendor. 05 Risks specific to the travel business. 06 A 30 / 60 / 90 plan before spending money. 07 Where Orbis fits (full disclosure).

Tap any node to jump to the detail. Read in any order.

01

Start with the problem, not the technology.

The worst AI adoptions start with "we need AI." The best start with "this week my team lost twelve hours chasing supplier confirmations."

Before you evaluate any tool — including mine — you owe yourself honest answers to four questions:

  • Where does my team's week actually go? (Do a real time audit. Not a guess.)
  • What do clients wait on longest, and why?
  • Where does information die — in someone's inbox, in someone's head, in a WhatsApp thread nobody reads back?
  • If I hired a perfect operational coordinator tomorrow, what would they spend their first week doing?

If you can't answer those, no AI will help. You'll pay to run your existing confusion faster.

Rule of thumbAI takes a team from 60% capacity to 90%. It does not take a team from 0% to 100%. If your operations are broken, AI will amplify the break.
02

What AI actually does well in 2026 — and what it doesn't.

Most vendors tell you what's possible. Fewer tell you what's reliable. For a travel business, the distinction matters.

What modern AI is good at:

  • Turning unstructured mess (emails, WhatsApp threads, voice notes) into structured records
  • Coordination — chasing overdue items, sending briefings, surfacing what's stale
  • Drafting — replies, itineraries, supplier messages that a human then reviews
  • Cross-channel memory — one view of a client regardless of where they last messaged
  • Pattern recognition — noticing that a client hates red-eyes without being explicitly told

What AI is still not good at:

  • Taste and judgment in edge cases. A human still decides what a VIP actually wants.
  • Fixing broken operations. It cannot compensate for an unclear process.
  • Building supplier relationships. Those remain human.
  • Guaranteed accuracy. Hallucinations are real. Any serious product ships with a human verification layer on anything client-facing.

The useful frame: treat AI as an operational second brain, not as an autonomous agent making binding decisions on your behalf. The moment a tool promises the second thing, ask harder questions.

03

The three paths — generic, vertical, partnership.

Every AI purchase falls into one of three buckets. They serve different needs and cost very different amounts.

Path What it is When it fits
Generic tools ChatGPT, Claude, Gemini. You + good prompts. Zero integration. Solo operators, early experimentation, drafting. Almost always part of the stack regardless.
Vertical SaaS Off-the-shelf travel-industry software with AI features bolted on. Commodity workflows you're happy to share with competitors — booking engines, reconciliation, content generation.
Productized partnership Your own dedicated AI instance. Your data, your voice, your channels. Not a login product — it lives where your team already works. Relationship-driven operations. Information in heads and chats. You want differentiation, not parity.

These aren't exclusive. Most serious travel operators end up using all three — ChatGPT for drafting, a vertical tool for GDS or reconciliation, and a partnership for the operational spine that ties clients, suppliers, and staff together.

The real question isn't which path?. It's which path for which problem?

04

Questions to ask any AI vendor.

If a vendor can't answer these crisply in a first call, that's itself an answer.

  1. Where does my data physically live — on your servers, mine, or a third party's?
  2. Can I export everything — contacts, conversations, history — in one click at the end of our contract?
  3. Who owns my client list and supplier list? Contractually.
  4. How are errors handled — silently dropped, or surfaced to me honestly?
  5. Is there a human verification layer on anything client-facing, or does the AI act unilaterally?
  6. Are my competitors running the same shared instance, or is mine isolated?
  7. What happens when your underlying AI model gets swapped for a new one — do I get notice, or do behaviours just change overnight?
  8. What's the failure mode when your service is down? Does my business stop?
  9. What does cancellation actually look like on day 1, day 90, and day three years?

Any vendor — including mine — should welcome these questions. A vendor that deflects is telling you something important.

05

Risks specific to the travel business.

Travel is not a generic industry. The things that make it wonderful — trust, relationships, human judgment under pressure — are exactly the things AI handles worst.

  • Client trust is the product. One hallucinated flight reference breaks a decade of relationship. Any AI that touches a client-facing message needs a human in the loop or deterministic verification behind it.
  • Compliance is not optional. PII everywhere, often payment data, often minors on family bookings. GDPR, PCI-DSS, and regional laws all apply. Ask every vendor where the data goes and under which jurisdiction.
  • Duty of care. If a traveller safety alert is quietly summarised away by an AI assistant, liability is still yours. Alerts must surface, not compress.
  • Suppliers are human. Hotels, DMCs, and ground operators often do not want to negotiate with a bot. Keep AI internal-facing — on your side of the wall — until you know what your supplier relationships tolerate.
  • High-touch clients expect invisibility. The best AI for HNW travel is the AI a client never notices. It shows up as faster responses, better briefings, nothing missed. It does not show up as a chatbot.
  • Language and cultural nuance. Travel is global. The subtext of an email from a Japanese supplier is not the same as from a Colombian DMC. AI handles this unevenly. Keep a human in the loop where it matters.
06

A practical 30 / 60 / 90 plan before spending money.

Most AI disasters trace back to someone skipping the audit and buying a platform to solve a problem they hadn't yet defined. The antidote is boring and works.

Days 1 – 30

Time audit.

Track where the team actually spends hours. Tag every activity. You will be surprised. This document alone is worth more than most AI tools.

Days 31 – 60

One workflow, end-to-end.

Pick a single workflow with obvious ROI — client briefings, supplier follow-ups, trip debriefs. Automate it end-to-end with one tool. Measure before and after.

Days 61 – 90

Measure and decide.

Did coordination overhead drop? Did response times improve? Did anything slip? Only now are you qualified to evaluate a bigger platform commitment.

If you follow this, you will make better vendor decisions than 90% of your competitors — regardless of whose product you pick. If a salesperson (including mine) tries to sell you a platform before you've done the time audit, push back.

07

Where Orbis fits — full disclosure.

I built Orbis for my own companies — an agency, a fund, a travel concierge — and productised it when the operational pattern held across all three. That's where my bias comes from. Here is the honest shape of it.

Not just software · your AI back of house

We're a team — not a log-in.

Most AI vendors sell you a tool and hand you the documentation. Orbis works differently. Think of us as a small senior team sitting behind your business — shaping the AI layer to your actual workflows, building custom integrations when off-the-shelf isn't the right answer, and staying on as things change.

The software is the starting point. The implementation — and the humans maintaining it with you — is the product.

For a travel business, that typically looks like:

  • A few weeks of shaping your instance to your real operations
  • Wiring it into the channels your team already uses — WhatsApp, Slack, email
  • Training it on your clients, suppliers, and preferences
  • Building custom automations when your workflow doesn't fit a template
  • Being on call when something needs a human touch or a judgment call

Not a subscription you log into. A partnership you call when things get real.

What Orbis is

A per-tenant AI chief of staff for relationship-driven businesses.

Your own dedicated instance. Your own database. Your own voice and rules. WhatsApp and Slack first — it listens in the channels your team already uses, not in a new portal you have to log into.

The core function is coordination overhead: follow-ups, briefings, stale-deal nudges, contact intelligence across channels, surfacing what's falling through the cracks. The things a great operations person does, done continuously.

Good fit if
  • Your business runs on relationships, not on a booking engine
  • Knowledge lives in heads and chats, not in a CRM
  • Your team already operates in WhatsApp / Slack
  • You want the AI to be yours — not shared with competitors
  • You value coordination reliability over raw automation volume
Not a fit if
  • You need a booking / GDS / reservations engine
  • You want a client-facing chatbot to replace concierges
  • Your operations are already systemised in vertical travel SaaS
  • You expect zero human verification on outbound client messages
  • You want a plug-and-play SaaS with no setup period

The questions in section 04 apply to Orbis exactly as much as to anyone else. Ask them. If any answer is softer than you'd like, push me on it before buying anything.