Best AI for a receptive travel agency (DMC): my take

I spent a year testing Claude, ChatGPT and Gemini to run Mexikoo, my DMC in Mexico. For a receptive agency, I recommend Claude. Not because it writes better, all three write well, but because it acts inside my own tools and it pushes back when I'm wrong. Here's the honest comparison, including where the others win, and a concrete example of what it actually changes.

One thing first, because it frames everything else: none of these AIs replace your operating software. At my agency, Odys stays the system of record, the place that owns the data. AI feeds Odys faster, it doesn't take its place. An agency that teaches its clients to work around its own production tool is shooting itself in the foot.

Prompting isn't enough: how far can AI really go inside a receptive agency?

Most people stop on the ground floor. They open ChatGPT, type a prompt, have it draft an email or a description, copy, paste. That's useful, but it's 5% of what the tool can do. The way I see it, there are three levels, and the real jump in value happens between the first one and the next two.

Level 1, assisted writing: you ask, the AI writes. An email, a reworded quote, a presentation blurb. Real but limited, because you stay at the controls at every step.

Level 2, the business skill: you stop writing a fresh instruction every time. You package an expertise into a reusable tool that runs a full task, the same way each time. This is where the time really opens up.

Level 3, the agent that acts inside your tools: the AI no longer just produces text, it takes control of your browser, opens your software, enters the data. The work happens while you do something else.

The difference between an agency that "uses AI" and one that gets real leverage from it is that move from level 1 to levels 2 and 3. And not every AI climbs those steps the same way.

Claude, ChatGPT or Gemini: which one for a DMC?

I pay for all three and use them for real, not for a three-day trial. Here's what each one does well and where it struggles, for our line of work specifically. Models and prices move fast, so this table is true for today, not forever.

If you already live entirely on Google Workspace and mostly want help inside your emails and spreadsheets, Gemini is a logical choice and often free within your plan. If you want the best-known, most versatile tool to start with, ChatGPT is a safe bet. But the moment you want to industrialize tasks specific to a receptive agency and have them run inside your software, that's where Claude wins for me.

Why I recommend Claude for a receptive agency

Honestly, it's not writing quality that settles it. It's three very concrete things.

First, Skills. A skill is an expertise you write once and the AI replays identically after that. Gemini's equivalent, Gems, stays mostly a saved instruction on top of a chat. A Claude skill goes further: it carries a method, steps, files, and it fires inside the context of your real work.

Second, Claude acts inside your actual browser. Through the Chrome extension, it reads your pages and enters information straight into the tools you're already logged into. That's the detail that turns help into execution. ChatGPT's agent can browse, but in a separate browser, which breaks the thread of your logins.

Third, and this may be my favorite part, it doesn't just obey. I've set my preferences so it never agrees by default, looks for blind spots, and behaves like a consultant rather than an assistant. The result is that it doesn't charge off to run a bad instruction: it asks a question, it flags a risk. With the other models I've used, I rarely got that reflex to push back, and that's the difference between a tool that executes and a tool that makes you better.

What is a skill, and how is it different from a chat?

A chat is a conversation. You ask a question, you get an answer, and next time you start from scratch. Perfect for exploring, not for repeating.

A skill is the opposite. You lock in, once and for all, the way a task is done, with its rules and its steps, and you rerun it whenever you want without re-explaining. The advantage over a chat is simple: consistency. The skill does the task the same way every time, without you depending on how good your prompt was that day. For an agency, that's the difference between a one-off hand and a process that holds.

Chat, Cowork or the Chrome extension: which one to run a skill?

For real automations, I don't use plain chat, I use Cowork. Cowork is the mode where Claude works on its own, reaches your files, chains the steps by itself and sees a long task through. Chat asks you to validate at every step; Cowork runs the whole workflow.

A lot of people think this kind of tool is only for developers, because it started on the code side. That's a mistake. The same power handles a contract, fills in software, produces a report. In practice, to run a skill, I use Cowork and let the Chrome extension take over the browser when something needs to be entered into a tool.

A concrete example at Mexikoo: the hotel contract handled in 10 minutes

Here's the case that convinced me. Every season we receive hotel contracts as PDFs, with rates by period, conditions, seasons. Before, someone keyed all of that into our tool by hand. Count thirty to forty-five minutes per contract, plus the data-entry error you pay for later.

Today, a skill reads the hotelier's PDF and turns it into a clean Excel file. Then the Chrome extension takes control of the browser and enters every season straight into Odys. What took thirty to forty-five minutes now takes one minute to drop in the PDF, two minutes to check the Excel, and the rest runs while I do something else. A forty-five-minute job drops under ten minutes.

That's the level 3 I was talking about. And it repeats across a whole swath of a receptive agency's tasks.

My 5 favorite skills for a receptive agency

I'll cover each of these skills in its own article, with the skill to grab. Here's the series at a glance:

• A competitor benchmark that analyzes my positioning, my client reviews against competitors, and the keywords that are saturated or worth taking.
• A roadbook quality scan, a kind of audit before sending: itinerary coherence, logistical gaps, inconsistent travel times, contradictions, and even safety alerts like an activity that includes peanuts when the client is allergic. It also pushes you to apply the principles of unreasonable hospitality, a concept popularized by Will Guidara, because a receptive agency stands out by offering what the client didn't even ask for.
• A hidden-gems hunt that searches online for the hotels and activities that are hard to find, the ones that aren't on Booking or the big platforms, buried deep in forums, and that make our added value as a DMC.
• A copy skill to write hotel, activity and destination descriptions with a soul, avoiding all the AI turns of phrase we've seen a thousand times.
• A sales performance analysis that takes an Odys export and produces sales trends, top sellers, conversion rates by market and by agency, time to close a file, with concrete actions to stretch certain KPIs.

Where to start

If you take away one thing: stop stopping at level 1. The real leverage for a receptive agency isn't drafting an email faster, it's automating a whole process and letting it run inside your tools.

Want to test one of these skills on your own case? Get in touch: we'll run it together on one of your contracts or roadbooks, I'll give it to you, and I'll show you along the way how it feeds Odys.

FAQ

Do you need a paid subscription to automate your agency with AI? For plain chat, no. But for the game-changing uses, Claude's Skills, Cowork and the Chrome extension, yes, a paid subscription is required. That's the entry ticket to move from assistance to execution.

Will AI replace receptive travel agencies? No. It replaces the repetitive part, data entry, formatting, the first round of research. It doesn't replace local judgment, the relationship, the improvisation when a flight gets cancelled at 11pm. A DMC's value was never the logistics, it's just that logistics ate 80% of our time. AI gives that time back.

Does Claude replace an operating software like Odys? No, and this matters. Claude executes and feeds, Odys owns the data and runs production. The two complement each other, one doesn't replace the other.

Which AI should I choose if I'm starting out? If you live on Google Workspace, try Gemini, often already included. If you want the most versatile, ChatGPT. If you want to automate business tasks and have them run inside your tools, Claude. My daily use leans toward Claude for that reason.