Three years after the launch of ChatGPT; after a 2023 spent building cool prototypes, a 2024 where you got the impression that you were really getting your hand at this and you were really “getting it” — fast-forward to today: EVERYONE is trying to implement GenAI. Everyone is trying to sell you a BOT and, you know what? What no one is telling you is that you are wasting your time.
Yeah, you got that right.
Your BOT is meaningless. Do you know why? Because the BOT that your customers are going to be using tomorrow to access, search, or perform actions is not the one you are about to deploy on your website or on WhatsApp.
The BOTs people are going to use tomorrow are those of the likes of Perplexity, Gemini, OpenAI, Alexa+, etc. As SEO ranking becomes progressively useless there is only ONE thing that will ensure visibility for your company: a robust MCP catalog.
MCP stands for Model Context Protocol, and you can think of it like a special door for AI agents to use. Companies that will quickly adopt the standard are those that will benefit the most from this epic shift to an internet where we all search less but interact more.
Take airlines as an example: today you don’t have to look on their own individual websites anymore when you are planning a trip, as you now have players like Skyscanner or others, which essentially can send an API request to multiple airline companies and allow you to easily see every possible combination for any possible price.
Today AI agents can do the same, as long as you have built the proper MCP endpoints for them.
So, to come back to my original point, focusing on your BOT is a waste of time in the long run, unless you are a pure player and your business makes money by building and selling BOTs.
Have you tried Perplexity’s BOT when looking for things? I am 99% sure that its BOT will perform better than whatever you might be able to produce on your own, for the simple reason that if you are not an AI company you cannot invest the same amount of energy/money/talent on that topic, and so your output will always be behind the bleeding edge. You will always play catch-up.
Don’t get me wrong, building this stuff is FUN, ok? I love it. I had LOADS of fun with my team and we have built some pretty damn cool stuff. But, if we have to be realistic, you need to admit that I am right: in the large scheme of things, individual BOTs will be irrelevant.
So, what should WE do? You need to think of your MCP catalog. You need to think of its architecture. And you need to do that NOW.
Every time there is a disruptive new technology you see the companies slow to adapt fade away, and those who move at the right pace rise above their competitors.
In a few months those BOTs will be able to book restaurant tables for you (Perplexity allows that for some in the USA), help you plan your next trip, your hotel, your cinema tickets.
Imagine a tourist planning a trip to Paris. He uses Alexa+ to plan the trip, find a place to eat where he can see the sunset, or maybe one that has a scenic terrace AND allows you to bring your dog with you. LLMs are perfect for that stuff. But what if your restaurant does not have an MCP? Well, it will simply not be proposed by the BOT, which will instead propose your competitor who happens to have an MCP endpoint.
Do you see my point? The revenues will not be driven by your company’s BOT. People are already moving away from a classic search experience, and we are exponentially moving towards a conversational experience. If your service is invisible to these BOTs, it will simply not “exist”, which is catastrophic—at the same level as businesses that could not be found on Google just a few years ago. No visibility = no traffic. No traffic = no business.
So, to wrap it up: do develop your BOT and deploy it to have fun, sure. But remember that what will really make the difference for your business tomorrow is for your services to be easily accessible by those LLM agents.
The best moment to have MCP endpoints was yesterday. The next best moment is NOW.