A recent estimate placed AI and related investments at around one-third of the total US economy. At the same time, Bloomsbury warns of a potential AI investment bubble driven by unrealized gains, which could trigger a recession due to the large share of the market invested in a single sector.

At this point, the question is not whether the technology is profitable or not, but whether your application of it and the core of your business should really be reselling tokenized APIs. I’ve seen it more than once: a group of entrepreneurs decide that applying a connector to a website in tandem with an AI model will make them millions.

Not only is this misguided, but it is also ignoring a basic business tenet: if you don’t own the IP or have a unique proposition, then you don’t own a company.

This is an almost classic commandment and fundamental principle in any basic business course, but the fake gold rush caused by LLMs has hoodwinked more than one person under the promise of easy money. 

To begin with, most plugins and APIs fall under very few actors in the market; this is the reason why both OpenAI and Anthropic make such huge waves every time there is a change. Many “companies” are effectively running as subsidiaries of their technology, built on their infrastructure and simply adding a wrapper on top of their product.

The problem is that not only is your idea vulnerable, but if your information is dependent on them, then anyone with a modicum of prompting capacity can completely bypass you. The company will function as long as you manage to sweet-talk your way to potential buyers, but Gartner expects around 70% of agentic AI implementations to disappear by 2029.

Asking the model to “impersonate” or simulate expertise is not really difficult to achieve, and is both unethical and imprecise. 

How can you identify a hallucination from an “expert doctor” if you have no medical knowledge yourself? Having a second agent “controlling” the result is also not a good solution if they hallucinate in tandem. But then you doubled your costs, effectively building the provider’s service instead of your own. Having no expertise means you are sitting on a time bomb until you produce a noticeable failure, big enough to sink the business, or a real expert takes you out.

The core business of AI companies is data. It’s unlikely that you will manage to develop or outcompete the vast swaths of data they own (let alone match their capital), but without a model of your own, you might not really have a business at all.

Many will even ask, “What about Amazon or Tesla? They are coming in late too!” But that is also a very naive comparison. To begin with, neither of those businesses has AI as a core sector. Tesla is an industrial company, and Amazon is e-commerce. For them, the models are a means to an end, not the reason for being.

In the end, solid business principles will remain robust; without a differentiator or clear expertise, you have no gap (and potentially no place) in the market. Tighter economic conditions will surely patch it up. 

Make sure, at least, that you’re not chasing fool’s gold by pawning your house (or anyone else’s, for that matter).

Notes:

Some quick reads worth a bit of time:

1. AI Now Accounts for a Third of US Market Value—What That Means For The Economy

2. AI Bubble Risk: Echoes of Dotcom? — Bloomsbury Intelligence and Security Institute (BISI)

3. 2026 Hype Cycle for Agentic AI | Gartner

---

The Pocket AI Guide is out!

📙 Amazon US: https://a.co/d/gCHHDax
📗 In Europe Amazon Germany: https://amzn.eu/d/3cmlIqa
(Available in other stores Amazon stores too in Europe)

Check the free resources in this website!