API only is moving within reach as a target image - API first may help

Early October 2025 - the AI news is coming thick and fast. On the one hand, Google announces its AI mode; on the other, openAI takes an evolutionary step towards the operating system and opens chatGPT with its App SDK for external apps. Two sides of the same coin - on the social side, the abolition of individual thinking is being pushed forward and on the other, the technical side, the further development from API first to API-only is taking place for all companies that still want to "play along" in the future.

Well, this is basically nothing new: the further re-education to abolish one's own way of thinking. Google has been doing this for decades and we are now well de-sensitized when it comes to giving thanks. Whether it's phone numbers, birthdays or addresses - we store a lot of things in our devices and leave the "thinking" to the machine, whether it's calling by voice control "OK Google, call mom" or "OK Google, navigate me home" - even for small tasks we no longer use our own brains, but outsource the thinking to the software.

The same happens with passwords, by the way. Who still knows their passwords off the top of their head? Due to numerous threats and the media emphasis on the danger of phishing, Trojans etc., we not only need two-factor authentication, but also the use of password managers. The master password is sufficient, everything else is stored in the tool. Fortunately, if this breaks or is leaked, everything is gone at once. You can then only hope that the forgotten password function works.

Effects on content/SEO agencies and click-based business models

Let's take a look at the business side of the AI revolution and what will happen as a result of the AI mode, which is already in full swing.

Google AI Overviews (formerly Search Generative Experience, SGE) will be expanded in the direction of Generative Engine Optimization (GEO). This will lead to even more "zero-click searches" - the answer is already on the SERP, which will put further pressure on the business models of publishers and content providers. The CTR (clickthrough rate) alone will have fallen by almost 35% by 2025. This will have a massive impact on the advertising revenues of content platforms and will ultimately mean that there will be very little free content on the internet. Much of it will disappear behind a paywall or not be produced at all. In order to remain relevant for Google and the various AI platforms, the only way will be to provide essential content via API or feed for the AI platforms and to be paid for it in return. Whether only for training or direct online searches remains to be seen - Reddit already took this first step at the beginning of 2024.

In combination with the recently introduced ACP (Agentic Commerce Protocol) from Stripe and openAI, online shopping will also change. Everything goes in the chat, right through to checkout and payment. Anyone who does not offer an API and API skills in their system/IT landscape has already lost out.

In the context of online shopping, the use of AI is also seen as an "option for personalization" or even the satisfaction of customer wishes for hyper-personalization - unfortunately understood a little too shallowly, because this is more than newsletter or chatbot personalization.

Hyper-personalization yes, although this customer need ("personalized products and offers") essentially relates to the product, not "only" to individual communication. A great chatbot or agent that guides customers through the range and curates products is just the beginning. We already had simple and pre-structured dialogs and filtering 20 years ago, back then it was called "guided selling". The necessary step here is the hyper-personalization of products, the individualization of what the customer ultimately uses or wants. Advising the customer is just one step in the customer journey and therefore in their user experience - but this doesn't start in the store and doesn't end with payment.

API only - sensible goal or dystopia?

In view of the thoughts already written about API first, I am currently asking myself whether we are not even heading in the direction of API onyl?

If we take the scenario we are currently facing a little further, then in my opinion there will soon be no need for classic UX designs for websites or online stores, will there?

In the context of AI, there are not only agents that can perform individual tasks, but also complete agent systems where an agent interacts with another agent to solve more complex tasks. If, as already anticipated, the use of external agents is added to your own agent system, why do you still need your own store, your own customer journey or your own emotionalized manufacturer brand? Isn't it enough to offer and deliver a personalized product at the best price?

I'm thinking of Amazon right now. My focus is less and less on the brand of the product on offer and also less and less on the manufacturer or shipper. Basically, I decide to buy even obvious Chinese goods because I know that if it goes wrong, Amazon is liable with its brand and will stand up for me. This means that it is not the manufacturer that is the anchor point, but the brand of the seller or marketplace.

The same will happen with AI and agents. If the user trusts the platform, they will use the agents and this will extend to their wallet or instant checkout. Stripe has intensified its collaboration with openAI in this area. Together, they offer a protocol that can be integrated into a chat platform in order to map the payment process as well as the shopping cart. This means you no longer need a store, a chat is enough. This basically means that the "consultation" and the "purchase" can actually be mapped in a personalized interaction model.

A look at the App SDK just presented by openAI also shows the same picture. In the first step, openAI wants to "communicate" with external apps in ChatGPT, i.e. retrieve, search, evaluate and display their offers within the ChatGPT UI. This will start with apps (providers) such as Booking, Expedia, Spotify and others. If openAI is as consistent as it has been so far, in the next step it will no longer be possible to see which external app is being communicated with, and it will probably no longer be necessary to log in to the external providers to use paid services - everything will be aggregated and centralized in ChatGPT, as with a central operating system (I somehow remember WeChat - it's not all that new what's happening, is it?). In the long term, this strategy will lead to providers of all stripes developing even more into pure fulfillment service providers or content generators, making their own brand irrelevant in the digital ecosystem.

With its App SDK offering, ChatGPT has opened the door, just like Amazon or Apple, to creating an ecosystem from which there will be no escape for end customers and providers (without thinking for themselves). Take it or leave it - the winner takes it all.

However, an essential prerequisite for being able to play in this league as a company or provider are fundamental skills that need to be developed or utilized.

Companies in the digital economy should do 3 things now to avoid being left behind.

API first - no API, no chance

As has been described often enough, the chat platform needs access to product information, prices and offers. The majority of the more static data (description, images) can probably be made accessible to the AI via a RAG integration via API, while prices and delivery times should be connected dynamically via real-time interfaces. In concrete terms, this means that data must flow, preferably automatically.

In the same context, if no generic chat AI platform is used, existing platforms such as ChatGPT or Perplexity or even Google must be given access to the data and interaction options. This requires publicly accessible interfaces. Anyone who does not yet have public APIs or hides everything behind logins or tokens will soon be left behind. Even the reddit approach of offering AI platforms content for training or consumption in return for payment will not be sustainable in the long term - the power of AI is already too great for this, and that of the providers too small in the medium term.

However, it is not only the API that is essential, but also its range of functions, up-to-dateness, management and documentation. If there is no dedicated AI/chat platform, then this API must be able to map and respond to all the different requests - and this API must be well documented with valid examples and publicly accessible.

Content - LLM and Knowledge Graph as a reliable combination

In this context, content does not necessarily only mean the product content, but also the way of communication, the style in which the communication between customer and chat(bot) should take place. One challenge lies in transforming beautifully structured databases into continuous text, as this is much better understood by the AI in the learning process. This can be done manually, with AI support, or so-called knowledge graph implementations can be used to feed the database as structured data. The combination of LLM with Knowledge Graph then delivers the best results.

Another challenge is the method of communication if a dedicated chat (bot) platform is planned. This can already be solved today by capturing the style of communication in texts as if you were talking to the customer yourself (sources for this can be transcribed hotline calls or support emails) and then making these available to the AI for training via RAG mechanisms and technologies. The AI then learns from these texts how the company or store would "speak" to customers and adapts this in the answers in the chatbot.

Integration - those interfaces again

The third component is the integration of the AI/chat platform into the company's own systems or management. Here too, ordering products or services is just one element of the customer experience. This is followed by booking, logistics and, if necessary, returns. As in the past, all of this should take place in stable, audit-proof backend systems, whereas customer communication must be quickly adaptable and flexible. In order to do this justice, it is therefore necessary to be able to import the order data (product, customer, delivery options, payment status, etc.) into the backend. Zack, next API or at least importer or other type of automatic connection.

In my opinion, if you believe in the triumph of agents, you also have to believe in the API-only approach, provided you still have your own business. Alternatively, you can hang up your retail business and become a manufacturer - which is a good idea anyway. Having your own brand plays less and less of a role; what is important is a product/product range that can be produced inexpensively, customizable and scalable.

Social consideration - the loss of one's own thinking leads to the loss of cognitive abilities and thus the loss of creative solution finding.

In my view, the biggest social challenge is the issue of information reliability. If we are satisfied with "zero-click results" and no longer question what we are actually consuming and if we increasingly and often unwittingly allow ourselves to be caught up in a (social media) bubble, then the power of big tech will grow immeasurably. That's what I'm most afraid of. That untruths and fake news or any other content will gain weight and thereby control and manipulate societies, countries or communities. As I said, we have been trained for years to neglect independent thought more and more - are agents and AI agent systems the next step towards a lack of willpower and how does the equation of loss of thought = loss of control affect society and manipulation?

I hope that the MIT study will be heard and heeded. One key message is that too much use of AI is dumbing us down and people's cognitive abilities are declining. If we don't use it at all, we will fall behind the possibilities, we will be too inefficient and therefore no longer competitive.

It's basically like with everything - the dose makes the poison. We should just be aware that every use of AI makes us think less - and that's why this article is actually written by us.

Conclusion: Between opportunity and loss of control

API only is more than a technical strategy - it is a future model for digital commerce. Companies that take interfaces, content and integration seriously can tap into new value creation potential.

At the same time, the social question remains: How do we preserve critical thinking in a world where AI chews up answers for us?
The solution probably lies in the right balance: using AI as a tool - without outsourcing our own thinking.

Recommended action for companies in the digital economy

1. Review and expand API strategy now.
2. Design content in such a way that it can be used by AI systems.
3. Observe agent ecosystems - and launch pilot projects.