There are two ways to deploy artificial intelligence in a business. The first is to integrate it into the core of your systems: connectors to the customer database, synchronisation with business tools, migration of historical data, proprietary formats. The second is to place it alongside them: an external harness that plugs into what already exists and can be detached. Koaee has chosen the second, and this article explains why that choice commits the provider more than the client.
What “forward engineering” means here
The term comes from software engineering: building forward from what has been specified, rather than deconstructing what exists in order to reproduce it. Applied to an AI deployment, it changes the order of operations. You do not begin by opening up the information system to bring a model into it; you begin with the process as it runs today, and you build outward.
The consequence is concrete: nothing that exists is rebuilt to accommodate the tool. The website remains the website, the booking database remains the booking database, the procedures remain the procedures. What is added is a conversational layer placed on top, which reads what it is given to read and writes nowhere else.
The model that locks you in is not malice, it is a slope
Deep integration is not a trap that has been set. It is the natural outcome of a commercial dynamic: the more a system is connected to others, the more useful it becomes, and the more useful it becomes, the more costly it is to leave. Each connector added is an immediate functional gain and, over time, one more dependency. The exit bill is not paid at installation; it is paid on the day you want to change your mind.
This slope always produces the same symptoms: data reformatted into a proprietary model, business rules rewritten in the provider’s tool rather than documented elsewhere, teams trained on one interface. None of these can be reversed in a day. This is not an accusation, it is a mechanism, and it is precisely the one an external harness avoids.
What an external harness looks like
The same event, wanting to change tools, does not carry the same cost depending on the architecture chosen. On one side, a chain of operations, each link of which must be negotiated; on the other, a single gesture. It is the difference between a load-bearing wall and a piece of furniture.
“Locks no one in”: three checks, not a promise
A promise of reversibility is only worth something if it can be tested. Here are the three questions to ask any conversational AI provider, including us.
First, how do you leave? At Koaee, the assistant is installed with one line of script on the pages where it should appear, and removed by deleting that line. There is nothing to uninstall anywhere else, because it installed nothing anywhere else. The next day, the site is exactly what it was before.
Second, where does the data go? The question is not rhetorical, and the honest answer is rarely a simple “with us”. Our servers are in Germany, in Frankfurt. Language processing runs through Mistral, and for French as for English, listening and voice do too: nothing leaves Europe for these two languages. For the other languages, voice still relies on engines outside Europe. We publish this on our transparency page rather than rounding it off.
Third, who owns the rules? The instructions that govern the assistant (what it should say, what it must not promise, how it guides the conversation) are an asset of the client, not a secret of the provider. They are written in readable text, not encoded in an interface. A client who leaves takes them with them.
What this choice costs us
It would be dishonest to present the external harness as superior in every respect. It is not. A deeply integrated AI sees more: it can cross-reference a purchase history, trigger an action in a business tool, write to a file. An external harness sees only what it is shown, and can act only where it has been explicitly connected. That is a real limitation.
We accept it because the trade-off seems more valuable to us for a business leader: the ability to change one’s mind. A tool that can be removed with one line is a tool you keep by choice, not because of the cost of leaving. And for the provider, it is a discipline: it has to be useful every month, since nothing holds the client back.
The regulatory framework points in the same direction
The European AI Act (Article 50) requires that people be informed that they are interacting with an AI system, unless this is obvious. This transparency obligation is easier to meet when the assistant is an identifiable layer placed on a website than when it is spread across around ten internal systems where no one knows any longer exactly what is responding. Here, technical reversibility and regulatory clarity are two effects of a single architectural decision.
It is also what allows us to name our engines plainly and, above all, to change them: because nothing has been fused into the client’s systems, replacing one model with another asks nothing of the client. The map of engines above is not a permanent commitment; it is a snapshot at a given date, and that is precisely what an external harness makes possible. These choices may change. What matters is that they can change without the client having to redo anything, because the client was never connected to the model, only to the harness.
Source: Eurostat, survey “Artificial intelligence by size class of enterprise” (isoc_eb_ai), businesses with 10 to 249 employees, France. Retrieved via API on 30 August 2026. The survey publishes neither 2022 nor 2026: the series stops at the latest available year.
In one sentence
We do not seek to become indispensable through entanglement. An assistant that installs with one line and is removed with one line obliges the one who sells it to remain useful, and leaves the one who buys it the one thing no contract gives back: the freedom to leave.
Further reading
- European Commission — European approach to artificial intelligence
- Regulation (EU) 2024/1689 on artificial intelligence — official notice from the Publications Office of the European Union
- European Commission, AI Act Service Desk — Article 50: transparency obligations for providers and deployers of certain AI systems
- Koaee — transparency page: hosting, service providers, limitations
