Sketching out a dashboard, adding features as you go, sharing the result with colleagues via a link – and all without writing a single line of code yourself: this is exactly the workflow currently described under the buzzword vibe coding.What is meant is development in which syntax is no longer typed, but rather what a tool should do is described in natural language.An assistant such as Claude Code generates the runnable version from this, which is then refined further in dialogue.
For business departments, this is a break with the usual sequence.Until now, the path to a small internal tool was: formulate the need, document the requirement, wait for prioritisation, commission implementation.Now the first versions are emerging where the need sits – in controlling, in sales, in marketing.
Where the efficiency gain actually arises
The benefit is greatest for tools that nobody builds today because they never paid off economically.The list of such gaps is long in almost every mid-sized company:
- Overviews that merge figures from multiple exports so that the same table does not have to be created by hand every week.
- Small input forms for recurring reports that previously ran by email or word of mouth.
- Checklists and calculators that map specialist logic which otherwise exists only in the heads of individual people.
- Prototypes that can be used to quickly demonstrate an idea before a decision is made about a proper project.
The second, less obvious gain lies in communication.Anyone who has built a working miniature tool themselves subsequently formulates requirements more precisely.Instead of a wish list, there is a tangible version on the table that can be used to discuss what is missing and what is superfluous.For project managers, this is often more valuable than the tool itself.
The difference between running and lasting
However, this is exactly where the zone begins in which self-building without guidance becomes expensive.A prompt-generated tool is initially only one thing: runnable.Whether it is sustainable is decided by properties that you do not see when trying it out.
Maintainability. Code that nobody in the company has read is, in doubt, also nobody's responsibility.As long as the generating tool takes over every change, this works.As soon as an interface changes, a dependency needs a security update or the person who created the tool leaves the company, the person who can assess the state is missing.An afternoon project then becomes a legacy system on which a department depends.
Data protection and data storage. The critical moment is rarely the generation of the code, but the first real data record.As soon as customer data, personnel information or calculations move into a self-built tool, the same requirements apply as for any other system: where the data is located, who accesses it, how long it remains there, how to get rid of it again.A dashboard shared via link is convenient – and the question of who can pass this link on belongs before first use, not after.
Integration. Isolated tools create duplicate work.If master data is held in the ERP, content is maintained in the TYPO3 backend and a new dashboard draws its figures from a manual export, then the overview is outdated from the moment it was created.The value of an internal tool rises abruptly as soon as it is connected to the existing systems – and that is precisely the part that requires architectural knowledge: permissions, interfaces, error cases, operation.
A pragmatic framework for companies
The answer to these risks is not a ban.Bans only move such tools into private accounts, where nobody has any insight anymore.What makes more sense is a clear framework in which self-building is expressly permitted – with a limit at which a project moves into professional implementation.
- Approved playground. Define which tools, in which environment and with which data experimentation is permitted.Test data instead of real data is the simplest and most effective rule.
- Criteria for the transition. Define when a prototype is no longer a prototype: as soon as personal data is processed, as soon as several departments depend on it, as soon as external parties have access or as soon as an outage blocks work processes.
- Inventory. Record which tools exist, who built them and for what purpose.This list is the only protection against shadow IT growing unnoticed.
- Transfer instead of rebuilding. A good prototype is a specification.It shows the data model, operating logic and special cases – and significantly shortens the subsequent clean implementation.
What this means for the roles in the project
Prompt-driven development lowers the barrier to entry to the first working version, not the effort required for operation, security and integration.The work shifts: away from writing simple interfaces, towards questions of architecture, data quality and the question of which tools a company wants to operate permanently.For developers, this means less routine and more consulting.
For specialist departments, this means more scope for design – and more shared responsibility for where their own expertise ends.
For web and software projects, it is worth actively planning for this development instead of ignoring it: self-built prototypes from the specialist departments are an excellent basis for requirements discussions because they show what is really needed. What is crucial is a defined handover point at which a prototype is transferred into a maintainable solution that is operated in compliance with data protection and integrated into TYPO3 or the existing system landscape. Anyone who clearly names this point gains speed without the silent accumulation of systems for which no one feels responsible later on.