Low-code Automation tools such as n8n, Zapier, and Make have gained tremendous popularity lately. Agencies offering automation, or even “AI Automation,” seem to be springing up everywhere.
Having been traditionally skeptical of low-code tools, I wondered what was actually driving the hype, what the distractions were, and what made sense to me.
While the use of Artificial Intelligence agents greatly helps with the design of workflows, and while the integration of AI agents into the workflow allows for data extraction and content creation features that were hard to implement, if not unthinkable, just a few years ago, I think it is not the use of AI that is the killer feature we need here. I believe it is the availability and usability of the tool itself.
The Platform as an Enabler
Designing and reliably running asynchronous, long-running workflows that integrate with existing services is quite non-trivial in general. Rightly so. As noted in other posts, achieving state management, robustness under failure, retry handling, and monitoring is not simple.
Although developers have always had access to libraries to build these features, the effort required to maintain a custom environment for “simple” integrations was often disproportionate to the value.
Tools like n8n provide a user-friendly,feature-rich and managed platform to design and execute logic across high-level business functions.
Such platforms have only recently become widely available. The availability of business application exposure via REST or SOAP web services may soon reach a level such that integration scenarios across business systems are no longer dependent on custom interface development.
A graphical representation of a workflow is inherently more comprehensible to a business stakeholder than a block of code. This transparency offers two distinct advantages:
Reduced Dependency: Stakeholders feel more confident when they can “see” the business logic, which makes the handover and maintenance of processes feel more manageable.
Collaboration: It allows tech-savvy business users and developers to “speak the same language” when discussing process improvements.
A Pragmatic Target Architecture
Adding the integration of collaboration tools, office systems, shared file storage and much more to the mix, we create a system architecture that enables high-level automation (or, even better,: orchestration) of tasks that were previously either too complex to implement or simply too much effort to develop and maintain, given the automation value:

For example, imagine a system that (for whatever business reasons) allows the preparation of production orders in a shop floor system:
- As a trigger, we use the reception of an email message of a given subject. We then extract basic facts such as lot data and dates via an LLM integration and store everything in a shared spreadsheet.
- Once that is done, we send an approval email with a link to the spreadsheet to another user – the approver.
- That user opens the spreadsheet at his discretion and checks for pending orders.
- Once approved, another workflow is triggered that invokes the shop-floor system to set up the production configuration.
- The status of the production order is monitored and updates are sent to the approver.
Limitations of Low-Code
While this approach looks great, graphical control flow modeling is very limited as a “programming” approach and the platforms we considered only hold up so far in terms of performance, reliability, and robustness.
Low-code tools are not ideal for critical automation control flows that require strong consistency and reliability. If a process runs many thousand times a day and requires manual intervention upon failure, you want more robustness than these tools can offer.
Low-code tools are not ideal for workloads involving high data loads, computation or transactions. Error handling and compensation are generally not that great and complex interactions may simply get too complex for graphical modeling as mentioned above.
Anything more complex than service orchestration belongs in the realm of specified, coded applications that are subject to a test-based quality assurance and a well-run DevOps process.
You do, however, get an “AI friendly Automation User Interface” essentially for free that connects your boring business application with the professional users that need to make the best of it!









