01. AI is not the solution
Many projects now begin with a familiar statement: We want to use AI. It sounds modern and ambitious. From an engineering perspective, however, it represents the same mistake as starting a project with: We need a new application.
In both cases, the tool is selected before the problem has been understood. Artificial intelligence is not the objective of a project. It is one possible way of solving a specific problem.
02. Not every problem requires a model
If a system needs to compare values against limits, calculate averages, validate schedules, generate reports or detect threshold violations, traditional algorithms are usually the better solution.
They are simpler, faster, easier to maintain and far more predictable. Adding AI to these tasks rarely increases the value of the system. More often, it simply increases complexity.
03. Where AI genuinely creates value
Some problems cannot be described by a simple collection of rules. They involve recognising patterns, interpreting images, classifying objects, estimating quality or predicting future behaviour from incomplete information.
These are the situations where artificial intelligence becomes useful. Not because it replaces engineers, but because it supports decisions where traditional algorithms reach their practical limits.
In engineering projects, AI should therefore solve problems that are difficult to describe explicitly rather than problems that are already well understood.
04. Process first, AI second
The obvious question is often: How can we use AI here?
A better question is: Which operational decision is currently too slow, too uncertain or too difficult to make?
If AI shortens the path to that decision, it deserves consideration. If it does not improve the quality or speed of decision-making, it probably does not belong in the architecture.
Process │ ▼ Data │ ▼ Deterministic logic │ ▼ AI (only where necessary) │ ▼ Decision │ ▼ Action
In well-designed systems, AI is never the centre of the architecture. It is one carefully selected component among many others.
05. Less AI often means a better system
Surprisingly, many successful projects end up using less AI than originally expected. Not because the technology is weak, but because most operational processes can already be described using clear, deterministic logic.
Artificial intelligence remains where deterministic rules stop being sufficient. This approach simplifies maintenance, improves reliability and keeps AI focused on the problems where it truly provides an advantage.
06. A principle that applies everywhere
This way of thinking is not limited to manufacturing, automation or computer vision. The same principle applies to logistics, healthcare, finance, administration and software engineering.
First understand the process. Then identify the decision that should become easier. Only afterwards choose the appropriate technology.
Sometimes that technology will be AI. Sometimes it will be a straightforward algorithm. Sometimes the best solution is simply improving the process itself.
07. Good technology becomes invisible
The best technical solutions rarely attract attention. Users do not think about models, frameworks or architectures. They simply make better decisions, more quickly and with greater confidence.
That should always be the objective. Not building an AI system. Building a system that solves the problem more effectively.
LOOKAS Perspective
AI does not replace engineering. Good engineering reveals where AI truly belongs.