AI and Automation

For years, the road to AI (let’s call them Generation 1 and 2) ran mainly through universities. But with the internet boom, the third generation of AI became accessible to everyone. Suddenly, knowledge was no longer locked behind academic walls.

Today, we see a flood of tutorials on YouTube and online learning platforms offering “AI in a few hours” — sometimes even “AI in minutes with no code.” But is that really possible?

Generative AI (G-AI) is a powerful tool, but it often behaves in ways that make automation unreliable, costly, and difficult to scale. "After three months of experimentation with multiple G-AI engines, our team at DesignAndBusiness developed State AI (S-AI) — a framework designed to mitigate these problems by introducing state-driven workflow control", says Arash Kharabi, who directed the project.

Have you tried using AI to solve a problem, only to end up with unpredictable or inaccurate results? One key reason is that we often rely on generative AI (G-AI), which produces outputs that can vary significantly over time. Today’s result might differ tomorrow, even with the same input. This is the nature of G-AI — it’s dynamic, evolving, and influenced by continuous training, feedback, and platform updates.