Automation, LLMs & Agents

Dec 7, 2025
AI Automation Innovation

AI is spearheading the tech industry. All we hear is ‘AI-this & AI-that’, AEO, GEO, and so on. But one term that really stands out is AI-enhanced workflows.

Early days of worflow automation

Considering that early concepts of workflow automation were pioneered in min 1900s, but first digital automations only took hold in mid 1990s, followed by modern automation tools in early 20th century.

Early to mid-2010

The mid-2010s, tools like Zapier and Microsoft Flow gave user ability to streamline task execution across different platforms. But those tools still primarily depended on human-defined rules.

2020s, the ‘Era of AI’

Introduction of generative AI, and ‘agent AI’ in 2020s, has revolutionised workflow automation. It is no longer a rigid, rule-driven process. We can now create, some-what intelligent, adaptive, reasoning-capable and self-improving systems.

With choices of automation tools, generative AI and agents, how do you select the right tools to build your workflow?

For automation there are rules. ‘If this…, then that…’. It is predictable and does the same thing most of the time. If tasks are straightforward and no human thinking is required, automation is your best bet.

LLMs are required if the goal is to generate something new. This is based on evaluation of input and requirements for output/result (text, code, transcriptions, media, etc.).

AI Agents combine LLM core, memory (short-term and long-term) and tools (eg. APIs) and multi-step planning. Agents work great when autonomous decision-making is required and the path is not predictable.

It is common to connect multiple agents to perform different tasks (eg. research, comparison, make decision, notifications, check performance, or even check work of another agent).

To sum it up (very top-level):
  • Automation is ideal for a predictable, step-based process
  • LLMs are great for task that require understanding input and generating content
  • Agents are well suited multi-step task execution, autonomous decision-making, tool usage or coordination about multiple systems.

In modern era AI-enchanted workflow automation is no longer a ‘luxury’, it is a ‘necessity’.

Back to blog