Businesses like predictability. AI agents alone won’t give you that

Photo by Christina @ wocintechchat.com on Unsplash

In his book “Atomic Habits”, author James Clear provides us with tools to build habits that aligns with our purpose in life: Get fitter, our dream job, find love… you name it.

One of the ideas that stick to me, and relates to fact that our human brain is built to save as much energy as we can by automating tasks. Habit building is then, in fact, an evolutionary mechanism of our brain to be as efficient as possible, but creating a pre-defined solution to existing problems that we do over and over again. We call them “habits”.

Get Sergio Sisternes’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

If we take this analogy from our own human background, and we explore how organisations are using GenAI today, especially when it comes to building AI Agents, we will then quickly spot a few interesting challenges:

1. We are basically overspending too much energy by not making agents into our “habit-building” feature. This is not only a financial issue, as most clients are using Public Clouds to build their Agentic solutions, but also from a sustainability perspective: We are using energy to think about how to approach a problem over and over.
2. The second challenge is, obviously, results are different each time. The outcome we might get from an agent dealing with the same question is not the same. And businesses like consistency. The largest and most successful companies in the world are built on top of systems. Identify patterns, apply processes, and reusability. Minimise the creativity and the amount of “grey matter”, the energy spent figuring out how to solve the problem, and “stick to the system books”.

It is essential, then, that we see Agents and Agentic AI not as a collection of Prompts, Tools and Data Sources plugged into an LLM model that is sucking out energy, water and your P&L, but as:

1. A platform to build systems at scale with LLMs
2. These systems, to leverage SLMs (Small and Specific Language Models), that are fit for specific purposes in the process
3. Build explainability and responsible AI right into the core of this platform, ensuring full transparency and traceability to get answers on the current behaviour and improve the system. This information can, of course, be fed back into our most powerful “System Design Agents” to rethink and enhance our system.

In consequence, if you are thinking about adopting Agentic AI, don’t forget to think how your System Design is, and how to create an effective integration with an Automation Platform.

Like the human habit-building process, perfected over millions of years of evolution, this will give you the predictability, savings and productivity that the business is looking for in GenAI.