Put simply, Agentic AI describes an AI system that operates with a high level of autonomy. It's a system that can decide and act on its own with no prompts or help.  AI's Large Language Models  require prompts and guidance - they do not act on their own without you guiding it. Not the case with Agentic AI systems - they analyse, plan and act independently. They are able to make decisions, complete actions without human oversight. An example: an AI agentic system in fashion could autonomously plan, forecast and order collections. All without having to supervise it. An AI Agentic customer service system could interact with your customers after purchase, enrol them in a personalised loyalty programme, send periodic followup emails to recommend new garments, new sizing and collect feedback. This AI Agent might then collect feedback and enrich your customers' data. Let’s contrast this with the AI chatbot of today. This is often a prompted, restricted dialogue that is rules based and limited. Not independent or autonomous and often frustrating! 

 

Low Agency & High Agency? AI agency in this context refers simply to the level autonomy. How much it can do on it's own. On the one hand, there are the current low agency systems - these are systems that are able to handle well defined tasks. Returning to the customer service agent example, we are already seeing AI customer service agents able to take orders and respond to open questions relatively autonomously. High agency systems go further learning from their environment, making decisions and performing tasks independently. This is where we are heading. 

 

So what’s it about? In a word: productivity. AI agents do not need to sleep! They handle large amounts of data really well. And fast. An AI agent could analyse your collection's sales performances across all your markets faster than you could whilst suggesting insights and actions, or already acting on them if you gave it permission to do so. So the tasks that are most suited are those that require automation, decision-making, and optimisation based on data. They excel in environments where they need to handle repetitive, complex, or data-driven tasks. 

 

We are not there yet.  There are challenges such as governance, there are questions over their trustworthiness as well as likely employee resistance.  Tools, and guardrails are required to ensure that Agentic AI is effective, helpful and productive. 

 

It is coming, however. And fast. By 2028, Gartner estimates 33% of enterprise software applications will include agentic AI, up from less than 1% in 2024, enabling 15% of day-to-day work decisions to be made autonomously.

 

Companies often overestimate the impact of short-term changes in technology and underestimate the effect of long-term changes. 

 

I’m not sure this is entirely the case just yet…. Judging by the acceleration of learning, results and investment in this space, we will see short term changes very quickly in addition to profound long term impacts. 

 

That’s it for this week.