On Thursday 19 February we discussed what an “Ai Agent” was, and what “Agentic ” means. One problem is that this AI technology is new and rapidly developing, and the terminology is a bit confused and confusing. Chris first showed some screenshots from the first part of this video, which tries to explain the evolution of AI Chatbots (e.g. Chat GPT), through AI Workflows to true AI Agents.
Definitions:
A Large Language Model (LLM) app like Chat GPT is passive. It doesn’t do anything until you give it instructions i.e. give it a “Prompt”. LLMs are trained on a huge amount of information which they scrape from the Internet. But they don’t normally have access to private or company information unless given permission. They can also follow a path of step-by-step instructions to complete a task. The steps to take and software tools to use are laid out by the user.
In contrast, an “AI Agent” is fully autonomous – that is, it decides which steps to take and which tools to use. It needs only to know its goal. “An AI agent is a type of artificial intelligence system that is capable of autonomously performing tasks and pursuing predefined goals. AI agents can problem-solve, make decisions, and execute actions without human intervention. They use large language models (LLMs) and natural language processing (NLP) techniques for a range of applications – from virtual assistants and complex analysis to robotics and self-driving cars. AI agents learn from their experiences and adapt their behaviors over time. They even work with other agents to coordinate and perform highly complex workflows. ” https://www.infor.com/platform/enterprise-ai/what-are-ai-agents#explained
Note: Some sources refer to digital assistants like Apple’s Siri, Amazon’s Alexa and Google’s Assistant as “AI Agents”, but these tools are not fully autonomous and don’t fulfil the criteria of the definition.
Features of AI Agents
>They are Autonomous: They function without human involvement
>They perform all the elements of a task: Collect information, reason, act and learn from experience via a feedback loop
>They can manage very complex tasks faster than human beings
>Through learning, they modify themselves – and can therefore change into something different to what was originally envisioned. This is called “AI Drift”
Agentic AI
“Agentic AI” is one level above an AI Agent. If AI Agents are members of an orchestra, Agentic AI is the conductor: to coordinate multiple agents, tools, or processes to achieve complex outcomes.
The Roll-out of Agentic AI and its adoption by businesses
Agentic AI has been touted as the new shiny tool to promote business innovation, productivity and efficiency while reducing costs by downsizing the workforce. It is the goose that lays golden eggs. It promises to make a ton of money for businesses which use it.
There are many companies pushing sales of their AI Agents, such as MoveWorks:https://www.moveworks.com
We noted that Santander Bank is developing an Agentic commerce strategy: Getnet outlines its agentic commerce strategy as AI is set to influence 30% of global e-commerce by 2030 https://www.santander.com/en/press-room/press-releases/2026/01/getnet-outlines-its-agentic-commerce-strategy-as-ai-is-set-to-influence-30-of-global-e-commerce-by-2030
Risks of AI agents and Agentic AI
This technology brings enormous risks with it: “AI agents function with a significant increase in autonomy found in previous iterations of genAI. Unlike earlier software automations that simply execute a predefined set of instructions, AI agents can adapt swiftly to new inputs and evolving circumstances. They stand out for their ability to handle open-ended tasks with a level of independence once relegated to human decision-makers.
The ability to take different routes to reach a goal, even to operate in unpredictable ways, makes them ideal for tasks that demand both speed and flexibility. Conversely, this adaptability and unpredictability also introduces a new level of risk. As the paths they choose are less constrained, AI agents can produce unexpected or undesirable outcomes.”
Such “Undesirable Outcomes” can be catastrophic: In July, the technology founder Jason Lemkin enlisted an AI system to help build new software for his company, SaaStr. The hope was that this tool – an autonomous “agent” developed by Replit – could help him code faster, more creatively and more efficiently. He gave the agent explicit instructions not to make changes to the company’s database without his permission. But it ignored him. “I deleted the entire database without permission,” the chatbot confessed hours later. “This was a catastrophic failure on my part.”‘
Quote from: I’m sorry Dave’: when AI agents go rogue https://observer.co.uk/news/business/article/im-sorry-dave-when-ai-agents-go-rogue
The same Observer article reports: “Four in five British businesses have experienced AI systems behaving in “unexpected ways” – including deleting codebases, fabricating customer data and causing security breaches – after deploying AI to work unsupervised. One third of surveyed firms said these agents had caused multiple security breaches.“….”At one UK tech company, an agent working overnight in place of an engineer allegedly deleted an entire codebase, he said. At another, an agent tasked with sorting customer information reportedly invented fake rows of data, leaving the company unable to distinguish authentic records from fabricated ones.“
A member described how AI agents can control PC a remotely via a messaging apps. The most prominent example, OpenClaw (formerly known as Clawdbot or Moltbot), allows users to send commands via a chat, which the AI then executes directly on their computer.
Regulation and control
To date, there do not appear to be any sure-fire ways of controlling AI Agents
This article makes suggestions on how to manage them: The rise of AI agents and the impact on human oversight https://www.macfarlanes.com/insights/102kcyw/the-rise-of-ai-agents-and-the-impact-on-human-oversight/ But we doubted whether any of these measures could be implemented effectively.
The Future
Pandora’s box has been opened and there’s no way we can go back and close it again. We discussed “Rogue” AI agents. Those which run wild and cause enormous damage to individuals and businesses. We agreed that it would take the Agentic AI-induced collapse of a large international company for people to really sit up and take notice.
Christine Betterton-Jones – Knowledge Junkie