Computer says, ‘YES!’. The rise of Agentic AI

Chatbots are evolving, drawing on advanced AI capabilities to become AI Agents – fully autonomous, sophisticated systems capable of advanced learning, contextual analysis and complex problem solving. So, what is Agentic AI and where has it come from?

Chatbot or chat-not

“What brings you to our website today?”

It’s a common question to be asked by a (usually) faceless entity sitting in the corner of a web page. Sometimes these things are unobtrusive, sometimes they are in your face and get in the way of your browsing. Sometimes they give you the answer you want, sometimes they can’t even understand what you’ve asked them. In some instances they can leave you screaming at the screen: “I just want to talk to a human!”

These chatbots have been with us for some time, starting out as rule-based systems with pre-programmed responses. The integration of neurolinguistic programming (NLP) helped with the understanding of user intent and allowed the chatbot to provide more human-like responses.

Hands up! We have a chatbot on the DeeperThanBlue website. It varies in functionality on different pages, in some instances drilling down into the problem you are trying to solve during your visit to our site. We admit it’s not perfect, but it might help you to find the information you want.

Hey Siri/Alexa/Cortana/Bixby/Google! What’s new?

AI Assistants have wormed their way into everyday life with hyperscalers and software developers creating their own flavours of interactive (even proactive) tools that you can summon to help you. With a cry of ‘Alexa’, you can create your grocery shopping list without a pen and paper, plan and book a holiday without trawling the internet or create a playlist to satisfy whatever mood you are in.

These assistants are pervasive, even invasive at times, waiting for your commands and willing to do your bidding.

This interaction is possible due to the integration of machine learning and advanced NLP, which allows the assistant to handle more complex tasks and requests. The assistance is more personalised and interaction more conversational, offering services like diary scheduling, information retrieval and integration with an individual’s smart devices.

Agentic AI: Autonomous Systems

The latest stage of evolution has seen the incorporation of autonomy and goal-orientated behaviour. Agents make decisions, adapt to dynamic environments and perform tasks without human intervention or explicit instructions.

AI agents will use real-time analysis of user data (such as browsing history, preferences and behavioural patterns) to predict the needs of the user and offer tailored answers and solutions before the user explicitly asks. They can also pick up on contextual cues, including intent and emotions and nuanced instructions to provide empathetic and personalised responses.

The ability to continuously learn during interactions lets agents refine their responses and adapt to changing conditions or complex scenarios.

Ethical considerations for Agentic AI

For some people, the thought of AI systems making autonomous decisions and becoming integrated into society will set alarm bells ringing, with people of a certain age having flashbacks to the 1983 film WarGames. Or maybe the prospect will instil a fear of always being watched or listened to, like Orwell’s dystopian novel, 1984.

Concerns will range from AI Agents unintentionally perpetuate biases present in their training data, to potential misuse of user data, and many more.

For this reason, transparency and accountability are key factors in the development of AI Agents, with developers conforming to data governance and regulatory frameworks.

Across the world, governments are introducing frameworks to address these concerns. These frameworks include:

• The EU AI Act, which emphasises risk-based regulation for high-risk applications.
• The US AI Bill of Rights, which protects against algorithmic discrimination and ensures privacy rights.

Of course, since the technology itself is still quite new, yet evolving at a rapid rate, these frameworks are likely to be playing catch-up for a long time.

Deploying Agentic AI where it makes a positive difference.

AI Agents aren’t the harbingers of doom. No. They are already being put to work for good.

For example. In the US, Federal Bank has an AI Agent powered by Microsoft Azure OpenAI Service that autonomously handles 90% of customer queries across multiple channels, improving customer satisfaction by 30%. (This and hundreds of other Microsoft AI case studies can be found in ‘How real-world businesses are transforming with AI’ on the Microsoft website.)

At IBM, their Agent Lab allows developers to create custom AI agents using the LangGraph framework. These agents can perform tasks like weather queries or research paper retrieval and are deployed on Watsonx.ai for scalable enterprise use. IBM Watsonx Orchestrate is their commercial release of this application. Available as a SaaS offering or on premise, IBM has developed a number of business applications such as Ask HR, Ask Procurement, Ask Sales, and Ask IT with the intention of making people more efficient and removing mundane tasks.

Google Cloud Contact Center AI (Contact Centre as a Service—CCaaS) provides advanced virtual agents for customer service automation, handling complex inquiries with high accuracy.

Agentic AI isn’t just for enterprise businesses, though. SMEs can get on board too. Agents can be used to:

• Handle FAQs, booking appointments, or order tracking in customer service scenarios
• Monitor stock levels, predict demand trends, and automate reordering processes
• Create and schedule social media posts, email campaigns, or product descriptions for marketing automation.
• Analyse sales trends and customer behaviour to make data-driven decisions without the need for advanced technical expertise.

• Share