Agentic AI refers to AI systems that can plan, make decisions, interact, use tools, and complete multi-step tasks with limited human direction or, in some cases, no direct human intervention.

It is gaining attention because of its potential to automate knowledge work, improve productivity, support research, manage workflows, detect anomalies, collaborate with groups of researchers, and assist in areas such as education, mathematical problem-solving, healthcare administration, customer service, and software development.

However, the hype around agentic AI often exceeds its current capabilities. These systems can still make mistakes, hallucinate, misunderstand goals, act on poor information, become vulnerable to prompt injection or external manipulation, and overstate their own reliability.

This gap between promise and performance creates serious risks when they are given too much autonomy. Ethics is central to agentic AI because these systems may act on behalf of people or organisations.

Key issues include accountability, privacy, transparency, bias, consent, and human control, and strong safeguards, audit trails, clear limits, and human oversight are needed to prevent harm and reduce over-reliance.

The main challenges include reliability, cybersecurity, regulatory compliance, data quality, public trust, and the risk of trusting these systems without adequate safeguards. Used responsibly, agentic AI can support human workers, reduce repetitive tasks, improve decision-making, and make complex processes more efficient, but only when those challenges are addressed.

This discussion will examine how agentic AI is already present in our lives and why adopting an ethical framework is essential to improve outcomes, augment human decision-making, and protect human rights.