Agentic AI - the latest evolution of this important technology - allows software to operate more independently.
The clue is in the name: agentic AI. Unlike generative AI, which primarily produces content based on user prompts, agentic AI can operate as an autonomous "agent", completing complex, multi-step tasks with minimal human intervention. This shift marks a transition from AI as a passive tool to AI as an agent capable of decision-making, action-taking, and driving outcomes independently.
What's so exciting here is the prospect of myriad real-life uses for this developing technology. It's much like moving from a smart digital assistant to an intelligent, goal-oriented self-starter that executes tasks end-to-end. The upside is clear: lower operating costs, faster execution, and new data-driven services. Companies that can harness agentic AI effectively look likely to profit through enhanced performance metrics.
Artificial intelligence is evolving faster than any other technology in history, reshaping industries and investment landscapes alike. Generative AI amazed the world, but its impact so far is just a hint of what is coming next.
Today, generative AI creates content like text, but needs user prompts and cannot operate independently. Although 80 % of companies use generative AI, many have seen only limited effects due to early-stage use and measurement challenges.
Agentic AI can be tailored to overcome specific challenges, but typically follows a sequence of fundamental steps:
Initial versions of agentic AI appeared in late 2024, in the form of smarter chatbots with deep search features. They were able to "think deeper", meaning they could solve problems using multi-step reasoning, and automate users' research processes. This year, several agentic AI products have been introduced that handle tasks even more independently, shifting AI from a responsive tool to an autonomous agent.
This seismic shift is disrupting traditional software businesses, as agentic AI can perform tasks like coding and customer service, sometimes replacing software applications.
The result? A potential challenge to traditional pricing and business models in the software industry. In response, software firms are developing and integrating AI-enhanced solutions into their offerings and increasingly, are increasingly adopting usage-based pricing models. While their installed user bases and unique data sets represent an advantage, competition from AI-native startups and tech giants is intensifying.
Agentic AI is ready to act - but humans must still decide where to draw the line
Agentic AI is expected to grow from handling focused tasks to managing complex, multi-faceted workflows, potentially transforming enterprise operations. Coresight Research, a global tech-focused research firm, estimates that 51 % of enterprise software will use AI agents by 2029. US-based Grandview Research, another tech-focused research company, expects the global AI agent market to grow from USD 5.4 billion in 2024 to a whopping USD 50.3 billion by 2030, which implies a compound annual growth rate of 45 %.
Certainly, agentic AI holds great promise. But equally, the development of this technology poses challenges that call for careful management. Perhaps most troubling to the general public is the question of autonomy versus oversight. Giving AI decision-making powers can feel reckless, so it's important to balance the loss of autonomy with human supervision to help avoid unwanted results, and ensure that ethical and legal standards are met.
It's also true that agentic AI's complex decision-making processes are often unclear, making it hard to judge the fairness of its outputs. Because of this, many customers will prefer human support, particularly for sensitive tasks in sectors such as healthcare and education. The use of sensitive data will also need to be closely monitored, given the risks associated with interconnected systems and cyberattacks.
Cost is an additional issue. Developing and running agentic AI is expensive because of the need to develop complex reasoning models, and to integrate tools and systems. Ongoing human oversight doesn't come cheap, either.
Finally, the rise of agentic AI could lead to significant job losses, raising social and economic concerns that must be addressed proactively.
Meeting these concerns will mean developing transparency frameworks, ethical guidelines, robust governance, and tailored deployment strategies to ensure safe and trusted usage as agentic AI matures.
All that said, the possibilities are almost limitless. Agentic AI's real-world use cases illustrate incredible versatility. The level of current adoption represents just the tip of the iceberg - with transformative impacts anticipated in every industry.
Take healthcare: from robots using agentic AI to assist in surgeries, to supporting disease diagnoses and interpreting medical images, this technology is already making a significant impact. Medical AI agents are analysing large datasets of patient histories, suggesting personalised treatment plans, and even helping to triage patients in emergency rooms. AI agents are also effective at automating administrative tasks like appointment scheduling and reminders, as well as prescription refills.
In finance, agentic AI can be found autonomously trading stocks, commodities, or cryptocurrencies by analysing vast datasets and market conditions to make real-time decisions. Unsurprisingly, agentic AI can enhance fraud detection using real-time monitoring of transactions, blocking threats before they escalate. Back-end operations like compliance checks and loan applications can be managed and optimised by agentic systems.
Self-driving cars use multi-agent AI to perceive the environment, make navigation decisions - entirely autonomously - and interact safely with other vehicles. Delivery drones and autonomous trucks use agentic AI for obstacle avoidance, real-time route optimisation, and mission completion, even in complex and fast-changing environments.
In the logistics industry, AI agents can autonomously manage and optimise inventory by predicting demand, placing supplier orders, and ensuring timely deliveries. Warehouse robots powered by agentic AI now coordinate package handling, sorting and shipping - without human oversight – which dramatically improves speed and accuracy.
And although cybersecurity is a concern in the operation of agentic AI, it is incredibly useful for detecting and halting cyberattacks by monitoring network traffic for vulnerabilities, phishing, or unauthorised access attempts.
Agentic AI is here today. It's being used increasingly widely and will only proliferate, offering opportunities for companies and investors alike. In the words of Andy Jassy, CEO of Amazon, "Many of these agents have yet to be built, but make no mistake, they're coming, and coming fast."
LGT's experts analyze global economic and market trends on an ongoing basis. Our publications on international financial markets, sectors and companies help you make informed investment decisions.
