Agentic AI for dummies: 101 on how marketers can leverage on the trend

Artificial Intelligence has come a long way, and marketers have strived to adapt to its rapid evolution. Now, with the rise of agentic AI, we're entering a new era where AI doesn't just analyse or respond, it can act with autonomy, make decisions, and execute multi-step campaigns on behalf of human marketers. This shift opens the door to unprecedented efficiency, creativity, and strategic precision. 

According to a 2025 Salesforce report, 75% of retailers say AI agents will be essential to help businesses stay competitive by 2026. Meanwhile, 67% of retailers believe autonomous AI will bring opportunities while 81% of retailers trust AI to act autonomously, with sufficient guardrails. 

Global HR leaders also believe digital labour is the future and its integration is critical to their role, with 80% believing that within five years, most workforces will have humans and AI agents working together, according to another recent Salesforce survey. Another 86% of CHROs say that integrating digital labour alongside their existing workforce will be a critical part of their job. 

With AI transforming how brands engage and learn from customers, marketers face new pressures and new possibilities.

What is agentic AI? 

Agentic AI refers to a new class of artificial intelligence systems that possess the ability to act with agency, meaning they can set goals, make decisions, and take autonomous actions to achieve those goals. Unlike traditional AI, which typically waits for a prompt and responds passively, agentic AI can proactively initiate tasks, break down complex objectives into manageable steps, and adapt its behaviour based on feedback or changing conditions. 

Slightly different from AI agents which can plan, use tools, interact with APIs, and repeat steps until a task is done, agentic AI has memory, can manage goals, adapt to feedback, and collaborate with other agents, according to Henson Tsai, founder and CEO, SleekFlow.  

“It’s not just task-driven, it’s outcome-driven. Instead of ‘Improve writing’ you’d say ‘research 50 articles, extract trends, update my workspace, and alert me if anything’s urgent.’ We're moving from apps to autonomous workflows and it's a big shift,” he added. 

Key players such as AWS, Google Cloud, IBM, Microsoft, and NVIDIA offer definitions emphasising this ability to understand goals, interact with environments (often through tools and APIs), and act autonomously, Jacky Chan, CTO, Votee AI, said.  

In marketing and retail, agentic AI is an autonomous, AI-powered system that assists customers with complex tasks such as checking live inventory, seeking professional advice, booking personalised appointments, or resolving intricate inquiries, according to Samson Fong, regional head of marketing, Beame. “Unlike traditional chatbots, it adapts dynamically to each customer’s needs, guiding them toward a booking or purchase.” 

Which industries need agentic AI the most? 

Agentic AI is vital for industries such as healthcare, hospitality, and financial services, where customer engagement exceeds simple FAQs or transactions, added Fong. “For instance, a Hong Kong hair implant centre might receive inquiries about surgery details, costs, safety, and scheduling. He added:

An AI agent can address these autonomously, escalating to humans only for niche questions, delivering a seamless experience indistinguishable from human support.

Similarly, in banking, AI can guide customers through loan applications or account management, tailored to their needs, he added. 

On the other hand, in APAC’s fiercely competitive food delivery sector, where customer loyalty hinges on speed and personalisation, agentic AI is the unseen force keeping customers engaged and businesses ahead of the curve, according to Eva Swainston, senior manager, APAC lifecycle growth, foodpanda.  

This is due to the industry’s operational necessity, including high transaction volumes and razor-thin margins which require next-level efficiency in customer service and retention, she said. “Hong Kong’s deal-savvy consumers expect real-time, tailored promotions, AI delivers dynamic offers that manual or rule-based processes can’t match.” 

On the agency front, using agentic AI internally can create content aligned with brand guidelines and schedule posts at optimal times, Tsai said. “The engineering team can leverage the coding agent for code completion, bug detection, and code generation. The admin AI agent can review documents and receipts to approve or reject claims. These applications enhance organisational efficiency and are particularly valuable for small teams.” 

For external use cases, client-facing industries with high volumes of online communication urgently need agentic AI such as financial services, eCommerce, and insurance services, he said. “These sectors manage repetitive yet critical interactions, making them ideal for agentic AI to enhance customer conversion and retention.” 

Opportunities and challenges 

While the potential for agentic AI is vast across numerous sectors, marketing and sales also benefit greatly by automating complex campaign workflows, leading to significant gains in efficiency and speed, according to Votee AI’s Chan. “This automation reduces the burden of operational tasks on human employees, freeing them to focus on higher-level strategy, creativity, and oversight.” 

However, adopting agentic AI also comes with significant limitations and challenges. The development, configuration, and integration of these sophisticated systems can be complex and costly, Chan added. “Its effectiveness heavily depends on clearly defined goals; ambiguity can lead to errors or unintended consequences. As autonomous entities, agents introduce risks related to safety, security, and control – flawed decisions stemming from biased data or faulty reasoning can have real-world impacts.” 

As such, ensuring ethical operation, accountability, and compliance with standards such as ISO/IEC 42001 becomes paramount, Chan said. “Significant technical hurdles exist, particularly in enabling effective collaboration between agents and ensuring coherent understanding through shared context.” 

SleekFlow’s Tsai said agentic AI often requires significant development and integration efforts, as the system performance is highly reliant on the quality and availability of data. It also raises questions around data privacy and decision-making transparency, he said. 

Industries with strict regulations or legacy systems may face hurdles in adopting agentic AI. Oftentimes, transformation is hard to get internal stakeholders’ buy-in.

How will the agentic AI trend evolve? 

Back in July 2024, OpenAI outlined a framework, known as the "five levels of super AI", which delineates the stages of AI development from basic functionalities to surpassing human capabilities. Currently, the world is moving from level two to level three, SleekFlow’s Tsai, said. 

The evolution of agentic AI is moving towards creating autonomous workflows rather than isolated task automation, he said, adding that moving forward, the world will see closer collaboration between agents, meaning agentic AI systems will work together to achieve complex goals, enhancing efficiency across industries. 

“In five years, I believe every company will have an AI hub that allows AI agents to do work that connects to various systems. At SleekFlow, we believe this is the future of work, and that’s why we are committed to building AI agents that not only respond to queries but also take meaningful actions. The goal is always to drive business growth,” he added. 

As customers embrace instant AI assistance via a chat-based interface, they’ll expect brands to deliver similar experiences, Beame’s Fong, said. “AI agents will handle complex request chains, such as bank validating transactions, waiving annual fees and setting up auto-pay in one go, or an airline planning and booking itineraries based on past trips, preferences, price trends or upcoming holidays.” 

Advancements in AI training will drive this shift over the next three to five years, so brands must ensure trust and data security to meet rising expectations.

While complex now, Votee AI's Chan anticipated the democratisation of agentic AI through platforms and tools that simplify development and deployment for businesses and perhaps even individuals. “This ties into the potential shift in business models – moving from ‘Software-as-a-Service' (SaaS) towards ‘Service-as-a-Software', where agents bundle functionalities previously requiring multiple SaaS products or services, potentially disrupting traditional software and service delivery.” 

Meanwhile, foodpanda’s Swainston said next-gen chatbots will utilise vocal and visual cues to detect customer sentiment and dynamically switch from upsell or promotional to a supportive mode such as abandoning promotional offers if a user sounds distressed.  

Join us this coming 17 June for #Content360 Hong Kong, an insightful one-day event centered around responsible AI, creativity VS influencers, Xiaohongshu and more. Let's dive into the art of curating content with creativity, critical thinking and confidence!

Related articles:

Agentic AI and the CX reset: The tech changing how brands serve and retain customers