The days of AI acting solely as a predictive tool or chatbot are numbered.
Armand Ruiz, IBM’s vice president of product management for its AI platform, told delegates at the SXSW Festival in Australia this week that AI agents will soon allow enterprises in the APAC region to automate complex, multi-step tasks, freeing employees to focus on more human-centric activities.
Ruiz explained that AI technologies have evolved from predictive models in traditional machine learning to the widespread use of chatbots. He predicted that the next leap will usher in an “agentic era,” where specialised AI agents collaborate with humans to drive organisational efficiencies.
“We have a long way to go to get AI to allow us to do all these routine tasks and do it in a way that is reliable, and then do it in a way that you can scale it, and then you can explain it, and you can monitor it,” Ruiz told the crowd. “But we’re going to get there, and we’re going to get there faster than we think.”
According to Ruiz, an AI agent is a system that can autonomously reason through complex problems, breaking down tasks, creating actionable plans, and executing those plans using a suite of tools. These agents exhibit advanced reasoning, memory retention, and the ability to execute tasks independently.
Ruiz identified four capabilities of AI agents: planning, memory, tools, and action.
AI agents are capable of advanced planning to address given tasks or prompts.
Self-reflection: Agents can self-reflect or check if their decisions make sense or not.
Self-criticism: Agents can use feedback, often from the same or different large language models, to critique and improve their plans.
SEE: Gartner’s top 10 strategic technology trends for 2025
Chain of thought: Agents can break down larger tasks into smaller steps to improve accuracy.
Sub-goal decomposition: They can also establish sub-goals by dividing larger tasks into manageable components.
AI agents leverage both short- and long-term memory to support their autonomous actions.
Short-term memory: This in-context memory enables agents to track actions within an existing session.
Long-term memory: AI agents can log past interactions, helping them learn from mistakes and continuously improve their performance over time.
AI agents will be connected to third-party tools to complete their tasks. With the right access and governance, they could leverage tools ranging from web search and code-generation platforms to enterprise systems, such as HR platforms, Microsoft Teams, CRM tools, cloud services, and data warehouses.
The true potential of AI agents lies in their ability to act autonomously on behalf of humans. Whether streamlining HR workflows such as recruiting, resolving software code issues, or tackling other enterprise challenges, these agents will transform AI from passive chatbot to proactive actors.
Enterprises are likely to have “millions of AI agents” working for them, Ruiz said. These agents, which will essentially act as coworkers or AI assistants for human employees, will be able to work collaboratively with one another on various tasks, enabling them to “solve problems end-to-end.”
Ruiz explained that AI agents can function as either single-step or multi-step systems, with their actions coordinated and guided by a Super AI.
One-step agents are those that can perform specific tasks or solve individual problems when prompted, executing them with the help of relevant tools. Tools are defined and the process still remains fairly manual, though these agents can access systems such as LLMs to produce results.
Ruiz warned that there may be times when these AI agents hallucinate or fail to work as well as desired.
Multi-step AI agents leverage iterative strategies in what Ruiz called a “thought, action, observation loop,” using one or more LLMs. “You have this loop that is very iterative, and it’s amazing how that improves the outcome, and provides better results until you get the final one,” he said.
Enterprises will deploy “Super AI” systems to coordinate networks of individual AI agents. According to Ruiz, these Super AIs will act as orchestrators, planning tasks, breaking them down into smaller components, and assigning them to the most suitable agents within the organsiation to complete the work efficiently.
“One AI agent might be very good at sales or product management or encoding, or very good at mainframe or a specific programming language. Each will have small language models that are very easy to train, very cheap to execute and they will have specific access to certain tools,” he said.
Ruiz identified three primary user groups likely to benefit from AI agents: developers, no-code business users, and end users.
Developers: Traditionally, AI, data science, and machine learning required highly specialised expertise. However, Ruiz explained that millions of developers now have access to these technologies via APIs. Additionally, frameworks like CrewAI allow developers to quickly build and deploy AI agents.
Business users: No-code tools will soon allow business users to build their own AI agents through a user interface. IBM’s new agent builder, set to debut at IBM’s TechXchange Conference, will empower employees across all levels of a business to create agents that can automate and perform organisational tasks without needing programming knowledge.
End users: A broad range of end users will also engage with AI agents, Ruiz said, noting that there will be “a whole spectrum” of end users adopting and using these tools in various ways.
Ruiz said factories were a good analogy for how work may transform. In the early 1900s, factories relied on manual, labor-intensive work by many people, which was very time-consuming and inefficient. However, upon the dawn of the industrial revolution, machines were introduced to help automate them and accelerate production.
He explained that AI is now evolving to help automate and augment mental work in the same way as machines automated physical labor in factories. Rather than being a replacement, he argues it will enable them to focus on more strategic and innovative tasks, improving overall productivity and efficiency.
“We’re seeing this already in marketing,” Ruiz added. “We’re going to see this in sales as well, and it’s going to start expanding throughout all the different job functions. Our goal is to get AI to free us from a lot of distractions and enable us to work on meaningful work and on human connections.
“The vision is for AI agents to work alongside humans in a complementary way, augmenting human capabilities rather than replacing human workers entirely. This will allow for greater productivity, work-life balance, and focus on higher-value activities.”
