Agentic AI: Next big leap in workplace automation

The next time a customer calls a healthcare helpline or seeks assistance for a delayed food delivery, they may unknowingly be speaking with an artificial intelligence system rather than a human representative, as corporations rapidly adopt AI agent technology.

Major technology companies including Zoom, Adobe, Microsoft, Cisco and Atlassian are making a significant pivot towards “agentic AI” — systems that can autonomously work towards goals with human-like intelligence.

“AI workers with human level intelligence,” is how Ram Menon, CEO and co-founder of Indian tech company Avaamo, describes this evolution. According to Menon, agentic AI operates with proactiveness at its core, driven by autonomous capabilities to “perceive, reason, act and learn” towards prescribed outcomes.

Industry developments showcase this trend’s momentum. Adobe has developed 10 purpose-built AI agents for enterprises, while Zoom has incorporated agentic capabilities into its platform. Microsoft has created Sales Agents designed to research leads and contact customers on behalf of businesses. Nvidia is collaborating with Wipro to develop agentic AI for government applications.

These developments represent what experts view as artificial intelligence’s third wave, following predictive AI and generative AI. This latest iteration utilises sophisticated reasoning and planning to autonomously solve complex, multi-step problems — essentially replicating skills that could replace some human office functions.

“Agentic AI is a major leap that will accelerate workplace transformation,” said Anjul Bhambhri, senior vice president of engineering at Adobe Experience Cloud. Their AI agents’ capabilities include content production, data engineering, experimentation and workflow optimisation.

The effectiveness of agentic AI depends on its training methodology, which typically incorporates large language models (LLMs), small language models (SLMs), reinforcement learning models, transformer architectures, generative adversarial networks and probabilistic reasoning.

China’s latest AI breakthrough, Butterfly Effect’s Manus AI agent, utilises existing models including Anthropic’s Claude and fine-tuned versions of Alibaba’s open-source Qwen, according to Peak Ji Yichao, the company’s co-founder and chief scientist.

Currently, agentic AI development focuses on enterprise applications and in-office workflows for individuals.

Smita Hashim, chief product officer at Zoom, identifies four core skills — reasoning, memory, task action and orchestration — that define their platform’s upgraded AI capabilities.

“Think of AI Companion as the second set of hands you’ve always wanted. An AI agent who knows you and operates with your guidance, using information across the Zoom platform, third parties, and the web,” Hashim said. The system can now schedule meetings, summarise missed ones, track team projects, draft communications and extract details from chats.

Cisco’s new Webex AI Agent aims to enhance organisational support services by working alongside humans to address “routine and high-volume customer questions and executes actions to fulfil customer requests,” with emphasis on natural conversational capabilities and real-time solutions.

“Enterprises are starting to realise the potential of agentic AI. It is reinventing what it means for people and technology to work together,” said Jeetu Patel, executive vice president and chief product officer at Cisco.

These agents are expected to extend beyond customer support to healthcare management, data analysis, sales, personal assistance, content creation, research and cybersecurity monitoring, though their effectiveness remains to be fully demonstrated.

Nvidia and Capgemini are developing 100 bespoke AI agents for automotive, financial services, life sciences, telecommunications and manufacturing industries.

Indian food delivery start-up Zomato has already seen benefits from their in-house agentic AI tool called Nugget, which handles queries via live chat and natural voice calls, streamlines ticketing and conducts quality audits.

“Built over 3 years as an internal tool, Nugget now powers 15 million support interactions per month for Zomato, Blinkit and Hyperpure,” explained founder and CEO Deepinder Goyal. Recently, Zomato Labs made Nugget available to businesses worldwide, offering it free in some cases for the remainder of existing agentic AI service contracts.

Companies developing or implementing agentic AI consistently highlight productivity gains, time savings and creative potential. Adobe emphasises “driving productivity gains for practitioners to free up time for creative ideation,” whilst Capgemini states that agentic AI enables clients “to solve complex business challenges and innovate at scale.”

However, a significant question remains unanswered: what ultimately happens to the human employees currently performing these tasks?

“The use case we are creating is a digital worker that displays human level intelligence to replace a function or a role,” said Avaamo’s Menon. Their Ava agentic AI can operate within healthcare systems, potentially replacing thousands of workers who currently guide callers to patient coordinators, provide doctor availability information and navigate insurance matters. “What we have built is an AI clone that replicates that role,” he added.

Molly Sands, head of the Teamwork Lab at Atlassian, offers a more optimistic perspective.

“Knowledge workers often juggle dozens of projects and a range of complex deliverables at once. This can quickly feel overwhelming. AI collaboration helps you stay on top of tasks and reclaim time,” she said, adding: “Our research shows 96% of the most mature AI collaborators in India deliver work on time, compared to just 91% of less experienced users.”

The current climate emphasises the need to perfect AI usage, though achieving this may prove challenging.