Kyndryl’s Ismail Amla discusses the company’s new agentic AI framework and why human oversight is still vital to this advanced technology.
One of the hot topics of the artificial intelligence (AI) domain at the moment is agentic AI.
Agentic AI differs from generative AI – which responds to an end user query using natural language processing to generate an output – by using machine learning and reinforcement learning to solve complex, often multi-step problems. Some of the oft-praised benefits of this tech is its adaptability and advanced reasoning capabilities.
The introduction of agentic AI is no small event, with some even considering the innovation to be the third stage or wave of AI technology – the first being predictive AI and the second being generative AI.
As was the case with generative AI, companies are now looking at how they can utilise agentic AI to streamline operations. But, as seen with previous reports on AI integration, companies may need assistance with adopting the technology.
Last month, enterprise technology service provider Kyndryl its own agentic AI framework which companies can utilise to deploy a portfolio of specialised and self-learning AI agents across operations – whether on-premises, in the cloud or in a hybrid IT setting.
According to the company, the Kyndryl Agentic AI Framework combines advanced algorithms, optimisation and AI agents that translate complex data into business insights.
“Kyndryl has developed an agentic AI approach that’s enterprise-grade from the ground up,” says Ismail Amla, senior VP of Kyndryl Consult. “It’s secure by design, interpretable by default, and always keeps humans in the loop.
“The Kyndryl Agentic AI Framework deploys intelligent agents that learn, adapt, and evolve, responding to human input and turning insights into real action across complex IT environments, making this AI that works with you.”
Security and human insight
One of the main talking points of the framework has been how its agents are “secure by design”. But what does that mean?
“The solution is built with security in mind from the ground up, rather than being bolted on after the fact,” explains Amla to SiliconRepublic.com. He tells us that the framework has been developed with zero-trust security, “privacy-by-design principles” and full traceability of the agents’ actions.
“Every agent is traceable and accountable. This ensures that the solutions are secure by design.”
With agentic AI being considered the next big step in the artificial intelligence evolution, many may expect the threat to human employees to be intensified. However, Amla believes that with Kyndryl’s new framework, the concept of the ‘human in the loop’ is still vital.
“The future of work isn’t AI agents versus humans,” he says. “It’s AI agents working with human oversight.” In a previous Q&A on the framework, Amla asserted that agentic AI can be incorporated as “digital teammates”.
“Trust is the foundation,” he tells us, adding that this is reflected in the framework’s emphasis on traceability.
“Building in full traceability of agent actions means that AI is deployed responsibly with human oversight and continuous improvement.”
A gap, an opportunity
In a report released in June, Kyndryl found that only four in 10 business leaders felt that their organisation has fully implemented an overall AI adoption strategy to take them from their current state to a future state – a gap that Amla says Kyndryl is focused on tackling.
Currently, Kyndryl’s framework is already being leveraged by two unidentified entities – a national government and a financial institution.
According to the company, the national government in question is evaluating the framework for a large-scale deployment across their IT estate, with the hopes of enhancing citizen experiences, improving public service, ensuring compliance and reducing regulatory risks.
Meanwhile, the unnamed financial institution is assessing the framework’s potential for automating compliance, optimising IT and accelerating service delivery.
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.