“We have created an approach to implementing agent AI in a secure, enterprise-grade environment. “We have created an approach to implementing agent AI in a safe, enterprise-grade environment. Ismail Amla, Senior Vice President of Kyndryl Consult, said:
Global Enterprise Technology Services Provider Kyndryl has introduced the Kyndryl Agent AI Framework. It aims to facilitate the deployment of Agent AI and work alongside human teams.
Kyndryl's and IT Industry's investment in Agent AI has followed Genai's widespread adoption, said Ismail Amla, senior vice president at Kyndryl Consult.
“genai was great for getting your content,” Amla told CRN. “All of a sudden we got all the content, not just the content, but all the content, reasoning, learning, acting, all with our agents.
[Related: Kyndryl CEO: Reselling ‘Other People’s Equipment’ Was ‘Empty Calorie Revenue’]
The investment in AI by multiple vendors of technology, including large-scale language models, inference engines and frameworks, is creating those opportunities, Alma said.
The big problem with the industry is that it means everything like customer service, HR, he said. It leads to the question of how to leverage agent AI in enterprise-grade workloads, not in the world that “showed the demos,” he said.
“Kyndryl understands the meaning of every opportunity people talk about for customers of all the technology currently available, and all of the hype was around Genai. “There were a lot of hype and demos of them, but many weren't implemented.”
Kyndryl's response is the launch of the Kyndryl Agent AI Framework, which aims to facilitate the deployment of Agent AI to human teams, Alma said.
“We've created an approach to implementing Agent AI in a safe, enterprise-grade environment,” he said. “We can deploy infrastructure for our customers, banks, governments, airlines and auto companies.
The Kyndryl Agent AI Framework can leverage thousands of infrastructure deployments and over 12 million AI-driven insights through its Kyndryl Bridge, the company's platform for integrating and coordinating complex IT infrastructures to better meet mission-critical requirements.
Combined, the Kyndryl Agent AI Framework and Kyndryl Bridge stated that they use advanced algorithms, self-learning, optimization, and design-by-design AI agents to turn complex data into clear insights.
“This agent AI framework allows us to connect these incredible technologies to a business environment, but we can do that in a way that addresses issues surrounding bias, security, enterprise grade, technology lock-in and more,” he said.
Kyndryl already uses the Kyndryl Agent AI framework with several large clients, Alma said.
In one example, we are working with the government to establish a native government. “We define policies from a human perspective, and agents who need little or no people come back to humans when there is something that may create risks that require human judgment,” he said.
In another example, Kyndryl also works with financial services companies.
“Genuine services are usually cutting edge in terms of some of these technologies, but we're also very concerned about security, risk, speed, and more,” he said. “We're working with this financial services company to employ certain processes. We can use things like customer experiences and actuarial support, such as fraud management and speeding up payments, to showcase agent AI and sit with humans and get more value for our customers.”
Alma refused to name one of these two clients.
Transparency within the Kyndryl Agent AI framework is essential to prevent bias from entering the Agent AI process, Alma said.
“By ensuring that you are learning in a fair and transparent way, bias is identified to address it rather than being reinforced for learning,” he said. “There are a lot of extra steps built in and it depends a little on the technology and model you are using, the maturity of the model. Also, as you pass the learning phase, it also changes from an inference perspective, from a transparency perspective, and from a large-scale language model to the way you construct large-scale language models and small-language models.”
How AI agents generally work depends on how they are treated, Alma said.
“Think of the workforce as a combination of agents and people,” he said. “In a sense, agents need to be treated like humans in the sense that they need to make choices to choose, mount, train and coach on a continuous basis, and maybe by the date of “selling” something else has come up and you might want to finish it.
Kyndryl plans to develop agent AI capabilities that focus on specific industries and how businesses can use specific data from their customers, says Alma.
“Our focus is not only on using generic technology for common issues, but also on making it very industry-aligned. We also keep the ecosystem open and enable our customers to use the framework as new technology is plugged in,” he said.
This story was originally published on the sister site CRN.
