With new division, AWS bets big on agentic AI automation

Amazon Web Services customers can expect to hear a lot more about agentic AI in future with the news that the company is setting up a dedicated unit to promote the technology on its platform.

The division will be headed by Swami Sivasubramanian, until now Amazon’s VP of AI and data services.

“I’m excited about leading a new organization focused on advancing the power of agents, and look forward to bringing the latest innovation into the hands of our customers,” Sivasubramanian wrote on Wednesday in a LinkedIn post that set out the company’s agentic AI ambitions.

“Agentic systems offer possibilities that extend far beyond today’s chatbots and will drive efficiency like we haven’t seen before. They will orchestrate complex workflows and solve problems with human-like reasoning, while maximizing performance and cost effectiveness at scale,” he wrote. “Agents help developers write and debug code more effectively, allow businesses to automate complex decision-making processes, and create systems that can learn and adapt to new challenges.”

AWS CEO Matt Garman announced the appointment internally on Wednesday, in an email leaked to Reuters.

He promoted the formation of the new division as critical for Amazon’s push into cloud AI services, which he said had the potential to “be the next multi-billion business for AWS.”

“We have the opportunity to help our customers innovate even faster and unlock more possibilities, and I firmly believe that AI agents are core to this next wave of innovation,” Garman wrote.

Autonomous agents

The term ‘agentic AI‘ (not to be confused with ‘AI agents’, which refers to more basic helpmates) is open to some interpretation. Broadly, it embraces the principle that AI systems should be autonomous and able to make decisions in real time without the need for human direction or intervention.

This is the dream: to create systems that can solve complex, multi-step problems on their own. For today’s organizations that find themselves weighed down by ever more complex processes, the ability to hand workload to self-learning, problem-solving systems would be an important evolution.

AWS is already a significant player in this nascent space with the Amazon Bedrock platform (which enables agentic AI creation), and tools such as the Multi-Agent Orchestrator and the Amazon Q developer service. If it were to put them inside a single division it might anchor these elements into a more coherent sales story.

“We’re at an inflection point where enterprises are moving beyond experimentation with AI to seeking tangible business value,” said Chris Ashley, vice president of strategy, GTM & partnerships at UK AI platform company Peak.

“AWS have already made key moves over the past couple of years with their development of AWS Bedrock. A dedicated agentic team and capability is the logical next step for AWS to build on the generative capabilities they’ve been building out.”

For AWS customers, Ashley believed that agentic AI was good news, allowing organizations to automate complex workflows and speed up decision making in ways that would be impossible without the technology.

“We’ve seen clients achieve 20-30% efficiency gains when AI agents are properly deployed within existing cloud architectures,” he said.

Portability in question

However, he said that agentic systems could also lead to platform lock-in: “Technology leaders will ultimately want to ensure their agents can operate fluidly and securely across all clouds, all systems and all processes across the business. This will be a key consideration for AWS to strike the balance between agentic portability and having customers leverage their AWS native services.”

His recommendation for anyone evaluating agentic AI is to prioritize features such as the ability to control, direct, and constrain agentic AI, the degree to which agentic AI will work with existing technology, and how easy it is to track its actions and decision making. 

“AWS’s approach appears strong on infrastructure integration, but enterprises should carefully assess how these aspects are addressed,” said Ashley.

The jury is still out. Agentic AI run from inside a cloud platform could accelerate the ease with which organizations can take on what is a promising but unproven technology. This is the lure of the integrated cloud platform – handing some of the hard work to someone else.

Equally, customers will still need to choose their cloud partners carefully in a rapidly evolving field where certainties about the future are still in short supply.