TECHnalysis Research Blogs
TECHnalysis Research president Bob O'Donnell publishes commentary on current tech industry trends every week at LinkedIn.com in the TECHnalysis Research Insights Newsletter and those blog entries are reposted here as well. In addition, those columns are also reprinted on Techspot and SeekingAlpha.

He also writes a regular column in the Tech section of USAToday.com and those columns are posted here. Some of the USAToday columns are also published on partner sites, such as MSN.

He also writes a 5G-focused column for Forbes that can be found here and that is archived here.

In addition, he also occasionally writes guest columns in various publications, including RCR Wireless, Fast Company and engadget. Those columns are reprinted here.

July 17, 2025
AWS Puts Agent-Focused Platform Center Stage

By Bob O'Donnell

AI-powered agents have unquestionably become the hot topic for all the largest tech vendors, as Amazon’s latest event in New York City once again demonstrated. However, as awareness of the complexities of creating and widely deploying agents has grown, so too has the number of questions regarding how to make that happen.

At AWS Summit, Amazon’s cloud computing division addressed those challenges head-on with a range of offerings covering development and deployment tools for agents, as well as a new marketplace from which customers can purchase finished agent products. In sum, it represented a comprehensive vision that AWS hopes will position it as a compelling platform for companies that want to start integrating agents into their environments, regardless of their level of IT sophistication.

One of the key offerings AWS unveiled was Amazon Bedrock AgentCore, which provides a suite of services for building custom agents that can leverage a wide range of different models and agent-building platforms, including open-source options such as Langchain, CrewAI and LlamaIndex. On top of that, AWS introduced its own new open-source agent-building tool it calls Strands Agents. As AWS VP Swami Sivasubramanian mentioned in his keynote, agents aren’t just the hot new AI buzzword, they represent a fundamental rethinking and evolution of how software applications are built, distributed, and used. That, in turn, makes tools like Strands Agents very important for the future of software and the platforms on which they run.

The key elements available through Bedrock AgentCore include a run-time environment for agents that integrates a range of security, reliability, and scaling capabilities. One of the big questions and concerns regarding agents is the potential impact—both positive and negative—that they can quickly have in a given environment because of their autonomous capabilities. To help build trust around agents, Amazon has integrated a series of guardrails, and it leverages its 10+-year old Lambda serverless architecture in which agents can be run. In addition, the company has integrated identity verification capabilities via AgentCore Identity, including the ability to tap into popular commercial and open-source authentication platforms such as OAuth. On top of that, one of the AgentCore services is an observability tool called AgentCore Observability that allows organizations to track how agents are performing, to see the type of data they’re accessing, and more.

Speaking of data, one of the key requirements for optimizing agents’ performance is to provide access to the specific data sets that need to be leveraged to perform certain workflows. AgentCore Gateway provides tools to access various APIs and other tools to make these connections possible. In addition, Amazon is supporting both the MCP (Model Context Protocol) and A2A (Agent-to-Agent) standards in a significant way via Gateway, giving companies more flexibility in the range of data sources, applications, and services to which they can connect.

Another important expectation for agents is the ability to build upon previous interactions through access to both short-term and long-term memory. The AgentCore Memory services makes the process of creating these types of often complicated memory structures much easier, allowing developers to create more compelling (and ultimately more successful) agentic experiences. Finally, for companies to monitor how the specific code of a given agent is processing data, Amazon has also created a service it’s calling AgentCore Code Interpreter that can be used to visualize more complex data interactions with certain types of agents.

In addition to the agent-building capabilities, AWS also announced several other key tools to help with the creation and fine tuning of AI models that can be used to drive these agentic applications. For the company’s family of Nova foundation models, Amazon announced the release of Nova Act, which can be used to drive autonomous actions within browsers. Given how critical browser-based applications and searches are to so many workflows, these capabilities will likely be used extensively for agentic AI applications. In fact, the AgentCore Browser service provides this specific capability within the AgentCore platform. Plus, because AgentCore supports the ability to use multiple models for different parts of an agentic workflow, organizations can leverage Nova Act for certain actions while using other LLMs for reasoning-based actions, for example.

For companies working to train or fine tune their models, there were two important announcements at Summit. First, the company unveiled new capabilities for its SageMaker AI model building tool that allows organizations to more easily customize Nova models by linking to their own data sources. Second, the company revealed a new type of S3 storage service called Amazon S3 Vectors that is specifically optimized for cost-effectively storing large amounts of the vector data that’s typically used for model training and fine tuning. Together these two offerings should make it easier for enterprises to build more customized and effective models. It’s also worth noting that these capabilities can be used independently of any agentic AI work but also fit nicely into the overall custom agent development process.

For companies that would prefer to buy instead of build when it comes to agents, AWS also launched a new Marketplace for commercially available agents. At launch, the marketplace already includes over 800 options ranging from general-purpose tools to industry and workflow specific options, highlighting how quickly this market is evolving. To put that in perspective, in conversations with an AWS Marketplace exec, he mentioned that the company’s SaaS-based Marketplace launched with only about 50 partners. The goal with this new Agent Marketplace is to provide a storefront for software partners as well as enable enterprises to easily find solutions to fit their specific needs.

Taken together, the announcements from AWS Summit reflect a very focused effort by the company to position itself at the cutting edge of the hottest new technologies. As with many big tech announcements, much of the news comes before many big real-world deployments of agentic AI have started to happen. But, as companies start to experiment with these technologies, it’s essential to provide a set of tools with which they can begin these efforts. Plus, by creating the Agent Marketplace, AWS is also giving companies an easy way to dip their toes into the agentic world. Similar to GenAI overall, we’re still very early in the agentic AI era, but as these latest announcements confirm, the pace of development continues to be blistering.

Here’s a link to the original column: https://www.linkedin.com/pulse/aws-puts-agent-focused-platform-center-stage-bob-o-donnell-fcgyc

Bob O’Donnell is the president and chief analyst of TECHnalysis Research, LLC a market research firm that provides strategic consulting and market research services to the technology industry and professional financial community. You can follow him on LinkedIn at Bob O’Donnell or on Twitter @bobodtech.