Last year at AWS Summit in Washington, DC, the AI conversation still had a lot of future tense in it.
AI agents were becoming more capable. Intelligent document processing was moving closer to real enterprise use cases. Organizations were beginning to see how generative AI could connect with internal systems and support more meaningful workflows.
This year, the tone was different. The possibilities are no longer hard to imagine. Most organizations have seen the demos, tested the pilots, and identified use cases that could create value. The harder work now is turning those ideas into secure, reliable systems that can operate in production.
That was the clearest takeaway from this year’s Summit. The conversation has moved from possibility to execution.
One of the strongest themes from this year’s keynote was that progress has always depended on builders.
During the keynote, Dave Levy, AWS Vice President of Worldwide Public Sector, connected that idea to the history of the United States as the country approaches its 250th anniversary. He spoke about the early founders of our nation, the people who saw one breakthrough and immediately asked, “If that is possible, what else is?” It was a useful way to frame the moment we are in now. Innovation has always come from people who can take new tools and find practical ways to apply them.
That framing matters because many organizations are not struggling with a lack of AI ambition. They have plenty of ideas. Teams can point to use cases across operations, compliance, analytics, customer experience, internal productivity, and citizen services. In many cases, they have already started experimenting.
Where things get harder is after the proof of concept.
A pilot can show that something is technically possible. Production requires the system to work inside the reality of the organization. It has to connect to the right data, fit into existing workflows, meet security and compliance requirements, be adopted by users, and be maintainable after the initial excitement fades. That is where many AI efforts are starting to slow down.
Over the past year, many AI conversations have started at the application layer. Agents, copilots, chat interfaces, automated workflows, and AI-assisted development have dominated the discussion because they are easy to see and easy to understand.
This year’s Summit brought the infrastructure layer back into focus.
AWS highlighted continued investment across GovCloud, classified workloads, AI infrastructure, modernization credits, and training support. The specific announcements matter, but the broader signal matters more. Production AI depends on the foundation underneath it.
That is especially true in the public sector and regulated industries, where sensitive data, mission-critical systems, compliance requirements, legacy environments, and edge use cases all shape the path forward. These organizations cannot rely on experimentation alone to get them into production. They need infrastructure that matches the risk profile of the work.
Modernization was also an important undercurrent throughout the Summit, for the same reason. Many organizations do not have an AI problem in isolation. They have a data, cloud, and workflow modernization problem that AI is making more visible. Fragmented systems, manual reporting, disconnected data, and unclear governance have always created friction. AI simply raises the cost of leaving those issues unresolved.
For us at Red Oak Strategic, the takeaway was clear. If organizations want AI systems that do more than generate ideas, they need environments built for execution. Modern data platforms, secure cloud foundations, governed access, scalable compute, observability, cost management, and a realistic path from prototype to production all become part of the AI strategy.
Without that foundation, AI remains promising but difficult to operationalize.
Last year, a lot of organizations were still asking whether they could build AI solutions in the first place. Could an AI agent answer questions about an internal knowledge base? Could it summarize documents? Could it interact with an API? Could it generate code or help analysts move faster? Increasingly, the answer is yes.
The next set of questions is more operational. Can the organization trust the data behind the system? Can users understand where outputs are coming from? Can teams monitor performance, manage cost, secure access, and keep people in the loop where judgment is required? Can the system become part of how work actually gets done?
Those questions are less flashy than the demo, but they are where the real value lives.
At Red Oak Strategic, this is the work we keep coming back to with customers. AI adoption depends on choosing the right use cases, preparing the data, designing the architecture, and building a realistic path toward production execution.
That path often starts small. A focused discovery session, a prioritized use-case matrix, a prototype, a dashboard modernization effort, or a targeted workflow automation can all create momentum. What matters is that the work is connected to a larger path forward. The goal cannot be experimentation for its own sake. The goal has to be eventual execution.
This year’s Summit reinforced the direction we have already been moving as a company. We believe the next phase of AI adoption will be won by organizations that can connect strategy to implementation. That means helping teams understand where AI can create value while also building the data, cloud, and workflow foundation required to make that value real.
It also requires honesty about where organizations are starting from. Not every AI idea should become a production project. Not every proof of concept deserves continued investment. And not every team is ready to move directly into advanced agentic workflows.
Sometimes the most valuable first step is improving data visibility, automating reporting, modernizing a pipeline, or creating a secure environment where teams can test use cases responsibly. Those steps may sound less exciting than launching a new AI agent, but they are often what make production possible.
In summary, last year’s Summit was a reminder that the AI market was maturing. This year, the question is whether organizations are ready to execute.