While we are still trying to understand what artificial intelligence can mean for the economy and the world of work, the time has come for the “border agents” who promise to dominate the conversation in 2026. This was one of the main concepts presented at the Re:invent 2025 event, held by Amazon Web Services (AWS) in Las Vegas from December 1st to 5th, where key trends in AI and cloud were revealed. THE NATION scheduled a conversation with Ben Schreiner, head of AI and data, who is driving the development of new architectures that define the next level of automation and working with companies to put AI into practice.
-What are the biggest misconceptions about implementing AI?
-This is something that is super simple, like downloading a free app on your phone and letting the AI do things automatically. The reality is that the apps people have on their phones solve a very different problem than the one an entrepreneur faces. The main difficulty for the non-technical entrepreneur is connecting the company’s internal data with AI, which is harder than you think. Another misconception is that we continue to automate things that should no longer exist, instead of thinking more holistically about whether they need to do this process or whether we can eliminate steps or entire functions.
-What specific business problems are agents supposed to solve?
-Long-standing problems that were previously resolved by hiring additional employees can be addressed. For SMBs, the biggest value will be agents having more time to focus on business growth. Unlike GenAI, which only answers questions, agents are able to reason, plan and take action, allowing them to solve more complex, manual and time-consuming workflows. The goal is for agents to increase the effectiveness of people so that the company grows faster.
– There is already a certain disappointment in the initial enthusiasm for AI, as the expected value is not recognized
–The rush to implement AI was followed by the realization that it is not as easy as previously thought. This is due to the data, because all solutions, including models, are only as good as the data to which they have access. The disappointment comes from the need to put significant effort into connecting the solution to the company’s data.
-What advice do you have for entrepreneurs on how to think about AI?
-How AI will transform your business and your competitive advantage in the next year or two. They should ask themselves how their customers are using AI and how their expectations are changing. It is expected that in the next two years everyone (businesses and consumers) will expect solutions tailored to their needs.
-If we talk about SMEs, how should they experiment with agents?
-You will experiment with AI and agents software that they already use (e.g. CRM, inventory or HR management). Many companies from software They work with AWS to improve their products with AI and agents, making them more valuable and easier to use. There is a huge opportunity for agent development and orchestration across multiple domains software which are often used by SMEs that do not communicate with each other to achieve greater efficiency.
-What are your recommendations for companies that value AI?
– Start with the problem that needs to be solved and make sure it is significant. Second, before considering models or agents, analyze what data is needed to solve the problem and whether you have access to it. Third, establish a metric to measure success (e.g. time saved or increased sales). Although AI saves time, its greatest value lies in reinvesting the time saved into higher-value activities, such as selling more, meeting more customers, or innovating.
-How should companies deal with large amounts of data and how can they determine which data is valuable for AI?
-Industry bound by regulations (banking, healthcare) are at an advantage because they are forced to organize their data. Other industries will recognize the direct connection between the quality of their data and the value they derive from AI. The data problem has become more complex because, in addition to internal data, many more external data sources are now available. The integration of this data enables hyper-personalization of the service, and in the near future customers will expect companies to know them well.
-What should companies do if they know their databases are deficient?
– First, they must prioritize solving the data problem and dedicate the time and sometimes money necessary to get the most out of AI. Secondly, it is advisable to look for partners who will help you solve the problem quickly. Third, they need to prioritize the data they need. It is important to work backwards from the problem you are trying to solve and focus only on the most valuable data, as trying to solve the problem of all data at once is not efficient.
-Which areas (HR, IT, Business) should be included in the design and implementation of agents within a company?
-AI represents a critical leadership moment affecting all types of business leaders who must decide how to use AI to drive growth. Good communication is required between different functions as all departments can benefit from AI in defining and implementing strategies. In addition, leaders must excel in communication and empathy, understand complex situations and be able to use agents bots in their teams.
-Given the concerns about the risks of using agents (e.g. data loss), what are the actual risks?
-The risk exists (cases of database deletions have been observed). For example, AWS applies the same mental model of security, scalability, and operational excellence to agents it uses internally. The key is to empower people to set parameters or boundaries (Guardrails) about what AI can and cannot do. This includes the use of AI guidelines that define permitted tools. “Observability” (monitoring) and evaluation of agents are critical steps in building trust and control. It is important that people define the rules.
Re:invert 2025’s final conference with 60,000 participants was led by AI and cloud guru Werner Vogels, CTO of AWS, who joked about the “death of the programmer” and called on them to appear in a new conference in which he called the programmer “Renaissance Developer, who is curious to learn and fail, who thinks in systems and communicates skillfully.” The main announcements introduced a new generation of AI models that can process text, images, videos and voice in a unified way, enabling companies to build smarter and more efficient applications called Amazon Nova. The concept of “Frontier Agents” was implemented, which are autonomous AI agents that work as virtual members of the development team, including the Kiro agent for programming, the AWS Security Agent for security, and the AWS DevOps Agent for operations. When it comes to processors, Trainium 3 was launched, a chip that offers up to 4.4 times more performance than the previous generation, making training possible. Larger, more complex AI models at lower costs and processors with five times more cache, delivering up to 25% better performance while maintaining superior power efficiency. Vogels quoted Jeff Bezos to describe this moment: “We are at the epicenter of multiple simultaneous Golden Ages.”
About The Author
