In early June 2026, the internet quietly passed a milestone: For the first time, more requests for web pages came from bots than from people, by about 57.5% to 42.5%. According to Cloudflare, much of that growth has come from a new generation of AI agents. Just three months earlier, Cloudflare’s CEO had predicted the crossover would come sometime in 2027. It arrived well ahead of schedule[1]. 

While the change won’t be apparent to the average internet user, it represents an enormous shift, from AI that simply answers questions to AI that accomplishes tasks. This shift opens up new questions as well as new opportunities for investors, and it is one that Aldrich Wealth is watching closely. 

Most people have experienced AI through assistants such as ChatGPT, Claude, or Google’s Gemini. Typically, users type a question or request into a chat window, and the AI assistant responds, answering questions or handling simple tasks like drafting an email or summarizing a report.  

An AI agent goes further. It takes a goal, breaks it into steps, uses tools and data to carry them out, checks its own work, and checks with a human when needed. If an AI assistant is like a well-read expert who can answer almost any question, an agent is more like an employee who has been given a specific job and the access needed to do it.  

Many people already use agents without realizing it. Amazon’s Alexa+ can book restaurant reservations, schedule home maintenance, and order delivery on a user’s behalf. Google Maps now uses Gemini to recommend trip adjustments in real time[2]. In the business world, companies are putting agents on routine customer service requests, like processing a return or updating an account. In software development, agents can write and test code with little human direction. 

Many big ideas in technology never caught on, so it’s fair to ask whether agentic AI will be any different. Unlike many fads, which offer new but not necessarily improved ways to do things, agentic AI aims to solve existing problems with greater speed and efficiency. 

A recent example of a tech fad is the metaverse: shared virtual worlds where people could work, shop, and socialize through virtual reality headsets. In 2021, Facebook was confident enough in the idea to rename itself Meta, yet few people ever made the metaverse part of their daily lives. 

In many respects, the metaverse was a solution in search of a problem. People could already shop online, FaceTime friends, and hold meetings over video. Virtual worlds did not make any of that meaningfully easier.  

Agentic AI, on the other hand, has the potential to streamline time-consuming and tedious tasks, making businesses more productive and cheaper to run. Companies that fail to adopt it are at real risk of being left behind by competitors that do. 

As Aldrich Wealth prepared presentation materials to discuss the state of agentic AI this past spring, the movement’s major milestones looked comfortably far off. But by the time the presentation was delivered in late summer, so much had changed that almost the entire deck had to be rebuilt. The only slide to survive was the one that defined what an agent is. 

But while adoption of agentic AI is widespread, it is still shallow. In McKinsey’s 2025 State of AI survey, 62% of organizations were at least experimenting with AI agents, but only 23% had begun rolling them out beyond testing, and usually in just one part of the business[3]. 

The biggest obstacle is trust. Can an agent be relied on to run payroll? To approve a customer refund? To send a wire transfer to the right account? Companies in heavily regulated industries like banking and healthcare will likely be slower to take full advantage of what agentic AI can offer. Early adoption is more likely to come from smaller, more nimble companies in less regulated fields, where the cost of an occasional error is lower and the payoff for moving quickly is higher. 

Much of the investment conversation around AI centers on the usual suspects: AI developers, such as OpenAI and Anthropic; tech giants, such as Google and Microsoft; and the data centers that power them. The companies building the tools that make agents reliable enough for real-world applications, however, are getting far less attention. 

Between an AI model and the business systems it touches, such as billing software, customer databases, and payroll, there exists a layer of software that manages how agents interact with those systems. It checks that agents are doing what they are supposed to, controls what they can access, and keeps a record of what they did. A growing number of companies, many of them early-stage startups, are building this layer. For example, one startup is developing a checkpoint that verifies the facts behind an agent’s action, such as issuing a refund or sending wire instructions, before that action goes through. 

AI models will likely get cheaper and easier to swap out over time. The software that keeps a company’s agents in check is another story: once it is built into daily operations, it is hard to replace. The companies that build that software may be well positioned to last. 

That said, not every company in this space will survive. When we look at early-stage opportunities, we start with two questions:  

• Is it disruptive? Does the business actually change how work gets done, rather than rebranding old automation as “agentic”? 
• Is it durable? Could an AI developer or an enterprise software vendor simply build the same capability into its own product? Being acquired can be a good outcome for investors. Being copied usually isn’t. For qualified investors, much of this opportunity is in private markets, well before companies reach public exchanges. Part of our role at Aldrich Wealth is to look three or more years ahead, bring emerging themes like this one to the families we work with, and help them understand how to access those opportunities as their portfolios evolve. 

If agents overtaking people in web traffic was the first milestone, the next may be agents spending more money online than people do. That would be a clear sign that agents have become a real force in the economy. 

The road there will not be smooth. Many of today’s agent projects will likely be canceled by 2027, and regulators and the public will raise hard questions about trust, privacy, and jobs. But many of these challenges strengthen the case we have outlined here. Every new demand for oversight, security, and accountability adds to the need for systems that monitor, verify, and govern what agents do. That is where we expect lasting value to form. 

We will continue to follow this space closely and share what we learn as it evolves. 

Advisory services offered through Aldrich Wealth LP (“Aldrich Wealth”), an Investment Adviser registered with the U.S. Securities & Exchange Commission. 

This article is provided for informational and educational purposes only and should not be construed as investment advice, a recommendation, or an offer to purchase or sell any security or investment product. Statements regarding emerging technologies, industry developments, and potential investment opportunities reflect current opinions and expectations, which are subject to change.  

 Investments in private companies and private markets involve significant risks, including illiquidity, limited transparency, valuation uncertainty, and the potential loss of the entire investment. Such investments may not be appropriate for all investors and may be subject to investor eligibility requirements. With any investment decision, it is important to consider the risks and objectives and match them to your own goals and risk tolerance. 

[1] Web traffic crossover: HotHardware, June 5, 2026, reporting Cloudflare data. Cloudflare measures HTTP requests, not time spent online. 

[2] Alexa+ and Google Maps examples: Amazon and Google, as cited in the Aldrich presentation. 

[3] Agent adoption figures: McKinsey, The State of AI 2025 (survey conducted mid-2025), as cited in the Aldrich presentation.