enteredIn June 2026, the global generative AI market is experiencing a profound disillusionment and rebirth. According to the latest market research data released on June 2 by well-known research institutions such as Roy Morgan, in Australia alone, more than 58% of the population over the age of 14 (about 13.6 million people) incorporate tools such as ChatGPT, Google Gemini, and Microsoft Copilot into daily high-frequency use. However, in sharp contrast to the enthusiasm of the consumer side, a bottom-level reflection on "how AI can truly make a profit" is erupting within the capital market and entrepreneurial ecosystem.
The risk of software being "de-intermediated"
Since the beginning of 2026, major hyperscalers such as Microsoft, Google, Amazon and Meta have spared no effort to promote the commercialization of AI, but the market has never stopped questioning: How high is Copilot's actual conversion rate? Are companies really paying for this?
Industry analysts pointed out that currentlyThe biggest pain point in the commercialization of AI lies in "disintermediation of software." In the past, Microsoft relied on office suites such as Excel and Word to monopolize the enterprise-level market; now, if an enterprise's business processes evolve into a "fully automated agent cluster" driven by a third-party model, traditional office software will be downgraded to pure "Back-end tools", and their core business value will flow out.
The law of life and death for entrepreneurs: owning a workflow
In response to this trend, the technology community reached a general consensus this week: just by covering large models with a beautiful layerThe era of "wrapper startup" for UI is completely over.
Well-known technology criticsVioletta Bonenkamp bluntly stated in an industry analysis on June 1 that underlying large models such as OpenAI have now become "enterprise infrastructure" like electricity and tap water. In this context, if startups continue to regard the results output by models as truth, ignore data boundaries, or blindly purchase AI before sorting out their business processes, they will face disaster.
AI companies that truly have high gross profit and can form barriers in the second half of 2026 must meet the following characteristics:
Verticalization and narrow areas: Focus on extremely sophisticated professional tools (such as specific medical compliance, legal text traceability).
Workflow reshaping: is no longerA general-purpose assistant that "helps me write a paragraph of copy" is an automated closed-loop that can complete the entire cumbersome process of "review-reconciliation-delivery-filing" for employees.
Humanity in the Ring (Human-in-the-loop）: In high-risk and highly regulated industries, provide traceable AI-assisted review services with compliant stamps.