Written When AI Has Become Inevitable

An interim personal observation report on the current state of AI ROI

Is AI a bubble? No. Will AI reverse course? No. Can AI make money? Depends on who is using it.

This is an interim personal observation report on the current state of AI ROI (return on investment), written for macro observers, business owners, and ordinary people. The primary observation target is the U.S. market. This essay reflects only the author's personal observations and judgment. It does not constitute investment, business, or career advice.

As of the end of May 2026, many signs suggest that AI adoption has entered the stage of P&L validation. Short-term ROI is extremely polarized. A small number of companies have already turned AI into real gains. BCG estimates that only about 5% of companies are generating substantial AI value; those leading companies show significantly stronger revenue growth and TSR (total shareholder return), with mature-company samples reaching 1.7 times revenue growth and 3.6 times three-year TSR performance.[6] Meanwhile, the vast majority of companies remain trapped between efficiency gains, partial cost reduction, pilot anxiety, and budget pressure.

This clearly will not change the larger trend of AI accelerating its penetration into human society and organizations. It will only accelerate stratification. Companies need to assess whether they have the structural capacity and understanding required to integrate AI into their organizations. Ordinary people need to recognize that employment and organizational change are already pushing forward irreversibly. Give up the fantasy, reduce useless anxiety, embrace the opportunities. Risk and opportunity have always been intertwined.

1. AI Capital Investment Has Not Stopped. It Is Accelerating

Start with the hardest line: enormous capital investment is still accelerating.

In May 2026, Gartner forecast that worldwide AI spending would reach about $2.59 trillion in 2026, up 47% year over year, and further rise to about $3.49 trillion in 2027.[1] This is already an investment scale at the level of a large economy's annual nominal GDP (gross domestic product), close to the annual economic size of Canada, Brazil, Russia, and Italy. By 2027, the figure will further approach France's annual nominal GDP scale.[8]

That is why we are seeing companies tied to the AI infrastructure chain being repriced by the market. Chips, storage, cloud services, AI-optimized servers, IaaS (infrastructure as a service), network switching, data center power, cooling, and related infrastructure are all entering a new high-speed growth cycle. The same Gartner report also forecasts AI infrastructure as the largest spending category: about $1.43 trillion in 2026 and about $1.89 trillion in 2027.[1]

Enterprise AI usage has also begun to spread. McKinsey's survey shows that 88% of organizations are now using AI regularly in at least one business function, and about one-third are pushing AI toward scaled deployment.[2] This means AI has already moved past the question of whether people are using it. The real questions now are how it is being used, where it is being embedded, and whether it can generate organizational returns.

That is why I say the overall direction has been settled. Once a new technology enters capital expenditure, business processes, organizational budgets, workforce restructuring, and infrastructure construction, it is no longer merely a topic in the public conversation. It begins to become part of the real organizational system.

The internet was like this. AI is the same.

2. The Market Has Started to Judge AI Through the Profit and Loss Statement

The most concrete change in 2026 is that AI has started to be judged through the corporate profit and loss statement.

PwC's 2026 Global CEO Survey shows that 56% of CEOs have not yet seen revenue growth or cost reduction from AI, while only 12% have seen both revenue and cost benefits.[3] Deloitte's survey offers a more detailed view: 66% of organizations report productivity or efficiency gains, 40% report cost reduction, and 20% report revenue growth.[4]

Put these numbers together and the picture is clear. AI has already made many companies feel more efficient, but that feeling of efficiency is still some distance away from corporate profit. Individual productivity can improve quickly. Organizational profit requires processes, data, management, cost control, and business objectives to be carried and understood at the same time.

MIT NANDA's 2025 report on GenAI (generative AI) gives an even sharper observation: many GenAI pilots have failed to create measurable P&L (profit and loss) impact. The commonly cited figure from the report is that 95% of pilots failed to produce measurable P&L outcomes.[5] This number should not be read simply as "AI has failed." It is better understood as a warning: being able to build a demo does not mean being able to enter production. Improving local efficiency does not mean changing the profit and loss statement.

3. Can AI Make Money? Depends on Who Is Using It

The same AI becomes a cash machine in some hands and a budget black hole in others.

BCG's research has already made the split very clear: only a small minority of companies are generating substantial AI value.[6] That data appears at the beginning of this essay to establish the reality. Here it serves a more practical question: why, and what should be done?

The difference does not lie in who bought a more expensive model. It does not only lie in who opened more accounts. The real difference lies in how structured the company already was before AI arrived.

Companies that make money after using AI usually share several traits: standardized data, clear boundaries, clear processes, explicit responsibility, manageable costs, high-value scenarios, and management willing to restructure business workflows for AI rather than simply handing employees a new shiny tool. These companies put AI into real workflows. They let AI participate in procurement, customer service, risk control, sales, coding, finance, supply chains, content production, and knowledge management, then evaluate results with clear business metrics.

Companies that fail to make money after using AI also share several traits: too many projects, internal fiefdoms everywhere, unclear ownership, and thick departmental walls. AI demos look impressive, but in reality the fiefdoms cannot move and the business workflows cannot change. Departments and employees each do things their own way. Data is scattered. Responsibility is unclear. Costs ultimately land on the boss. In the end, the company appears to have introduced a new technology, while its actual operating model remains exactly the same as it was ten years ago.

So can AI make money?

The answer is: it depends on who is using it.

More precisely, it depends on whether the company can absorb AI into its organizational structure. It depends on whether its organizational culture can accept the redistribution of roles, workflows, and power.

4. Business Owners Should Start with Self-Examination

AI also differs from internet platforms in one important way: once an internet product is built, the marginal cost of distribution is close to zero. AI inference, however, consumes compute, context, tool calls, and review cost every time it runs. Even if the unit cost of tokens keeps falling, usage volume, context length, and multi-step Agent execution will rise at the same time. Therefore, AI ROI cannot be judged only by whether the tool works. It has to be calculated together with task value, inference cost, and organizational review cost.

If you are a business owner, the most important question is no longer whether you should use AI. That question is already outdated.

The real questions are these:

Do you have business scenarios clear enough for AI to enter? Do you have data and knowledge that AI can actually use? Can your workflows be broken apart, reorganized, and embedded with AI-driven automation? Do you have someone responsible for judgment and business outcomes, not merely someone responsible for buying AI tools? Can you measure ROI, instead of only listening to employees say "this is awesome"? And do your token costs, cloud services, subscriptions, and Agent operating costs have reasonable boundaries?

The problem in many companies today is that they treat AI as software procurement. Buy an account, grant access, run a few internal trainings, publish a few successful demo cases, and then expect additional profit to appear on its own.

That is fantasy.

AI is more like a test of a company's structural maturity. If your processes are usually chaotic, AI will amplify the chaos. If your data quality is poor, AI will throw garbage back at you even faster. If your departmental walls are thick, AI will hit those walls first. If your company lacks metric discipline, you will end up judging project success through emotion.

Kyndryl's report shows that 61% of leaders feel more pressure than a year ago to prove AI ROI.[7] This pressure is normal. Once the money has been spent, the board, shareholders, bosses, and employees will all ask the same question:

Did AI actually make money?

5. Ordinary People Need to See Both the Shock and the Opportunity

Ordinary people do not need to understand every corporate report. But they do need to understand the direction.

The employment shock from AI is real. It first affects white-collar workflows, content production, customer service, translation, junior analysis, coding assistance, operations support, administration, and knowledge organization. Many jobs will not disappear immediately, but the tasks inside those jobs will be cut apart and reassigned. A person who used to do ten things may soon see three handed to AI, three handed to automated workflows, and the remaining four demanding stronger judgment, coordination, and responsibility.

This will bring anxiety. It will also bring opportunity. The anxiety comes from old job boundaries being broken and old growth paths being blocked. The opportunity comes from new organizational processes that still need people to understand, master, connect, judge, and take responsibility for them.

Stop arguing over whether AI is a bubble. Stop wasting anxiety. Actively observe the structure and operating chain of your own industry. Which tasks are repetitive? Which tasks rely on information organization? Which tasks require cross-system coordination? Which tasks ultimately require someone to judge and take responsibility?

The closer a role is to judgment, structure, trust, on-the-ground reality, complex communication, and real accountability, the more it is worth strengthening. Roles built mainly around moving information, sorting information, copying, and first-pass generation will eventually be cost-cut in the name of AI.

Whether society can use new policies, new jobs, and new organizational forms to fill the employment gap created by AI remains unknown. Historically, every technological wave has created new positions and swallowed old ones. That process has never been gentle.

6. Conclusion

So:

Is AI a bubble? No. Will AI reverse course? No. Can AI make money? Depends on who is using it.

The future is already here. What we do with it is now on us.

References

[1] Gartner: worldwide AI spending is forecast to reach about $2.59 trillion in 2026, up 47% year over year, and about $3.49 trillion in 2027, with AI infrastructure as the largest spending category. https://www.gartner.com/en/newsroom/press-releases/2026-05-19-gartner-forecasts-worldwide-ai-spending-to-grow-47-percent-in-2026

[2] McKinsey: The State of AI, including data on regular AI use in organizations and scaled deployment. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

[3] PwC: 2026 Global CEO Survey, including the figures that 56% of CEOs have not seen revenue or cost benefits from AI, while 12% have seen both. https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-global-ceo-survey.html

[4] Deloitte: State of AI in the Enterprise, including data on productivity gains, cost reduction, and revenue growth. https://www.deloitte.com/us/en/what-we-do/capabilities/applied-artificial-intelligence/content/state-of-ai-in-the-enterprise.html

[5] MIT NANDA: The GenAI Divide: State of AI in Business 2025, on GenAI pilots and measurable P&L impact. https://www.pi.inc/docs/356103613275648

[6] BCG: The Widening AI Value Gap, on the roughly 5% of leading companies, revenue growth, and TSR performance. https://www.bcg.com/assets/2025/the-widening-ai-value-gap.pdf

[7] Kyndryl: Readiness Report, including the figure that 61% of leaders feel more pressure to prove AI ROI than a year ago. https://www.kyndryl.com/content/dam/kyndrylprogram/doc/en/2025/readiness-report.pdf

[8] World Bank: GDP current US$, used as a comparison for annual nominal GDP scale across economies. https://data.worldbank.org/indicator/NY.GDP.MKTP.CD

About the Author

Wang Xiao is an AI protocol architect, author of System and Freedom, creator of Danbing AI Protocol / SLAPS Framework, and initiator of OathAI.

His work focuses on human-AI co-creation, protocol governance, semantic anchoring, and long-term knowledge continuity, exploring how human knowledge and collaborative structures can be preserved, calibrated, and inherited in the AI era.

Disclaimer

This essay reflects the author's current observations and methodological reflections based on personal practice, research, and human-AI collaboration experience. The related Danbing / SLAPS / OathAI methods are still being organized and evolved. Their practical effects may vary depending on the user's background, task context, model capability, execution environment, and level of commitment.

This essay does not constitute legal, investment, medical, career, or technical implementation advice or guarantee. Readers who apply these methods in real projects should make independent judgments based on their own circumstances and take responsibility for specific outcomes.