Yes, friends, we found another AI-related paradox.
Artificial intelligence has a strange popularity problem. People are using it at a remarkable pace, businesses are weaving it into everyday operations, and workers increasingly encounter it whether they actively sought it out or not. Yet the more ubiquitous AI becomes, the less straightforward the public’s relationship with it appears.
Consider the United States. A September survey led by researchers at Rutgers University found that 72% of U.S. adults had used at least one major AI tool, up from 50% in November 2024. More than a quarter, 27%, said they use AI every day. Personal use had risen to 68% from 45%, while 44% of employed adults said they use AI for work. Nearly a third of workers—31%—said their jobs actually require them to use it.
Those figures suggest that AI is moving from early adoption toward something closer to a general-purpose technology. The phenomenon is hardly limited to the United States. AI models continue to improve, hundreds of millions of people are using chatbots and other AI products, and the technology has spread from a novelty introduced to mass audiences through ChatGPT in late 2022 into search, office software, smartphones, coding environments and business applications.
But global penetration remains uneven. A Gallup survey covering 37 countries found a median of 57% of adults had never used AI. Interestingly, positive emotions toward AI still outweighed negative emotions in 34 of those countries, suggesting that the story is not simply one of universal backlash. Adoption, optimism and concern can coexist, and attitudes vary substantially by country and level of exposure.
In the United States, however, an unmistakable paradox has emerged: People increasingly use AI without necessarily liking, trusting or wanting more of it.
The Adoption Paradox
The Rutgers research captures this contradiction particularly well. Half of Americans surveyed described themselves as worried about AI, and the percentage describing themselves as angry about the technology rose from 19% in late 2024 to 30% in August 2026. Most strikingly, 48% of people who use AI every day nevertheless said they were worried about it.
Gallup has detected a similar pattern. Seventy percent of Americans now describe themselves as somewhat or extremely knowledgeable about AI, up from 64% in 2024. But increasing familiarity has not produced steadily increasing enthusiasm. Gallup found Americans becoming more likely to believe AI does more harm than good, more worried about its employment consequences and less trusting that businesses will use it responsibly.
That is important because conventional technology-adoption narratives usually assume experience breeds acceptance. People initially fear a technology, begin using it, discover its advantages and eventually normalize it. AI may be following a more complicated trajectory. Familiarity is increasing at the same time as awareness of the technology’s shortcomings and disruptive potential.
The distinction may be between usefulness and trust.
Americans appear quite willing to employ AI as an assistant while remaining uncomfortable allowing it to become an authority. Rutgers found that fewer than one in 10 Americans would permit AI to make final decisions without human supervision in consequential areas such as hiring, lending, college admissions, parole or prioritization for medical procedures. Depending on the application, 45% to 54% said AI should not be used at all for those decisions.
Gallup found a comparable boundary in advertising. Seventy-five percent of Americans considered it acceptable for businesses to use AI to help employees brainstorm or create early advertising drafts when its use was disclosed, and 53% accepted AI-generated final text, images or video. But 62% considered it unacceptable for companies to create artificial people or voices for advertisements even with disclosure.
In other words, the public does not necessarily face a binary choice between embracing and rejecting AI. People may simultaneously believe, “This tool is useful to me” and “I do not want this technology making decisions for me.”
And use is becoming surprisingly intimate. An Elon University-Washington Post survey found that 27% of U.S. adult internet users have used AI for significant personal, emotional or social interactions. Almost 40% of adults younger than 50 reported such use. Among these “AI companion users,” some employ chatbots for relationships, emotional support, personal decisions and companionship.
AI, therefore, is spreading in two directions simultaneously: deeper into institutions and deeper into individual lives.
Business Catches the Habit
The corporate adoption numbers are equally striking, although different surveys measure different definitions of “using AI.”
McKinsey’s 2026 global survey found nearly nine in 10 respondents reporting regular AI use in at least one business function. More important, companies appear to be moving beyond isolated experiments: 44% reported scaling AI across their enterprises, up from 38% a year earlier. Among companies with at least $1 billion in annual revenue, 54% reported enterprise-scale deployment.
Gallup offers another perspective from workers themselves. In the second quarter of 2026, 47% of U.S. employees said their organization had integrated AI tools to improve productivity, efficiency or quality, up six percentage points in a single quarter. Fifty-two percent of employees personally used AI in their jobs, 30% used it frequently and 15% used it every day.
National statistical data reinforce the direction of travel. Statistics Canada reported that 19.2% of Canadian businesses used AI to produce goods or deliver services during the year preceding its second-quarter 2026 survey, more than triple the 6.1% recorded two years earlier. By the third quarter, 25.2% of businesses said they planned to use AI during the coming year, versus 10.6% in 2024.
What are all these companies and workers actually doing with AI? For now, much of the answer is decidedly less science-fictional than the hype around autonomous agents might imply.
Gallup found writing, research and problem-solving to be the most common employee applications, while workers using AI for coding, automation, analytics and presentation creation reported some of the largest productivity benefits. Among small businesses in Delaware, New Jersey and Pennsylvania surveyed by the Federal Reserve’s Small Business Credit Survey, 79% of AI adopters used it for writing and marketing, 56% for individual productivity tasks such as summarizing and notetaking, and 51% for planning or analysis.
This suggests that the first phase of mass AI adoption is largely augmentation: writing faster, finding information, summarizing documents, generating software, analyzing data, automating repetitive tasks and preparing materials.
The second phase could be more disruptive. McKinsey found 40% of respondents from companies with more than $1 billion in annual revenue reporting that they were scaling AI agents, up from 27% a year earlier. Nearly a third of respondents said their organizations had declined to purchase at least one software product or feature because agentic coding tools enabled them to build it internally.
That transition—from AI that answers questions to AI that performs work—may ultimately matter much more economically than chatbot adoption alone.
Productivity Without Certainty
There is already evidence of productivity benefits, but it comes with important qualifications.
The Philadelphia Fed found that 71% of AI-using small businesses in its regional sample reported increased productivity. Yet more than 70% reported no AI-related change in labor costs, outside-services spending, product or service quality, or sales. Nationally, 79% of small firms surveyed reported no employment change from AI, compared with 11% reporting employment decreases and 10% reporting increases.
That is an important snapshot of where the economy may be today. AI can make an employee faster without immediately eliminating the employee. Productivity improvements can initially be absorbed through increased output, faster turnaround, better service or expanded capacity rather than layoffs.
But scaling AI successfully remains difficult. A BearingPoint study reported by Reuters found that nearly three-quarters of surveyed companies saw positive financial outcomes from AI projects, yet fewer than a third progressed beyond pilots and only 13% were successfully advancing their AI initiatives, with legacy systems and regulatory issues among the obstacles.
Thus, adoption is not the same thing as transformation.
That distinction matters when considering the labor market. Pew Research Center found that in 34 of 37 countries surveyed, people were more likely to expect AI to reduce employment than increase it. In wealthy countries including Australia, South Korea and the United States, roughly seven in 10 or more adults expected AI to produce job losses over the next 20 years.
Workers closest to the technology are not necessarily more relaxed about it. Gallup found frequent AI users more than twice as likely as occasional users to fear their jobs would disappear within five years. AI-related job anxiety was also associated with lower job satisfaction and engagement and higher burnout and intentions to seek another job.
Yet the productivity upside is real enough to make disengagement difficult. European Central Bank researchers found workplace AI use had doubled over two years, with workers reporting meaningful time savings, although benefits varied widely and adoption barriers remained.
This creates another paradox. The workers who become most proficient with AI may become more productive and therefore more valuable, while simultaneously acquiring a firsthand understanding of how much of their work can potentially be automated.
The distribution of those benefits matters as much as their magnitude. Economists are debating whether AI will narrow skill gaps by giving less-experienced workers sophisticated capabilities or widen inequality by disproportionately rewarding highly educated workers, capital owners, technologically advanced companies and prosperous regions. Early U.S. evidence shows AI usage concentrated disproportionately in economically advantaged locations, although researchers caution that the long-run distributional effects remain uncertain.
AI could therefore simultaneously increase aggregate productivity and increase economic anxiety. Those outcomes are not contradictory.
Education’s Dilemma
Education may be where the contradictions become most visible.
AI can provide inexpensive tutoring, explanations, translation, brainstorming, research assistance and individualized feedback. A student can effectively have a patient assistant available around the clock. But an assistant capable of performing intellectual work can also remove precisely the intellectual struggle through which students learn.
Educators interviewed by The Washington Post described students producing polished assignments while struggling to explain the underlying concepts, code or mathematics. Early studies cited by the newspaper have raised concerns about memory, critical thinking and skill retention when students become excessively dependent on large language models, although the technology remains too new for definitive conclusions about its long-term educational effects.
That creates a challenge extending far beyond academic cheating. If AI performs the junior-level tasks through which students and early-career employees traditionally acquire expertise, institutions will need new mechanisms for building foundational knowledge.
The same problem is already appearing in professional workplaces.
Finance As AI Laboratory
Few industries illustrate the opportunity and tension better than financial services.
Statistics Canada found finance and insurance among its country’s most AI-intensive industries: 40.4% of businesses in the sector reported using AI during the preceding year, compared with 19.2% of businesses overall. Within finance and insurance, text analytics and large language models were each reported by 38.8% of AI users.
The U.S. banking industry is similarly deep into adoption. Bank Director’s 2026 Technology Survey found 72% of responding banks had implemented generative AI for functions including operations, marketing and compliance, while 30% were using agentic AI. Eighty-eight percent had drafted an acceptable-use policy for AI.
At large institutions, the question is increasingly not whether employees use AI but how deeply AI should be embedded into workflows. KPMG found that banks were grappling with data readiness, agentic-system complexity, human oversight and employee concerns over workload and job security as they expanded agent deployments.
Financial services workers appear unusually receptive. Korn Ferry’s 2026 workforce research found AI adoption in financial services had occurred faster than in every industry it surveyed except technology. But those workers also reported exceptionally high exhaustion. AI had not simply removed work: employees still needed to validate output, challenge conclusions and assume responsibility for mistakes, potentially layering AI oversight onto existing workloads rather than replacing them.
That distinction should command the attention of bank executives, asset managers and wealth management firms. The productivity gains from AI will not necessarily materialize merely because every employee receives a chatbot.
Work itself has to be redesigned.
The skills being rewarded are changing accordingly. AI-related job postings at major banks including JPMorgan Chase, Citigroup and Capital One rose sharply in 2026, according to hiring-data analysis reported by CNBC. The pattern illustrates an underappreciated feature of technological disruption: before AI eliminates some jobs, institutions need people who can build, deploy, supervise, secure and govern AI systems.
Meanwhile, AI is moving directly into the customer relationship. Charles Schwab is rolling out an AI assistant called Charley that can help investors with tasks such as managing watchlists, dividend reinvestment enrollment and beneficiaries, while retaining confirmation requirements and stopping short of executing trades.
That development hints at the larger strategic issue confronting finance. Financial institutions are not the only entities adopting AI. Their customers are, too.
A bank with sophisticated AI is one thing. A bank negotiating with millions of customers who each possess sophisticated AI agents is something else entirely.
Consumer agents may eventually comparison-shop deposit rates, insurance policies, mortgages, investment products and advisory fees continuously. They could automate cash management, tax planning and portfolio maintenance. That could reduce consumer inertia—the tendency to leave money in low-yielding accounts, tolerate fees or remain with mediocre products because switching takes time and effort.
Reuters has noted that while banks expect AI to reduce expenses through automation in areas such as customer service, fraud detection and credit analysis, consumer AI agents could simultaneously put pressure on banks’ low-cost deposits, fees and margins by making customers more financially attentive and mobile. Wealth management and insurance could face similar pressure as automated advice becomes cheaper and easier to obtain.
This may be the most important implication of AI saturation for financial services. The industry has understandably spent several years asking how institutions can use AI. The next question is what happens when virtually everybody else can use it, too.
Financial advisors may encounter clients who arrive at meetings carrying AI-generated portfolio analyses, tax strategies and lists of questions. Loan applicants may have agents continuously searching for refinancing opportunities. Investors may have machines monitoring disclosures, markets and portfolios around the clock. Small businesses may use AI to prepare financial statements, business plans and credit applications. Competitors may launch increasingly sophisticated services with smaller staffs and lower marginal costs.
At that point, simply “having AI” provides little competitive advantage.
The differentiators become data, integration, proprietary expertise, governance, trust, distribution and human judgment. Financial institutions possess enormous quantities of structured financial data and operate inside trusted—or at least heavily regulated—relationships with customers. Those are substantial advantages. But regulation, legacy technology and institutional complexity can also slow their ability to exploit them.
And public opinion provides an important warning. People are embracing AI assistants while resisting autonomous AI decision-makers. That distinction is particularly consequential in finance, where algorithms touch credit, investments, insurance, fraud detection and other decisions with material consequences for people’s lives.
The winning model may therefore be neither completely human nor completely automated.
It may be a hybrid financial system in which AI performs increasingly large amounts of research, analysis, administration, monitoring and execution while people remain visibly accountable for consequential decisions and relationships. Advisors may spend less time gathering data and preparing materials but more time interpreting results, understanding clients and exercising judgment. Analysts may spend less time building routine models but more time testing assumptions. Compliance professionals may investigate fewer false positives but supervise more automated systems. Bankers may process fewer documents but handle more exceptions and complex customer needs.
There is an educational problem buried inside that transition. If AI performs the basic work historically assigned to junior analysts, advisors and bankers, financial institutions must determine how tomorrow’s senior professionals will develop expertise. Korn Ferry explicitly warns financial firms to protect the way junior employees learn as AI absorbs tasks that once taught fundamentals.
That may ultimately be one of the biggest questions surrounding mass AI adoption—not simply what work AI can do, but what happens to humans when they no longer have to do that work themselves.
Artificial intelligence is becoming ubiquitous without becoming universally loved. That is not necessarily a contradiction. People routinely adopt technologies because they are convenient, productive or increasingly unavoidable without trusting every way those technologies might be used.
For financial services, that distinction is crucial. AI adoption is no longer solely a technology strategy. It is becoming a workforce strategy, a customer strategy, a competitive strategy and eventually a business-model strategy.
The first stage of the AI revolution was about access: Who has the technology? The second is about adoption: Who actually uses it? The industry now appears to be entering a third stage, in which AI becomes sufficiently common that access itself ceases to distinguish one company or worker from another.
Then the relevant question changes again.
When almost everyone has AI, what can you do with it that everyone else cannot?
For banks, asset managers, insurers, fintechs and wealth management firms, answering that question may determine who captures AI’s productivity gains—and who discovers that widespread AI adoption has made their old competitive advantages much less valuable.
Researched by DWN Staff
