On a typical workday, the AI menu can look like a row of icons at the edge of the screen. ChatGPT is the catch-all assistant workers know for writing, brainstorming and fast answers. Gemini is the Google-native choice, close to search, Android and shopping discovery. Microsoft Copilot sits inside Word, Excel, Outlook and Edge, ready to summarize meetings, draft emails and rewrite sentences. Claude is the dutiful reader and aspiring long-form thinking partner. Perplexity is a quick research aide for people who want links and answers. Meta AI shows up inside social and messaging. A worker may start with one tool to polish a memo at 10 a.m., then reach for the same tool at 8 p.m. to plan a trip, compare a purchase or make sense of a health question.
PYMNTS Intelligence’s latest Consumer AI Benchmark report shows that work is now one of the strongest on-ramps to consumer AI, and that path is often a captive one. Employers don’t just give workers access. Many hand them a single tool or mandate a specific model, and when one assistant is the only one within reach during the workday, workers learn it, lean on it and carry it home.
The key finding is that consumer AI choice is shaped less by advertising and more by workplace access and mandates, default placement at work desks and the simple fact of being the dominant tool at hand.
AI in the Workplace
AI in the workplace has already reached critical mass.
The U.S. AI workforce has become too large to treat as an early-adopter niche. PYMNTS Intelligence finds that 86 million Americans use AI at work. That is 33% of U.S. adults and 53% of employed Americans. Those workers earn an estimated $7 trillion to $8 trillion in annual wages, with $16.3 trillion in GDP flowing through workers who use AI. The wage number alone is larger than the GDP of any other country in the world except the United States and China.
This measure counts the wage base touched by AI adoption. It doesn’t claim that AI created those wages or the output tied to those workers. That caveat is useful because it puts the trend in perspective: AI already sits within the working lives of a labor pool whose paychecks rival those of the world’s largest economies.
That economic scale also maps onto consumer adoption. Twenty-seven percent of U.S. consumers now use AI for both work and personal tasks. That’s roughly 70 million always-on users. Another 34% use AI only for personal tasks, and 6% use it only for work. The always-on 27% plus the work-only 6% make up the 33% captive workplace base used in this report. Add the personal-only group, and 67% of consumers have some current AI exposure. This is a broad market where work and life are starting to share the same tools.
From Work to Home
Workplace AI tools provided by employers often follow employees home.
For workers whose employers provide them with AI tools, the crossover into personal life is even clearer, often because that tool is the only one within reach. Among workers whose employer provides or mandates an AI platform, 78% use that workplace-provided tool for personal tasks. Sixty-two percent use it for personal tasks during work hours. Forty-four percent use it for personal tasks outside work hours. Just 22% keep the tool confined to work.
This is the behavioral jump in the benchmark, and the captive setup at work drives it. A company license or mandate gives workers a primary reason to try an AI tool. When it’s the only one available during the workday, that tool becomes the default. Daily job tasks give the worker a reason to learn it. The convenience of an already-open assistant gives the worker a reason to keep using it after hours. The path from work to personal life can start with a meeting summary, then move to a family schedule, a purchase comparison, a vacation plan or a personal finance question.
Workplace policy sets up that captivity. Only 13% of consumers say their workplace prohibits the use of AI. Meanwhile, 24% say their workplace provides an AI license. Eleven percent say the workplace requires a specific model, which for those workers is often the only AI they can access during the workday. Another 13% say the employer provides a license while allowing other tools. That means that employer choice introduces AI to a large share of consumers, not just personal exploration.
Workplace Lift
Some AI platforms follow workers home more than others.
The platform race looks different depending on the lens, and one lens hides how much a platform depends on the captive workplace. If the question is which AI work platform consumers will most likely use personally, ChatGPT appears to dominate. Among workers who use ChatGPT at work, 91% also use it for personal tasks. Gemini follows at 77%, Meta AI at 68%, Copilot at 63% and Claude and Perplexity at 60%.
But that ranking is misleading. ChatGPT already has a deep consumer base, so its 91% dual-purpose rate is a habit consumers have formed elsewhere, not proof that the workplace drove it.
Workplace influence is measured by distinguishing between captive exposure and personal use.
To see which platforms depend on the captive workplace, PYMNTS Intelligence looks at crossover: the difference between personal use among consumers who use AI in both work and their personal lives, and personal use among consumers who use AI only for personal tasks. That gap isolates the personal use that workplace exposure can claim.
By that measure, Copilot leans most on the workplace, with a 10-percentage-point lift. Gemini, Claude and Perplexity each show a nine-percentage-point lift. ChatGPT shows only a four-point lift, and Meta AI shows a one-point lift. Applied to the 70 million always-on consumers, Copilot’s roughly 7.1 million work-attributed personal users are consumers it would likely never have reached without employer mandates. Gemini adds 6.5 million, Claude 6.3 million and Perplexity 6.2 million on the same captive logic. ChatGPT adds just 2.8 million and Meta AI 0.9 million, because their personal use is already there.
Copilot’s workplace advantage is driven by enterprise distribution rather than user preference.
Copilot’s strength should be read as a captive-distribution signal, not a verdict on preference. The AI tool is built into the Microsoft work stack that many employees already use, and in many large enterprises, workers have Copilot on work devices while other AI tools are restricted. ChatGPT, by contrast, began as a consumer-focused tool for writing, recipes, poems and general questions, then moved deeper into work. Copilot often travels in the other direction, from mandated office software to personal use.
Copilot’s crossover is not uniform. Overall, 63% of consumers who use Copilot for work also use it in their personal lives. That rises to 79% among baby boomers and seniors, 75% among power users and 71% among workers who say AI is essential at work. It falls to 48% among workers who say AI is unnecessary at work and to 49% among macOS users. Copilot travels home when the mandate is tied to a work routine or long-standing office habits. It weakens when job use is shallow or when the device environment points elsewhere, a reminder that captive reach holds only while the tool stays on the path of least resistance.
AI Platform Choice
Worker choice and workplace access highlight different platforms.
Where a captive tool wins on convenience, a chosen tool wins on value, and Claude is the clearest example. Claude is among the platforms workers use the least under mandate but rate highest on impact. Eighty-one percent of consumers who use Claude at work say they couldn’t do their job without AI or would be substantially slower without it. That compares with 66% of workers who use other AI at work. The same comparison holds for each platform: Perplexity at 79% versus 68%, Meta AI at 76% versus 67%, Copilot at 74% versus 66%, Gemini at 71% versus 68%, and ChatGPT at 71% versus 67%.
This is where captive reach and chosen value separate. Copilot has the strongest captive pipeline, while Claude has the strongest productivity association. A platform can travel home because it is convenient, familiar and the only one licensed. Another can earn the highest productivity scores precisely because workers chose it for deeper work rather than an employer assigning it to them. Both are forms of adoption, but only one rests on the mandate.
These productivity results are correlational. Workers who seek out Claude or Perplexity may already be more focused on productivity, analysis or complex tasks. Even so, the pattern is useful. It shows that the captive workplace should be measured in more than one way: reach, lift, frequency and perceived value all show different parts of the adoption story.
Conclusion
The story emerging from this is simple. Work has become one of the strongest consumer AI gateways, and for a large share of workers, it’s a captive one. When the office provides a single tool, it tends to become the default people carry home. ChatGPT remains the broad consumer favorite on its own merits. Copilot shows how a workplace mandate, more than preference, can shape consumer behavior. Claude shows how deep work value can build loyalty where workers get to choose. Gemini shows how defaults and broad utility can cross both settings.
For banks, payments companies and FinTech providers, the takeaway is that AI adoption shouldn’t be measured only by app downloads, model preference or marketing reach. The AI assistant that an employer requires an employee to open every morning can become the assistant they trust for personal decisions, whether or not they would have chosen it on their own. That may influence how consumers search, plan, compare, budget, shop and, eventually, manage financial tasks. As AI assistants move further into productivity suites, banking apps and commerce sites, the strongest adoption path may run through the captive tools consumers already have open at their jobs.
Methodology
“Captive at Work, Adopted at Home: The Hidden Force Driving Consumer AI Choice,” the eighth installment of the Consumer AI Benchmark Series, is based on a PYMNTS Intelligence survey of 2,663 U.S. adults conducted in April 2026. Samples were balanced to match the U.S. adult population by age, gender, education and income. Some findings use smaller subgroup bases, including 648 workers whose employers provide or mandate an AI platform, 856 consumers who use AI at work and 422 consumers who use Copilot at work. The economic footprint estimate starts with the 33% of roughly 260 million U.S. adults who report using AI at work, then compares that workforce with employment, wage and GDP data. It estimates the economic weight of workers who have adopted AI, while excluding any claim about incremental productivity gain.
