TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
AI is moving from back-and-forth chatbots toward agents that can carry out tasks, a shift that may bring the technology into everyday routines. Wharton professor Ethan Mollick argues that people should use AI to understand its capabilities, while concerns about jobs, reliability and dependence remain unresolved.
AI agents are beginning to handle practical tasks such as making restaurant reservations and shopping, sharpening a broader debate about how the technology will fit into daily life. In his new book, Co-Existence: The Next Phase of AI, Wharton professor Ethan Mollick argues that people need to use AI to understand its capabilities, even as concerns about job losses and other harms persist.
Mollick, an associate professor at the Wharton School and faculty director of its Generative AI Labs, says the change is not just that chatbots can give better answers. He argues that newer AI agents can carry out extended work with less step-by-step direction than earlier systems. In an interview with Vox, he said some systems can now work for hours, days or weeks, though the source does not independently verify that description or define the tasks involved.
Examples in the source suggest how agents may move from workplace tools into ordinary routines. AI agents can make restaurant reservations, Walmart has enabled agents to shop for people, and Apple announced an AI-ready home display that can control lights. These developments point to possible uses in shopping, home control and personal administration; they do not establish how widely the products are available or how reliably they work.
Mollick described using an agent to cancel an $8 monthly airline Wi-Fi subscription. He said he asked it to resolve the problem, and it contacted customer service before the subscription was canceled. The account illustrates the appeal of delegating tedious tasks, but also leaves open what actions the system took and what access or oversight was required. Mollick’s broader message is that AI’s practical effects are becoming harder to ignore, while users still have to decide what to delegate.
Everyday AI · October 2026
Can Humans And AI Share Daily Life?
AI is moving from answering questions toward taking on multi-step tasks. That could make everyday routines easier, while raising unresolved questions about reliability, privacy, jobs, and how much control people retain.
Which tasks are useful and safe to delegate—and which should remain under direct human control?
Examples point to possible uses, but do not establish how widely available or reliably these systems work.From chat to task-taking
Wharton professor Ethan Mollick argues that newer agents can attempt extended work with less step-by-step direction. The source offers examples, but no independent performance benchmarks.
Set an outcome
A person describes what they want done, such as resolving a service issue.
Work through steps
An agent may navigate forms, websites, or customer-service channels.
Check what happened
Users need clear visibility into actions, access, and outcomes.
Keep or reverse
Useful delegation depends on the ability to intervene or undo actions.
Where daily life may change
The cited developments suggest practical applications across home, shopping, and personal administration. They do not confirm broad availability or consistent performance.
Customer service
Mollick described an agent contacting an airline to cancel an $8 monthly Wi-Fi subscription. The account leaves oversight and access details unspecified.
Buying on someone’s behalf
The source says Walmart has enabled agents to shop for people. Pricing, permissions, and the range of availability are not detailed.
Connected devices
Apple announced an AI-ready home display that can control lights. The source provides no product milestone or reliability data.
Convenience comes with conditions
Delegation can reduce tedious navigation, but agents may need access to accounts and personal data. Trust depends on dependable results and meaningful user oversight.
Less friction
- Handle multi-step tasks across services
- Save time on repetitive administration
- Make complex software easier to use
More access, more stakes
- Can users inspect and correct each action?
- What happens if an agent misunderstands?
- How are sensitive data and permissions handled?
Concern is not a forecast
A Pew Research Center study, as described in the source, measured what respondents expect—not jobs already lost or a proven estimate of future displacement.
Expected AI to lead to fewer jobs over the next two decades. The result records public expectations, not a prediction of how many jobs will disappear.
The source describes a seven-point rise. Other signals it mentions—employee departures, corporate buyouts, investment concerns, and AI research applications—do not by themselves establish a single cause or broad trend.
Learn through informed use
Mollick presents experimentation as a way to understand what current tools can do. Familiarity can support better individual choices, but cannot decide how institutions deploy AI.
“You need to use AI in order to understand it.”
Ethan Mollick · Wharton professor · Interview with Vox
Evidence before wider trust
The next test is whether agents complete routine tasks consistently and make their actions easy to review, stop, and reverse.
How often do tasks succeed?
Clear performance evidence is needed across real-world tasks, including error rates and how much human review is required.
What can an agent access?
Consumers need specific information about account access, data handling, costs, and safeguards before delegating.
Can people intervene?
Useful systems should make actions legible and give people practical ways to stop, correct, or undo them.
What we know—and what remains open
What is an AI agent?
A system designed to carry out tasks, sometimes through multiple steps, instead of only answering prompts. The source provides no independent performance measurements.
What everyday tasks are cited?
Restaurant reservations, shopping, customer-service tasks, and control of home lighting through an announced Apple display. Availability and reliability vary by product.
Does the jobs figure predict displacement?
No. The 71% figure describes respondents’ expectations about the next two decades; it does not measure jobs already lost or prove a future outcome.
What should people decide now?
Whether a specific task is useful and safe to delegate, based on the system’s access, oversight, privacy practices, and ability to correct mistakes.
The Trade-Off in Everyday AI
Agents could make software easier to use by taking on tasks that otherwise require navigating websites, forms or customer-service chats. That may save time, but the same delegation raises questions about accuracy, privacy and control: an agent acting through a person’s accounts may need access to sensitive information, and users may not know every step it takes.
The stakes also extend beyond convenience. A Pew Research Center study published in August, as described in the source, found that 71 percent of Americans thought AI would lead to fewer jobs over the next two decades, up seven percentage points from two years earlier. The finding records public expectations, not a forecast that those job losses will occur. Still, it helps explain why enthusiasm for useful tools can exist alongside anxiety about work and economic change.
Mollick’s call for people to experiment with AI is a way to build practical familiarity, not a resolution to those wider concerns. Learning what a system can do may help people make informed choices, but it does not give individuals control over how employers, technology companies or governments deploy AI.
Products Worth Considering
Amazon Kindle Paperwhite (12th Generation, 2024), 16 GB, 7-inch Display, Jade
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Apple iPad Air (2025) 13‑inch Wi‑Fi 256 GB – Space Gray (M3 chip)
As an affiliate, we earn on qualifying purchases.
From Chatbots to Task Agents
Earlier consumer AI use often involved a person asking a chatbot questions and refining prompts in a back-and-forth exchange. Mollick says agents are changing that pattern by attempting multi-step tasks with less supervision. He also cautions, in effect, that people who have not followed recent developments or tried paid systems may have an outdated sense of what current tools can do. The source gives no benchmark data to compare systems or quantify the change.
The shift comes amid conflicting signals about AI. The source describes employees at Anthropic and Google leaving over concerns about the technology, researchers using AI to identify vulnerabilities in critical systems, Microsoft offering buyouts to 60,000 employees while increasing AI spending, and economists warning about a possible bubble. It also points to AI applications in scientific research and mathematics. These accounts span separate developments and claims; together, they show why public discussion combines potential benefits with concerns about safety, jobs and investment.
Mollick’s book frames the issue as coexistence: people will need to decide how AI fits into their work and personal lives as the systems develop. The source presents this as his argument, not as a settled consensus about the pace or outcome of AI adoption.
““You need to use AI in order to understand it.””
— Ethan Mollick, Wharton professor, in an interview with Vox
Reliability, Access and Job Effects
The source does not specify how often agents complete tasks correctly, what safeguards they use, or how much human review is needed. Mollick’s account of canceling a subscription is one example, not evidence that agents can reliably manage all customer-service interactions. Details about the availability, pricing and privacy practices of the cited products are also not provided.
The pace and scale of job changes remain uncertain. The Pew figure describes what respondents expect over two decades; it does not measure jobs already lost or prove that AI will cause a particular amount of displacement. The source also does not establish whether the reported employee departures, corporate buyouts or warnings about an investment bubble are directly caused by AI or indicate a broader trend.
It is also unclear how much control users will retain when agents act through email, shopping accounts or connected devices. The convenience of delegation depends on access, but more access can create risks if an agent misunderstands a request, makes an unwanted purchase or handles personal data in ways a user did not anticipate.
How Daily Use Develops
The next test is whether agents can perform routine tasks consistently and give users clear ways to review or reverse their actions. As companies add agent features to shopping, home devices and personal services, consumers will need specific information about permissions, data handling, costs and human oversight before deciding what to delegate.
Mollick’s book makes the case for learning through use, but the source names no formal launch dates, product milestones or regulatory deadlines. For now, the practical question for readers is narrower: which tasks, if any, are useful and safe to hand over, and which should remain under direct human control. Evidence about reliability and real-world effects will be needed to answer that as these tools reach more people.
Key Questions
What is an AI agent?
An AI agent is a system designed to carry out tasks, sometimes through multiple steps, rather than only answering prompts. The source says newer agents can require less step-by-step guidance, but it provides no independent performance measurements.
What everyday tasks are agents being used for?
The source cites agents making restaurant reservations, shopping on a person’s behalf and handling customer-service tasks. It also describes an announced Apple home display that can control lights. Availability and reliability vary by product, and the source does not give detailed specifications.
Does the Pew study show that AI will eliminate jobs?
No. The source reports that 71 percent of Americans surveyed thought AI would lead to fewer jobs over the next two decades. That is a measure of expectations, not proof of future job losses.
What does Mollick recommend?
Mollick argues that people should use AI to understand it and decide how it might fit into their lives. His advice does not settle concerns about job disruption, privacy, reliability or how much work people should delegate.
Source: Vox
