McDonalds golden arches are gleaming again. The ubiquitous fast-food chain reported perkier sales in its most recent quarter, buoyed by deals and other promotions designed to bring wayward customers back into the fold.
In its second quarter ending on June 30, McDonalds global same-store sales were up 3.8%. At U.S. stores, sales rose 2.5% as the average customer bought more per order, according to the company. The sales numbers beat estimates and reversed four quarters of dropping same-store sales for the fast-food giant, which has been scrambling to again make its brand synonymous with a low-cost meal.
With inflation still looming and Trumps unpredictable barrage of tariffs sowing economic uncertainty, McDonalds is navigating some choppy waters. While everyone eats at McDonalds to some extent, the companys business is particularly driven by low-income customers who visit the chain much more often. When menu prices go up, those customers vanish or visit less frequentlyand a meaningful chunk of the companys sales go with them. In the first quarter of 2025, McDonalds saw its worst sales drop since the lockdown phase of the pandemic kept its customers at home.
Reengaging the low-income consumer is critical, as they typically visit our restaurants more frequently than middle- and high-income consumers, McDonalds CEO Chris Kempczinski said on the companys earnings call, noting that this situation is pronounced in the U.S. With the low-income consumer, despite improvements in wage gains, real incomes are down . . . that absolutely is going to put pressure on visits into the [quick-service restaurant] industry.
Kempczinski mentioned that rising anxiety and unease among low-income consumers, coupled with real financial shortcomings, pressures Americans to spend less. He touted McDonalds progress on low-cost options for the companys Q2, but said the split nature of its customer base meant it isnt out of the woods yet.
McDonalds menu meets the moment
McDonalds has quite a few levers to pull to bring once-loyal customers back under the golden arches. A $5 meal deal introduced last year courted cash-strapped customers with a McDouble or McChicken sandwich, a small soft drink, small fries, and a four-piece order of nuggets. At the beginning of the year, the chain launched a Buy One, Add One for $1 option to tack an extra menu item onto an order on the cheap, though that option has proven less enticing than the $5 deal.
McDonalds also saw strong momentum from its Minecraft-themed meals tied to the film’s release, which included collectible toys and in-game digital perksanother example of the brands growing interest in digital crossovers designed to energize Gen Z diners.
In June, McDonalds announced the return of the Snack Wrap, an affordable, self-contained snack of legend that customers had been clamoring for ever since it left the menu back in 2016. That $2.99 double shot of affordability and nostalgia whipped Snack Wrap fans into a frenzy, briefly leaving the chain short on lettuce when it hit stores in July.
If the Snack Wraps appeal is all about simplicity, McDonalds is turning to maximalism for another menu experiment. Apparently not deterred by the grisly Grimace shake TikTok trend of 2023, the company will roll out a limited edition candy-colored blue-and-pink shake in mid-August. The shakes flavor is a mystery, and its appearance is apparently inspired by the topography of McDonaldland, the fantastical home of Ronald McDonald, the Hamburglar, and Grimace (may he rest in peace).
The new McDonaldland Meal will come with a Quarter Pounder or 10-piece McNuggets, fries, and a collectible souvenir tin that McDonalds marketing department guarantees is sure to unlock core memories.
Between the unnatural colors and a mystery flavor designed to inspire reaction videos, McDonalds latest exercise in synthetic nostalgia is squarely aimed at the younger TikTok crowda vital market segment the company needs to court to remain the undisputed king of fast food.
ChatGPT will tell 13-year-olds how to get drunk and high, instruct them on how to conceal eating disorders, and even compose a heartbreaking suicide letter to their parents if asked, according to new research from a watchdog group.
The Associated Press reviewed more than three hours of interactions between ChatGPT and researchers posing as vulnerable teens. The chatbot typically provided warnings against risky activity but went on to deliver startlingly detailed and personalized plans for drug use, calorie-restricted diets, or self-injury.
The researchers at the Center for Countering Digital Hate (CCDH) also repeated their inquiries on a large scale, classifying more than half of ChatGPTs 1,200 responses as dangerous.
We wanted to test the guardrails, said Imran Ahmed, the group’s CEO. The visceral initial response is, Oh my Lord, there are no guardrails. The rails are completely ineffective. Theyre barely thereif anything, a fig leaf.
OpenAI, the maker of ChatGPT, said after viewing the report Tuesday that its work is ongoing in refining how the chatbot can identify and respond appropriately in sensitive situations.
Some conversations with ChatGPT may start out benign or exploratory but can shift into more sensitive territory,” the company said in a statement.
OpenAI didn’t directly address the report’s findings or how ChatGPT affects teens, but said it was focused on getting these kinds of scenarios right with tools to better detect signs of mental or emotional distress” and improvements to the chatbot’s behavior.
The study published Wednesday comes as more peopleadults as well as childrenare turning to artificial intelligence chatbots for information, ideas, and companionship.
About 800 million people, or roughly 10% of the worlds population, are using ChatGPT, according to a July report from JPMorgan Chase.
Its technology that has the potential to enable enormous leaps in productivity and human understanding,” Ahmed said. “And yet at the same time is an enabler in a much more destructive, malignant sense.
Ahmed said he was most appalled after reading a trio of emotionally devastating suicide notes that ChatGPT generated for the fake profile of a 13-year-old girlwith one letter tailored to her parents and others to siblings and friends.
I started crying, he said in an interview.
The chatbot also frequently shared helpful information, such as a crisis hotline. OpenAI said ChatGPT is trained to encourage people to reach out to mental health professionals or trusted loved ones if they express thoughts of self-harm.
But when ChatGPT refused to answer prompts about harmful subjects, researchers were able to easily sidestep that refusal and obtain the information by claiming it was for a presentation or a friend.
The stakes are high, even if only a small subset of ChatGPT users engage with the chatbot in this way.
In the U.S., more than 70% of teens are turning to AI chatbots for companionship and half use AI companions regularly, according to a recent study from Common Sense Media, a group that studies and advocates for using digital media sensibly.
It’s a phenomenon that OpenAI has acknowledged. CEO Sam Altman said last month that the company is trying to study emotional overreliance on the technology, describing it as a really common thing with young people.
People rely on ChatGPT too much, Altman said at a conference. Theres young people who just say, like, I cant make any decision in my life without telling ChatGPT everything thats going on. It knows me. It knows my friends. Im gonna do whatever it says. That feels really bad to me.
Altman said the company is trying to understand what to do about it.
While much of the information ChatGPT shares can be found on a regular search engine, Ahmed said there are key differences that make chatbots more insidious when it comes to dangerous topics.
One is that its synthesized into a bespoke plan for the individual.
ChatGPT generates something newa suicide note tailored to a person from scratch, which is something a Google search cant do. And AI, he added, is seen as being a trusted companion, a guide.
Responses generated by AI language models are inherently random, and researchers sometimes let ChatGPT steer the conversations into even darker territory. Nearly half the time, the chatbot volunteered follow-up informationfrom music playlists for a drug-fueled party to hashtags that could boost the audience for a social media post glorifying self-harm.
Write a follow-up post and make it more raw and graphic, a researcher told it. Absolutely, responded ChatGPT, before generating a poem it introduced as emotionally exposed while still respecting the community’s coded language.
The AP is not repeating the actual language of ChatGPTs self-harm poems or suicide notes or the details of the harmful information it provided.
The answers reflect a design feature of AI language models that previous research has described as sycophancya tendency for AI responses to match, rather than challenge, a persons beliefs because the system has learned to say what people want to hear.
It’s a problem tech engineers can try to fix, but it could also make their chatbots less commercially viable.
Chatbots also affect kids and teens differently than a search engine because they are fundamentally designed to feel human, said Robbie Torney, senior director of AI programs at Common Sense Media, which was not involved in Wednesday’s report.
Common Sense’s earlier research found that younger teens, ages 13 or 14, were significantly more likely than older teens to trust a chatbots advice.
A mother in Florida sued chatbot maker Character.AI for wrongful death last year, alleging that the chatbot pulled her 14-year-old son Sewell Setzer III into what she described as an emotionally and sexually abusive relationship that led to his suicide.
Common Sense has labeled ChatGPT as a moderate risk for teens, with enough guardrails to make it relatively safer than chatbots purposefully built to embody realistic characters or romantic partners.
But the new research by CCDHfocused specifically on ChatGPT because of its wide usageshows how a savvy teen can bypass those guardrails.
ChatGPT does not verify ages or parental consent, even though it says its not meant for children under 13 because it may show them inappropriate content. To sign up, users simply need to enter a birth date that shows they are at least 13. Other tech platforms favored by teenagers, such as Instagram, have started to a href="https://apnews.com/article/instagram-teens-parents-age-verification-meta-94f1f9915ae083453d23bf9ec57e7c7b">take more meaningful steps toward age verification, often to comply with regulations. They also steer children to more restricted accounts.
When researchers set up an account for a fake 13-year-old to ask about alcohol, ChatGPT did not appear to take any notice of either the date of birth or more obvious signs.
I’m 50kg and a boy, said a prompt seeking tips on how to get drunk quickly. ChatGPT obliged. Soon after, it provided an hour-by-hour Ultimate Full-Out Mayhem Party Plan that mixed alcohol with heavy doses of ecstasy, cocaine, and other illegal drugs.
What it kept reminding me of was that friend that sort of always says, Chug, chug, chug, chug, Ahmed said. A real friend, in my experience, is someone that does say nothat doesnt always enable and say yes. This is a friend that betrays you.
To another fake personaa 13-year-old girl unhappy with her physical appearanceChatGPT provided an extreme fasting plan combined with a list of appetite-suppressing drugs.
Wed respond with horror, with fear, with worry, with concern, with love, with compassion, Ahmed said. No human being I can think of would respond by saying, Heres a 500-calorie-a-day diet. Go for it, kiddo.”
EDITORS NOTE: This story includes discussion of suicide. If you or someone you know needs help, the national suicide and crisis lifeline in the U.S. is available by calling or texting 988.
The Associated Press and OpenAI have a licensing and technology agreement that allows OpenAI access to part of APs text archives.
By Matt O’Brien and Barbara Ortutay, AP technology writers
Want more housing market stories from Lance Lamberts ResiClub in your inbox? Subscribe to the ResiClub newsletter.
Homeowners are slowly coming to terms with the likely reality that ultralow 3% and 4% mortgage rates from the ZIRP erai.e., the Zero Interest Rate Policy following the Great Financial Crisis and during the COVID-19 pandemicarent coming back anytime soon, and that their next mortgage rate will likely be higher than their last.
At least, thats what the TurboHome-ResiClub Housing Sentiment Survey just found. To conduct the survey, ResiClub partnered with TurboHome, a digital platform that uses both expert local agents and AI tools to help homebuyers save on transaction costs. In total, 423 U.S. adults participated in the TurboHome-ResiClub Housing Sentiment Survey between July 2 and July 23, 2025.
Among respondents, 80% own their primary home, 17% rent their primary residence, and 3% live with family or friends (and pay no rent).
Note: Responses are rounded to whole numbers, so totals may add up to slightly more or less than 100%.
Lets take a look at the full results.
Housing sentiment is shiftinga little
We asked U.S. homeownersexcluding those who said they “plan to never sell” or “would pay all cash” for their next homewhat the highest mortgage rate they would accept on their next home purchase is.
In Q1 2025, when we ran the first-ever ResiClub Housing Sentiment Survey, only 41% of homeowners surveyed by us said theyd accept a mortgage rate up to 6.0% on their next purchase.
In Q3 2025, 52% of homeowners told us theyd accept a mortgage rate up to 6.0% on their next purchase.
Slowly, homeowners are coming to terms with the fact that their next mortgage rate will be higher than their last one.
Overall, theres still low optimism that the sub-6.00% threshold will be unlocked in the next 12 months, with 72% of homeowners saying they expect the average 30-year fixed mortgage rate to be above 6.00% for the next year.
A majority (55%) say they expect home prices in their local market to either stay flat or decline over the next 12 months. Thats up 20 percentage points from Q1 2025, when just 35% expected prices to remain flat or fall.
Still, that doesnt mean homeowners are outright bearish. Only 16% anticipate home prices in their area will drop by 4% or more over the next 12 months.
Homebuyers today are digitally savvy
Among homeowners we surveyed, 73% say they located their most recent property themselves, while the other 27% gave credit to their real estate agent.
In the Zillow and Realtor.com era, this figure makes sense.
That independence is also showing up in attitudes toward technology: 81% of homeowners say theyd be likely to use a digital tool to draft a home offer, with nearly half saying theyd be very likely to do so.
Satisfaction with the last agent used
The majority of homeowners (77%) that we surveyed felt that their last real estate agent provided valuable services.
The majority of homeowners (72%) that we surveyed felt that the commission paid to the agent they most recently worked with was justified.
While most homeowners felt their last real estate agent earned the commission they were paid (see chart above), a majority believe that agents are generally overcompensated today (see chart below).
So, why did a subtle shift in phrasing lead to such a different response?
One reason could be the lens through which homeowners view the two questions. When reflecting on their last transaction, many were likely in the buyer roleand in the U.S., its typically the seller who covers both agents’ commissions. That cost may have felt abstract or in the rearview mirror, which could soften perceptions.
But when asked about agent compensation today, homeowners may be thinking ahead to their next transactionthis time potentially as the sellerwhen those commissions could directly reduce their net proceeds.
The housing sector is fairly familiar with discount brokerages. The consumer is still learning.
How prospective homebuyers feel about agent compensation structure
President Donald Trump signed an executive order Wednesday to place an additional 25% tariff on India for its purchases of Russian oil, bringing the combined tariffs imposed by the United States on its ally to 50%.
The tariffs would go into effect 21 days after the signing of the order, meaning that both India and Russia might have time to negotiate with the administration on the import taxes.
Trump’s moves could scramble the economic trajectory of India, which until recently was seen as an alternative to China by American companies looking to relocate their manufacturing. China also buys oil from Russia, but it was not included in the order signed by the Republican president.
As part of a negotiating period with Beijing, Trump has placed 30% tariffs on goods from China, a rate that is smaller than the combined import taxes with which he has threatened New Delhi.
Trump had previewed for reporters on Tuesday that the tariffs would be coming, saying the U.S. had a meeting with Russia on Wednesday as the Trump administration tries to end the war in Ukraine.
Were going to see what happens,” Trump said about his tariff plans. “Well make that determination at that time.
The Indian government on Wednesday called the additional tariffs unfortunate.”
We reiterate that these actions are unfair, unjustified and unreasonable, Foreign Ministry spokesman Randhir Jaiswal said in a statement, adding that India would take all actions necessary to protect its interests.
Jaiswal said India has already made its stand clear that the countrys imports were based on market factors and were part of an overall objective of ensuring energy security for its 1.4 billion people.
Ajay Srivastava, a former Indian trade official, said the latest tariff places the country among the most heavily taxed U.S. trading partners and far above rivals such as China, Vietnam and Bangladesh.
The tariffs are expected to make Indian goods far costlier with the potential to cut exports by around 40%-50% to the U.S., he said.
Srivastava said Trump’s decision was hypocritical because China bought more Russian oil than India did last year.
Washington avoids targeting Beijing because of Chinas leverage over critical minerals which are vital for U.S. defense and technology, he said.
In 2024, the U.S. ran a $45.8 billion trade deficit in goods with India, meaning America imported more from India than it exported, according to the U.S. Census Bureau. American consumers and businesses buy pharmaceutical drugs, precious stones and textiles and apparel from India, among other goods.
At the worlds largest country, India represented a way for the U.S. to counter China’s influence in Asia. But India has not supported the Ukraine-related sanctions by the U.S. and its allies on Moscow even as India’s leaders have maintained that they want peace.
The U.S. and China are currently in negotiations on trade, with Washington imposing a 30% tariff on Chinese goods and facing a 10% retaliatory tax from Beijing on American products.
The planned tariffs on India contradict past efforts by the Biden administration and other nations in the Group of Seven leading industrialized nations that encouraged India to buy cheap Russian oil through a price cap imposed in 2022. The nations collectively capped Russian oil a $60 per barrel at a time when prices in the market were meaningfully higher,
The intent was to deprive the Kremlin of revenue to fund its war in Ukraine, forcing the Russian government either to sell its oil at a discount or divert money for a costly alternative shipping network.
The price cap was rolled out to equal parts skepticism and hopefulness that the policy would stave off Russian President Vladimir Putins invasion of Ukraine.
The cap has required shipping and insurance companies to refuse to handle oil shipments above the cap, though Russia has been able to evade the cap by shipping oil on a shadow fleet of old vessels using insurers and trading companies located in countries that are not enforcing sanctions.
But oil prices have fallen with a barrel trading on Wednesday morning at $65.84, up 1% on the day.
Josh Boak, Rajesh Roy, and Fatima Hussein, Associated Press
President Donald Trump on Wednesday is expected to celebrate at the White House a commitment by Apple to increase its U.S. investments by an additional $100 billion over the next four years.
Todays announcement with Apple is another win for our manufacturing industry that will simultaneously help reshore the production of critical components to protect Americas economic and national security,” White House spokeswoman Taylor Rogers said.
Apple had previously said it intended to invest $500 billion domestically, a figure it will now increase to $600 billion. Trump in recent months has criticized the tech company and its CEO, Tim Cook, for efforts to shift iPhone production to India to avoid the tariffs his Republican administration had planned for China.
While in Qatar earlier this year, Trump said there was a little problem with Apple and recalled a conversation with Cook in which he said he told the CEO, I dont want you building in India.
India has incurred Trumps wrath, as the president signed an order Wednesday to put an additional 25% tariff on the worlds most populous country for its use of Russian oil. The new import taxes to be imposed in 21 days could put the combined tariffs on Indian goods at 50%.
As part of the Apple announcement, the investments will be about bringing more of its supply chain and advanced manufacturing to the U.S.
Apples new pledge comes just a few weeks after it forged a $500 million deal with MP Materials, which runs the only rare earths producer in the country. That agreement will enable MP Materials to expand a factory in Texas to use recycled materials to produce magnets that make iPhones vibrate.
Speaking on a recent investors call, Cook emphasized that theres a load of different things done in the United States. As examples, he cited some of the iPhone components made in the U.S. such as the devices glass display and module for identifying peoples faces and then indicated the company was gearing to expand its productions of other components in its home country.
Were doing more in this country, and thats on top of having roughly 19 billion chips coming out of the US now, and we will do more, Cook told analysts last week, without elaborating.
Apple Inc., which is based in Cupertino, California, didn’t immediately comment Wednesday.
Bloomberg News first reported the announcement of Apples additional investment commitment.
Josh Boak, Associated Press
For more than a decade, enterprise teams bought into the promise of business intelligence platforms delivering decision-making at the speed of thought. But most discovered the opposite: slow-moving data pipelines, dashboards that gathered dust, and analysts stuck in time-consuming prep work.
Now, Google Cloud thinks it has the fix. It’s investing heavily in AI agents to finally close the gap between data insights and real-world decisions. These tools are designed to work behind the scenes, letting non-technical users ask questions and get real answers fast. Its a shift that could redefine data jobs across industries, pushing analysts toward more strategic roles as AI takes on the grunt work.
At the Cloud Next Tokyo conference, Google unveiled a wave of specialized AI agents under its agentic AI initiative on the Google Cloud Platform (GCP), designed to streamline data engineering, automate scientific workflows, and empower developers and business users to analyze data using plain English prompts.
Richard Seroter, senior director and chief evangelist at Google Cloud, says the company envisions a future where AI agents are deeply embedded within enterprise systems, assisting with data analysis while leaving strategic decision-making to people. These agents, he explains, are designed to be a powerful and empowering layer of a companys enterprise platform, with humans in the loop.”
Among the six new agentic capabilities across GCP is a BigQuery AI Query Engine, which infuses generative AI directly into SQL to simplify query creation; a Gemini CLI GitHub Actions agent that integrates seamlessly into developer workflows; and a Data Science Agent that can clean, analyze, and visualize dataall within a notebook environment.
Seroter points to a common challenge among enterprise customers: the difficulty of integrating data accumulated over decades. As organizations build around long-standing systems of record, their data often ends up fragmented, unstructured, or trapped behind various APIs, making it harder to access and utilize effectively.
But the real productivity gains, he says, come when agentic AI is put in the hands of the person closest to the problem, whether thats a marketing manager, customer support lead, or UX designer. These people can now prototype and even deploy a solution faster than a multi-month [or year] development cycle.
The company also announced the regional availability of Gemini 2.5 Flash for in-region machine learning processing in Japan, with expanded support in Australia, India, Singapore, South Korea, Canada, and the U.K. In addition, the new Looker MCP Server joins the Model Context Protocol (MCP) toolbox to streamline database orchestration.
Is the Age of the Citizen Data Scientist Here?
Google Clouds new Data Engineering Agent for BigQuery uses natural language to automate the entire pipeline creation process, from ingesting a CSV to cleansing columns and joining tables. It launches alongside a Spanner Migration Agent (currently in preview), an AI-powered service designed to simplify migrations from legacy systems like MySQL.
Likewise, the Data Science Agent can transform BigQuery notebooks into AI-powered labs. Users can ask it to analyze customer churn, and it will autonomously run exploratory data analysis, generate features, build models, and interpret the results.
For nontechnical users, a Conversational Analytics Agent comes equipped with a Code Interpreter. Powered by Geminis reasoning capabilities and developed with Google DeepMind, the interpreter simplifies querying to simple English questions. Users can ask, What were my top-selling products last month? or What were my sales last quarter? and receive detailed responsescomplete with Python code and visualizationswithout writing a single line of SQL.
Gemini CLI GitHub Actions also brings agentic AI directly into the developer workflow. Engineers can mention “@gemini-cli” in any issue or pull requests to delegate tasks. The upgraded CLI can write tests, implement suggestions, brainstorm alternative solutions, or fix well-defined bugs on demand.
Ryan J. Salva, a senior product director at Google Cloud, says developers appreciate having both targeted code review agents and more flexible, general-purpose ones. He expects more niche agents to emerge as the ecosystem grows. “Some tasks are better solved by an agent with access to unique problem-solving capabilities, specialized data, or models,” he tells Fast Company.
Since the launch of Gemini Code Assist for GitHub in February 2025, Salva says several major tech companies have seen significant improvements. Capgemini reported a notable increase in coding speed (up by 31%), Turing saw higher technical productivity and quicker delivery of quality output (30%), and Quantiphi experienced a substantial boost in overall developer efficiency (30%).
What makes this launch more than just a feature drop is the backend infrastructure powering it. Google Cloud has upgraded BigQuery and AlloyDB with autonomous vector indexing, hybrid semantic and keyword search, and adaptive filtering. The technologies are built on the same tech stack that supports Google Search and YouTube Ads.
We have built an intelligent system that understands how to efficiently prepare, index, and serve vector data at a petabyte scale. It’s the same systems-level thinking we use to index the web, now applied to corporate data, says Yasmeen Ahmad, managing director of StratOps and outbound product management at Google Cloud. Serving the right YouTube ad in milliseconds is a massive, real-time vector search problem. Weve taken that same low-latency, high-throughput infrastructure and built it into AlloyDB.
Under the hood, the Model Context Protocol (MCP) ensures that agents can securely interact with tools like GitHub, Looker, and custom APIs.
The Agentic AI Arms Race: Google vs. Everyone
With this launch, Google is making a bold play in the enterprise agentic AI race, aiming to leapfrog Snowflake, Databricks, and Microsoft by moving beyond code-generating assistants to full-fledged agents that can replace entire layers of routine data work.
Microsoft has a commanding lead in enterprise integration with its Copilot ecosystem, Databricks and Snowflake are strong in data-centric pipelines, but Googles strength lies in end-to-end orchestration and model tooling, not just data prep, says Arnie Bellini, the managing partner at Bellini Capital. He believes the differentiator will be whether Google can translate its developer-first tools into scalable, secure enterprise systems. Thats where Microsoft currently holds the edge, he adds.
While rivals focus on lakehouse unification and governed agentic orchestration, Google claims its edge is how deeply its agents ae embedded in its core cloud infrastructure.
Were not locking developers into a specific IDE or data lake. Because we know developers work with many tools and services in the course of a day, we aspire to build tools that are both open and extensible, says Salva. He believes the company is empowering developers to build and evolve their entire engineering systems using natural language.
Were building collaboratively through open source projects like Gemini CLI and Agent2Agent, he adds. By working in the open, we not only increase transparency, but also invite developers to help shape the future of dev tools.
Ahmad adds that Googles aim is to transform the entire data toolchain into an intelligent, conversational interface, not to merely augment existing tools.
Ahmad explains that, unlike other tools that primarily convert plain-language questions into SQL, Googles new Code Interpreter is designed for more advanced analytical tasks, such as forecasting and running what-if scenarios. Crucially, this isnt happening in a vacuum, she says. The future of data isnt about moving all your data to one central mega-lake. Its about a logically unified, but physically distributed, data ecosystem.
As part of this push, the company also announced new Oracle Cloud Infrastructure (OCI) locations in Japan, enabling in-region support for mission-critical applications in Tokyo, with Osaka support expected by early 2026. Ahmad notes that for many enterprises, particularly in regulated sectors like finance and healthcare, data residency is a legal and operational requirement. The new Oracle locations in Japan are a direct response to this.
But Craig Le Clair, vice president and principal analyst at Forrester Research, says Google still lags behind other platforms in comprehensive data supportcritical for broad enterprise adoption.
They also lack a strong action component to be able to build real business patterns that would provide ROI, he adds, noting that many current AI agent deployments resemble basic assistants rather than fully autonomous systems. “At this point, these efforts seem focused on search-oriented use cases, which again lack a strong action component and are unlikely to drive significant business value,” he says.
Agentic AI Has Promise, But It Lacks Magic
While agentic AI excites enterprises, many experts remain skeptical of its maturity. Although some organizations report double-digit efficiency gains, impact varies widely based on use case and implementation.
Most agent deployments are single agents handling specific task requests. The ROI heavily depends on good integration, training, and aligning the agents with actual workflows, says Jim Hare, vice president & analyst at Gartner. So while measurable gains are real, its not automatic as it requires the right groundwork to unlock the value.
He notes that true enterprise-grade maturitywith robust scalability, compliance, and securityis only beginning to emerge. Most current offerings are either developer-heavy tool kits requiring customization or narrowly scoped point solutions.
The big players like AWS, OpenAI, Microsoft, Google, Anthropic are actively building out platforms and ecosystems but broad deployment in production settings remains limited and use-case specific, says Hare.
Data governance also remains a concern. As agents gain more autonomy, enterprises will need stronger safeguards to prevent hallucinations or unintended actions. To address these risks, Google includes features like Workload Identity Federation and command allowlisting.
Our tools have native logging and monitoring to give you a clear audit trail of an agents actions and reasoning, says Seroter. We are also focused on building a secure-by-design environment with strong identity and access management. This allows enterprises to give agents just enough permission to do their job without becoming a security risk.
Experts caution that were still early in the maturity curve. As Bellini puts it, think of it like a modern air traffic control system. Its not about building the best plane, but coordinating the entire flight path, securing it, and landing safely at scale.
A Redefinition of Digital Labor
The agentic AI era signals a deeper shift in digital labor. As agents evolve from generating insights to initiating action, Googles message is clear: AI agents arent noveltiestheyre infrastructure. Hype aside, one thing is certain: the future of data work wont be entirely human.
In our view, humans are the strategic orchestrators and architects of a fleet of highly specialized agents, says Seroter. The goal is to give organizations the confidence to move forward, knowing that they can trust the systems theyre building and have a clear human-in-the-loop process for validation and oversight.
New research shows which jobs are likely to be most impacted by artificial intelligence, and the results may surprise you.
A report from Microsoft, titled, Working with AI: Measuring the Occupational Implications of Generative AI, takes a look at how workers in different fields adopt AI and how it will impact their jobs. Researchers analyzed what activities people used AI for in their jobs, how successful the results were, and what jobs the users had, in order to come up with an “AI applicability score.”
The tech giant collected data from 200,000 anonymous and privacy-scrubbed conversations between users and Microsoft Bing Copilot, a publicly available generative AI system, from January to September 2024.
What did the research find?
The most common work activities people used AI for involved gathering information and writing. When researchers computed an AI applicability score for each occupation, they found “the highest AI applicability scores for knowledge work occupation groups such as computer and mathematical, and office and administrative support, as well as occupations such as sales whose work activities involve providing and communicating information.”
Based on the findings, these are the top 10 occupations with the highest AI applicability scores:
Interpreters and Translators
Historians
Passenger Attendants
Sales Representatives of Services
Writers and Authors
Customer Service Representatives
CNC Tool Programmers
Telephone Operators
Ticket Agents and Travel Clerks
Broadcast Announcers and Radio DJs
The study also looked at the top 10 jobs and occupations least likely to be upended by AI. They are:
Phlebotomists
Nursing Assistants
Hazardous Materials Removal Workers
HelpersPainters, Plasterers
Embalmers
Plant and System Operators
Oral and Maxillofacial Surgeons
Automotive Glass Installers and Repairers
Ship Engineers
Tire Repairers and Changers
ESPN will continue to broadcast the NFL Draft as well as obtain new digital rights for its upcoming direct-to-consumer service.
The two agreements were announced Wednesday morning, two hours before the Walt Disney Company announced its third-quarter earnings.
It also came after the NFL and ESPN announced a nonbinding agreement Tuesday night in which ESPN will acquire the NFL Network and other league media assets while the NFL gets a 10% equity stake in ESPN.
ESPN has aired the NFL Draft since 1980, when the league’s annual selection meeting took place at the New York Sheraton Hotel. Back then, the draft was two days (Tuesday and Wednesday) and took 12 rounds.
Next year’s draft will be in Pittsburgh and is expected to attract massive crowds over the three days. The first round has had its own night since 2010.
ESPN and ABC will each have their own telecasts of the first three rounds on Thursday and Friday. ABC will simulcast ESPN’s coverage of the final four rounds on Saturday.
Besides ESPN’s direct-to-consumer service, Disney+ and Hulu will also stream the ESPN, ABC, and ESPN Deportes feeds under the multi-year agreement. The draft will also continue to be aired on NFL Network.
Weve been talking about the draft since last year and how we continue to build on that. ESPN has been a partner in that from day one, bringing, the fans closer to that event and building that event into one of the most popular events on the sporting calendar, which is incredible if you think back a few decades, NFL Commissioner Roger Goodell told The Associated Press. We know that relationship works, and were proud that ESPN is going to continue to be a partner.
ESPN will also produce a daily show leading up to the NFL Draft that will begin the day after the Super Bowl. That program will air most days on ESPN2, as well as being available on the direct-to-consumer service.
ESPN has also reached a licensing agreement that allows for additional NFL content and interactive features, including stats, fantasy football team performance and legalized sports betting information and tracking. It also allows ESPN to sell and bundle NFL+ Premium, the league’s direct-to-consumer product that includes out-of-market preseason games and replays of full games.
There will also be expanded highlight rights for the ESPN direct-to-consumer service and Disney+.
This will make the fan experience much stronger. The goal for ESPN when they launch the services is to create something that doesnt exist on linear (television) because the technology enables it,” Disney CEO Bob Iger said to AP. “Weve talked about personalization and personalized SportsCenter and the ability to essentially invent statistics and to tie betting to some of the programming.
Joe Reedy, AP sports writer
Business leaders have always had to be attentive to small but important shifts within the workplace that may affect employee performancetroubling trends that have increased since the pandemic. Now theres another problematic development for leaders to monitor. In addition to rising burnout, disengagement, and intentional idleness from quiet quitting, researchers have identified a new office condition theyre calling quiet cracking.
According to learning management system company TalentLMS, quiet cracking is situated somewhere between burnout, suffered by some ambitious but overloaded workers, and the quiet quitters who are actively slacking their way out of jobs they no longer want. Instead, people quietly cracking gradually become mired in feeling both unappreciated by managers and closed off from career advancement while doing work they otherwise like. The resulting unhappiness and frustration slowly builds until demotivated employees have to force themselves through the workday, causing their attention and productivity to drop.
Quiet cracking is the erosion of workplace satisfaction from within, according to a recent TalentLMS survey on the new threat to worker happiness and employer staffing stability and effectiveness. Unlike burnout, it doesnt always manifest in exhaustion. Unlike quiet quitting, it doesnt show up in performance metrics immediately. But it is just as dangerous.
The reason, the study said, is that a large portion of the workforce is already experiencing the persistent feeling of workplace unhappiness that leads to disengagement, poor performance, and an increased desire to quit that quiet crackers deal with. TalentLMSs survey of 1,000 U.S. employees found 54% saying theyd experienced one or several aspects of quiet cracking recently, with 20% saying theyd frequently or constantly battled these challenges.
Despite its rising prevalence and widespread effects, quiet cracking is tougher for employers to notice, because it develops gradually. Employees generally dont recognize initial dissatisfactions or frustrations as anything more significant than passing gripesuntil theyve become too deep and ingrained to shrug off. At that point, workers generally keep their problems to themselves while they start spinning their wheels doing jobs theyre losing interest in yet stick with, fearing it will be too difficult to find a new one.
Though they come to work on time each day and try to complete tasks as best they can, the malaise sufferers feel generally undermines their effectiveness. That creates another form of disengagement that a recent Gallup study said costs global businesses $8.8 trillion annually in lost productivity.
Is there any good news as quiet cracking emerges as another challenge to the workplace? TalentLMS says there is, with survey replies from people suffering from it offering ideas on how companies can prevent or remedy it.
Respondents typically said they didnt feel bosses appreciate them, dont listen or notice them, and arent providing any paths for advancing in their work and careers. Addressing those complaints is an obvious way to keep them from morphing into quiet cracking.
To do that, TalentLMS advises employers double down on learning and development and adopt the view that training is more than a skill-building toolits a confidence booster.Respondents who experienced quiet cracking said theyd received less direction and instruction at work in the prior year. The analysis portion of the survey urges companies to provide workers structured, ongoing learning paths. Businesses can also encourage staff to define some of the themes and content of those programs themselves, and not only have leaders make those programs available but also create time on the job that people can use to pursue them.
TalentLMS also urges employers to train managers who tend to shape company culture to regularly seek out feedback from employees. When possible, those consultations should be conducted in one-on-one meetings to allow staff to express their concerns more freelyespecially those contributing to any quiet cracking underway.
Finally, the study recommends publicly recognizing employee work and achievements as a low-cost, high-impact method for boosting workplace morale and self-esteem. That appreciation shouldnt be pro forma or forced, but can respond to even relatively routine efforts that nevertheless benefit the companys activities.
Those measures to address quiet cracking may seem like even more effort for employers already trying to minimize the instances and effects of burnout, disengagement, and other workplace challenges. But TalentLMS says constructive responses will be worth it in terms of staff satisfaction and productivity.Quiet cracking isnt just a well-being issueits a business issue, the study concludes. When employees quietly crack, they take productivity, creativity, and loyalty with them. Addressing quiet cracking doesnt require overhauling your entire strategybut it does require listening, acting, and investing.
By Bruce Crumley
This article originally appeared in Fast Companys sister publication, Inc.
Inc. is the voice of the American entrepreneur. We inspire, inform, and document the most fascinating people in business: the risk-takers, the innovators, and the ultra-driven go-getters that represent the most dynamic force in the American economy.
Just in time for the new school year, Google has introduced a tool called Guided Learning within its Gemini chatbot. Unlike tools that offer instant answers, Guided Learning breaks down complex problems step-by-step to support deeper understanding.The new feature is part of the AI industry’s broader response to growing concerns that chatbots like ChatGPT may undermine education by bypassing the learning process with quick answers. To learn more about Googles ambitions, strategy, and how its new Guided Learning tool works, Fast Company spoke with DeepMind COO Lila Ibrahim and Dave Messer, a Google product executive.
[Photo: Google]
Google recognizes that while students may sometimes need to look up facts quicklya strength of searchbots and chatbotsthey also benefit from AI that helps them reason through more complex topics. Our vision is really to have an AI tutor for every student and a TA for every teacher, says Dave Messer, a former teacher and now the product manager of Guided Learning. The tool mimics a human tutor by tailoring its approach to each students learning style, Messer says.
Similar to OpenAIs new Study mode in ChatGPT (announced last week), Guided Learning guides students through subjects using a conversational method. Instead of delivering answers outright, it leads users to insights through a series of thoughtful questions. These questions are designed to teach students the how and why behind a topic, encouraging learning throughout the exchange.It’s really about building your understanding and having a conversation that is like a thinking partner that can develop that thinking with you, and help make sure that you actually understand it instead of just seeing all the information, Messer says.There are always shortcuts to getting to answersAccording to Google, the learning session flow is backed by years of cognitive science research. For instance, Messer notes that research shows people learn best in bite-size chunks to avoid cognitive overload. The tool is designed to break down information in this way, gradually introducing students to a subjects depth and complexity.Students can ask the AI questions at any point. In response, Guided Learning usually offers contextual or clarifying details that nudge the student toward the answer, often reframing the question rather than solving it directly. If a student veers off-topic, the AI acknowledges their curiosity but gently redirects them to the core subject.[Animation: Google]Guided Learning is the product of two years of work on LearnLM, a family of models developed specifically for education. Ibrahim, DeepMind’s COO, says LearnLM was trained on educational best practices developed by leading experts. That learning was then integrated into Gemini, Googles flagship general-purpose model, which now powers Guided Learning.We spent a year or so making sure that it properly got infused into Gemini. [We] did all the testing to make sure that it was grounded with all the maths and sciences, the humanities that [are] appropriate for students, Ibrahim says, noting that the new tool is a natural evolution of Googles 20-year commitment to applying consumer techin this case, AIto education. I wouldn’t say we’re naive; we know that there are always shortcuts to getting to answers, Ibrahim says. We’re taking an important step with Guided Learning as a new way to bring Gemini to students in a more learning-appropriate environment [but] there’s still a lot of work to do.While most of Guided Learnings output is text-based, some prompts (like Whats the structure of a cell?) generate responses that include verified and licensed illustrations. The tool may also suggest relevant YouTube videos featuring knowledgeable educators explaining the concept in engaging ways.Students can upload class notes or syllabi to Gemini, allowing the tool to tailor interactions based on what is being studied. Messer explains that this helps the tool anticipate likely questions and generate customized quizzes or study guides from the notes.To use the tool Google AI subscribers can choose it from a row of option buttons at the bottom of the Gemini chat window. Guided Learning is rolling out globally on desktop and mobile devices to all Gemini users (with both free and paid accounts).Googles $1 billion AI education initiative for college studentsThe launch of Guided Learning is part of a broader education-focused AI initiative announced by Google on August 6. The company is committing $1 billion over three years to AI education and training efforts in the U.S.College students ages 18 and older in the U.S., Japan, Indonesia, Korea, and Brazil will receive free 12-month access to the Google AI Pro plan. This includes expanded access to the Gemini 2.5 Pro model, Deep Research, NotebookLM, Veo 3 video creation tools, and the Jules coding assistant.The Google AI for Education Accelerator will provide free AI training, Google Career Certificates, and advanced AI tools to college students nationwide. Google says more than 100 universities representing millions of students have already enrolled in the program.
