ChatGPT’s most advanced models recently served me a surprising statistic: US productivity grew faster in 2024 than in any year since the 1960s. Half that jump can be linked to generative AI tools that most workers hadn’t even heard of two years earlier.
Tools & Platforms
ChatGPT and Google Gemini can’t replace you. But they’re great coworkers.
The only problem is that it’s not true. The AI made it up.
Despite its much-documented fallibility, generative AI has become a huge part of many people’s jobs, including my own. The numbers vary from survey to survey, but a June Gallup poll found that 42 percent of American employees are using AI a few times a year, while 19 percent report deploying it several times a week. The technology is especially popular with white-collar workers. While just 9 percent of manufacturing and front-line workers use AI on a regular basis, 27 percent of white-collar workers do.
Even as many people integrate AI into their daily lives, it’s causing mass job anxiety: A February Pew survey found that more than half of US employees worried about their fate at work.
Unfortunately, there is no magic trick to keep your job for the foreseeable future, especially if you’re a white-collar worker. Nobody knows what’s going to happen with AI, and leadership at many companies is responding to this uncertainty by firing workers it may or may not need in an AI-forward future.
“If AI really is this era’s steam engine, a force so transformative that it will power a new Industrial Revolution, you only stand to gain by getting good at it.”
After laying off over 6,000 workers in May and June, Microsoft is laying off 9,000 more workers this month, reportedly so the company can reduce the number of middle managers as it reorganizes itself around AI. In a note on Tuesday, Amazon CEO Andy Jassy told employees that the company would “roll out more generative AI and agents” and reduce its workforce in the next few years. This was all after Anthropic CEO Dario Amodei warned AI would wipe out half of all entry-level white-collar jobs in the same timespan, a prediction so grim that Axios coined a new term for AI’s imminent takeover: “a white-collar bloodbath.”
This is particularly frustrating because, as my recent encounter with ChatGPT’s tendency to hallucinate makes clear, the generative AI of today, while useful for a growing number of people, needs humans to work well. So does agentic AI, the next era of this technology that involves AI agents using computers and performing tasks on your behalf rather than simply generating content. For now, AI is augmenting white-collar jobs, not automating them, although your company’s CEO is probably planning for the latter scenario.
Maybe one day AI will fulfill its promise of getting rid of grunt work and creating endless abundance, but getting from here to there is a harrowing proposition.
“With every other form of innovation, we ended up with more jobs in the end,” Ethan Mollick, a Wharton professor and author of the newsletter One Useful Thing, told me. “But living through the Industrial Revolution still kind of sucked, right? There were still anarchists in the street and mass displacement from cities and towns.”
We don’t know if the transition to the AI future will be quite as calamitous. What we do know is that just as jobs transformed due to past technological leaps, like the introduction of the personal computer or the internet, your day-to-day at work will change in the months and years to come. If AI really is this era’s steam engine, a force so transformative that it will power a new Industrial Revolution, you only stand to gain by getting good at it.
At the same time, becoming an AI whiz will not necessarily save you if your company decides it’s time to go all in on AI and do mass, scattershot layoffs in order to give its shareholders the impression of some efficiency gains. If you’re impacted, that’s just bad luck. Still, having the skills can’t hurt.
Welcome to the AI revolution transition
It’s okay to be scared of AI, but it’s more reasonable to be confused by it. For two years after ChatGPT’s explosive release, I couldn’t quite figure out how a chatbot could make my life better. After some urging from Mollick late last year, I forced myself to start using it for menial chores. Upgrading to more advanced models of ChatGPT and Claude turned these tools into indispensable research partners that I use every day — not just to do my job faster but also better. (Disclosure: Vox Media is one of several publishers that have signed partnership agreements with OpenAI. Our reporting remains editorially independent.)
But when it comes to generative AI tools and the burgeoning class of AI agents, what works for one person might not be helpful to the next.
“Workers obviously need to try to ascertain as much as they can — the skills that are most flexible and most useful,” said Mark Muro, a senior fellow at Brookings Metro. “They need to be familiar with the technology because it is going to be pervasive.”
For most white-collar workers, I recommend Mollick’s 10-hour rule: Spend 10 hours using AI for work and see what you learn. Mollick also recently published an updated guide to the latest AI tools that’s worth reading in full. The big takeaways are that the best of these tools (ChatGPT from OpenAI, Claude from Anthropic, and Google Gemini) can become tireless assistants with limitless knowledge that can save you hours of labor. You should try different models within the different AI tools, and you should experiment with the voice features, including the ability to use your phone’s camera to share what you’re seeing with the AI. You should also, unfortunately, shell out $20 a month to get access to the most advanced models. In Mollick’s words, “The free versions are demos, not tools.”
“If I have a very narrow job around a very narrow task that’s being done repetitively, that’s where the most risk comes in.”
You can imagine similar advice coming from your geeky uncle at Thanksgiving circa 1984, when personal computers were on the brink of taking over the world. That was the year roughly the same percentage of white-collar workers were regularly using PCs at work as are using AI today. But the coming AI transition will look different than the PC transition we’ve already lived through. While earlier digital technologies hit frontline workers hardest, “AI excels at supporting or carrying out the highly cognitive, nonroutine tasks that better-educated, better-paid office workers do,” according to a February Brookings report co-authored by Muro.
This means AI can do a lot of the tasks that software engineers, architects, lawyers, and journalists do, but it doesn’t mean that AI can do their jobs — a key distinction. This is why you hear more experts talking about AI augmentation rather than AI automation. As a journalist, I can confidently say that AI is great at streamlining my research process, saving me time, and sometimes even stirring up new ideas. AI is terrible at interviewing sources, although that might not always be the case. And clearly, it’s touch-and-go when it comes to writing factually accurate copy, which is kind of a fundamental part of the job.
That proposition looks different for other kinds of white-collar work, namely administrative and operational support jobs. A Brookings report last year found that 100 percent of the tasks that bookkeepers and clerks do were likely to be automated. Those of travel agents, tax preparers, and administrative assistants were close to 100 percent. If AI really did make these workers redundant, it would add up to millions of jobs lost.
“The thing I’d be most worried about is if my task and job are very similar to each other,” Mollick, the Wharton professor, explained. “If I have a very narrow job around a very narrow task that’s being done repetitively, that’s where the most risk comes in.”
It’s hard to AI-proof your job or career altogether given so much uncertainty. We don’t know if companies will take advantage of this transition in ways that produce better products and happier workers or just use it as an excuse to fire people, squandering what some believe is a once-in-a-generation opportunity to transform work and productivity. It sucks to feel like you have little agency in steering the future toward one outcome or the other.
At the risk of sounding like your geeky uncle, I say give AI a try. The worst-case scenario is you spend 10 hours talking to an artificially intelligent chatbot rather than scrolling through Instagram or Reddit. The best-case scenario is you develop a new skill set, one that could very well set you up to do an entirely new kind of job, one that didn’t even exist before the AI era. You might even have a little fun along the way.
A version of this story was also published in the User Friendly newsletter. Sign up here so you don’t miss the next one!
Tools & Platforms
AI Shopping Is Here. Will Retailers Get Left Behind?
AI doesn’t care about your beautiful website.
Visit any fashion brand’s homepage and you’ll see all sorts of dynamic or interactive elements from image carousels to dropdown menus that are designed to catch shoppers’ eyes and ease navigation.
To the large language models that underlie ChatGPT and other generative AI, many of these features might as well not exist. They’re often written in the programming language JavaScript, which for the moment at least most AI struggles to read.
This giant blindspot didn’t matter when generative AI was mostly used to write emails and cheat on homework. But a growing number of startups and tech giants are deploying this technology to help users shop — or even make the purchase themselves.
“A lot of your site might actually be invisible to an LLM from the jump,” said A.J. Ghergich, global vice president of Botify, an AI optimisation company that helps brands from Christian Louboutin to Levi’s make sure their products are visible to and shoppable by AI.
The vast majority of visitors to brands’ websites are still human, but that’s changing fast. US retailers saw a 1,200 percent jump in visits from generative AI sources between July 2024 and February 2025, according to Adobe Analytics. Salesforce predicts AI platforms and AI agents will drive $260 billion in global online sales this holiday season.
Those agents, launched by AI players such as OpenAI and Perplexity, are capable of performing tasks on their own, including navigating to a retailer’s site, adding an item to cart and completing the checkout process on behalf of a shopper. Google’s recently introduced agent will automatically buy a product when it drops to a price the user sets.
This form of shopping is very much in its infancy; the AI shopping agents available still tend to be clumsy. Long term, however, many technologists envision a future where much of the activity online is driven by AI, whether that’s consumers discovering products or agents completing transactions.
To prepare, businesses from retail behemoth Walmart to luxury fashion labels are reconsidering everything from how they design their websites to how they handle payments and advertise online as they try to catch the eye of AI and not just humans.
“It’s in every single conversation I’m having right now,” said Caila Schwartz, director of consumer insights and strategy at Salesforce, which powers the e-commerce of a number of retailers, during a roundtable for press in June. “It is what everyone wants to talk about, and everyone’s trying to figure out and ask [about] and understand and build for.”
From SEO to GEO and AEO
As AI joins humans in shopping online, businesses are pivoting from SEO — search engine optimisation, or ensuring products show up at the top of a Google query — to generative engine optimisation (GEO) or answer engine optimisation (AEO), where catching the attention of an AI responding to a user’s request is the goal.
That’s easier said than done, particularly since it’s not always clear even to the AI companies themselves how their tools rank products, as Perplexity’s chief executive, Aravind Srinivas, admitted to Fortune last year. AI platforms ingest vast amounts of data from across the internet to produce their results.
Though there are indications of what attracts their notice. Products with rich, well-structured content attached tend to have an advantage, as do those that are the frequent subject of conversation and reviews online.
“Brands might want to invest more in developing robust customer-review programmes and using influencer marketing — even at the micro-influencer level — to generate more content and discussion that will then be picked up by the LLMs,” said Sky Canaves, a principal analyst at Emarketer focusing on fashion, beauty and luxury.
Ghergich pointed out that brands should be diligent with their product feeds into programmes such as Google’s Merchant Center, where retailers upload product data to ensure their items appear in Google’s search and shopping results. These types of feeds are full of structured data including product names and descriptions meant to be picked up by machines so they can direct shoppers to the right items. One example from Google reads:
Ghergich said AI will often read this data before other sources such as the HTML on a brand’s website. These feeds can also be vital for making sure the AI is pulling pricing data that’s up to date, or as close as possible.
As more consumers turn to AI and agents, however, it could change the very nature of online marketing, a scenario that would shake even Google’s advertising empire. Tactics that work on humans, like promoted posts with flashy visuals, could be ineffective for catching AI’s notice. It would force a redistribution of how retailers spend their ad budgets.
Emarketer forecasts that spending on traditional search ads in the US will see slower growth in the years ahead, while a larger share of ad budgets will go towards AI search. OpenAI, whose CEO, Sam Altman, has voiced his distaste for ads in the past, has also acknowledged exploring ads on its platform as it looks for new revenue streams.

“The big challenge for brands with advertising is then how to show up in front of consumers when traditional ad formats are being circumvented by AI agents, when consumers are not looking at advertisements because agents are playing a bigger role,” said Canaves.
Bots Are Good Now
Retailers face another set of issues if consumers start turning to agents to handle purchases. On the one hand, agents could be great for reducing the friction that often causes consumers to abandon their carts. Rather than going through the checkout process themselves and stumbling over any annoyances, they just tell the agent to do it and off it goes.
But most websites aren’t designed for bots to make purchases — exactly the opposite, in fact. Bad actors have historically used bots to snatch up products from sneakers to concert tickets before other shoppers can buy them, frequently to flip them for a profit. For many retailers, they’re a nuisance.
“A lot of time and effort has been spent to keep machines out,” said Rubail Birwadker, senior vice president and global head of growth at Visa.
If a site has reason to believe a bot is behind a transaction — say it completes forms too fast — it could block it. The retailer doesn’t make the sale, and the customer is left with a frustrating experience.
Payment players are working to create methods that will allow verified agents to check out on behalf of a consumer without compromising security. In April, Visa launched a programme focused on enabling AI-driven shopping called Intelligent Commerce. It uses a mix of credential verification (similar to setting up Apple Pay) and biometrics to ensure shoppers are able to checkout while preventing opportunities for fraud.
“We are going out and working with these providers to say, ‘Hey, we would like to … make it easy for you to know what’s a good, white-list bot versus a non-whitelist bot,’” Birwadker said.
Of course the bot has to make it to checkout. AI agents can stumble over other common elements in webpages, like login fields. It may be some time before all those issues are resolved and they can seamlessly complete any purchase.
Consumers have to get on board as well. So far, few appear to be rushing to use agents for their shopping, though that could change. In March, Salesforce published the results of a global survey that polled different age groups on their interest in various use cases for AI agents. Interest in using agents to buy products rose with each subsequent generation, with 63 percent of Gen-Z respondents saying they were interested.
Canaves of Emarketer pointed out that younger generations are already using AI regularly for school and work. Shopping with AI may not be their first impulse, but because the behaviour is already ingrained in their daily lives in other ways, it’s spilling over into how they find and buy products.
More consumers are starting their shopping journeys on AI platforms, too, and Schwartz of Salesforce noted that over time this could shape their expectations of the internet more broadly, the way Google and Amazon did.
“It just feels inevitable that we are going to see a much more consistent amount of commerce transactions originate and, ultimately, natively happen on these AI agentic platforms,” said Birwadker.
Tools & Platforms
CarMax’s top tech exec shares his keys to reinventing a legacy retailer in the age of AI
More than 30 years ago, CarMax aimed to transform the way people buy and sell used cars with a consistent, haggle-free experience that separated it from the typical car dealership.
Despite evolving into a market leader since then, its chief information and technology officer, Shamim Mohammad, knows no company is guaranteed that title forever; he had previously worked for Blockbuster, which, he said, couldn’t change fast enough to keep up with Netflix in streaming video.
Mohammad spoke with Modern Retail at the Virginia-based company’s technology office in Plano, Texas, which it opened three to four years ago to recruit for tech workers like software engineers and analysts in the region home to tech companies such as AT&T and Texas Instruments. At that office, CarMax has since hired almost 150 employees — more than initially expected — including some of Mohammad’s former colleagues from Blockbuster, which he had worked for in Texas in the early 2000s.
He explained how other legacy retailers can learn from how CarMax leveraged new technology like artificial intelligence and a startup mindset as it embraced change, becoming an omnichannel retailer where customers can buy cars in person, entirely online or through a combination of both. Many customers find a car online and test-drive and complete their purchase at the store.
“Every company, every industry is going through a lot of disruption because of technology,” Mohammad said. “It’s much better to do self-disruption: changing our own business model, challenging ourselves and going through the pain of change before we are disrupted by somebody else.”
Digitizing the dealership
Mohammad has been with CarMax for more than 12 years and had also been vp of information technology for BJ’s Wholesale Club. Since joining the auto retailer, he and his team have worked to use artificial intelligence to fully digitize the process of car buying, which is especially complex given the mountain of vehicle information and regulations dealers have to consider.
He said the company has been using AI and machine learning for at least 12-13 years to price cars, make sure the right information is online for the cars, and understand where cars need to be in the supply chain and when. That, he said, has powered the company’s website in becoming a virtual showroom that helps customers understand the vehicles, their functions and how they fit their needs. Artificial intelligence has also powered its online instant offer tool for selling cars, giving customers a fair price that doesn’t lose the company money, Mohammad said.
“Technology is enabling different types of experiences, and it’s setting new expectations, and new types of ways to shop and buy. Our industry is no different. We wanted to be that disruptor,” Mohammad said. “We want to make sure we change our business model and we bring those experiences so that we continue to remain the market leader in our industry.”
About three or four years ago, CarMax was an early adopter of ChatGPT, using it to organize data on the different features of car models and make it presentable through its digital channels. Around the same time, the company also used generative AI to comb through and summarize thousands of customer product reviews — it did what would have taken hundreds of content writers more than 10 years to do in a matter of days, he said — and keep them up to date.
As the technology has improved over the last few years, the company has adopted several new AI-powered features. One is Rhodes, a tool associates use to get support and information they need to help customers, which launched about a year ago, Mohammad said. It uses a large language model combining CarMax data with outside information like state or federal rules and regulations to help employees quickly access that data.
Anything that requires a lot of human workload and mental capacity can be automated, he said, from looking at invoices and documents to generating code for developers and engineers, saving them time to do more valuable work. Retailers like Target and Walmart have done the same by using AI chatbots as tools for employees.
“We used to spend a fortune on employee training, and employees only retained and reliably repeated a small percentage of what we trained,” said Jason Goldberg, chief commerce strategy officer for Publicis Groupe. “Increasingly, AI is letting us give way better tools to the salespeople, to train them and to support them when they’re talking to customers.”
In just the last few months, Mohammad said, CarMax has been rolling out an agentic version of a previous buying and selling assistant on its website called Skye that better understands the intent of the user — not only answering the question the customer asks directly, but also walking the customer through the entire car buying process.
“It’ll obviously answer [the customer’s question], but it will also try to understand what you’re trying to do and help you proactively through the entire process. It could be financing; it could be buying; it could be selling; it could be making an appointment; it could be just information about the car and safety,” he said.
The new Skye is more like talking to an actual human being, Mohammad said, where, in addition to answering the question, the agent can make other recommendations in a more natural conversation. For example, if someone is trying to buy a car and asks for a family car that’s safe, it will pull one from its inventory, but it may also ask if they’d like to talk to someone or even how their day is going.
“It’s guiding you through the process beyond what you initially asked. It’s building a rapport with you,” Mohammad said. “It knows you very well, it knows our business really well, and then it’s really helping you get to the right car and the right process.”
Goldberg said that while many functions of retail, from writing copy to scheduling shifts, have also been improved with AI, pushing things done by humans to AI chatbots could lead to distrust or create results that are inappropriate or offensive. “At the moment, most of the AI things are about efficiency and reducing friction,” Goldberg said. “They’re taking something you’re already doing and making it easier, which is generally appealing, but there is also the potential to dehumanize the experience.”
In testing CarMax’s new assistant, other AI agents are actually monitoring it to make sure it’s up to the company’s standards and not saying bad words, Mohammad said, adding it would be impossible for humans to look at everything the new assistant is doing.
The company doesn’t implement AI just to implement AI, Mohammad said, adding that his teams are using generative AI as a tool when needing to solve particular problems instead of being forced to use it.
“Companies don’t need an AI strategy. … They need a strategy that uses AI,” Mohammad said. “Use AI to solve customer problems.”
Working like a tech startup
In embracing change, CarMax has had to change the way it works, Mohammad said. It has created a more startup-like culture, going from cubicles to more open, collaborative office spaces where employees know what everyone else is working on.
About a decade ago, he said, the company started working with a project-based mindset, where it would deliver a new project every six to nine months — each taking about a year in total, with phases for designing and testing.
Now, the company has small, cross-functional product teams of seven to nine people, each with a mission around improving a particular area like finance, digital merchandising, SEO, logistics or supply chain — some even have fun names like “Ace” or “Top Gun.”
Teams have just two weeks to create a prototype of a feature and get it in front of customers. He said that, stacked up over time, those small new changes those teams created completely transformed the business.
“The teams are empowered, and they’re given a mission. I’m not telling them what to do. I’m giving them a goal. They figure out how,” Mohammad said. “Create a culture of experimentation, and don’t wait for things to be perfect. Create a culture where your teams are empowered. It’s OK for them to make mistakes; it’s OK for them to learn from their mistakes.”
Tools & Platforms
Available Infrastructure Unveils ‘SanQtum’ Secure AI Platform for Critical Infrastructure
Available Infrastructure (Available) publicly unveiled SanQtum, a first-of-a-kind solution that combines national security-grade cyber protection and the world’s most-trusted enterprise artificial intelligence (AI) capability.
In the modern era, AI-powered, machine-speed decision-making is crucial. Yet a fast-evolving and increasingly sophisticated threat landscape puts operational technology (OT) and cyber-physical systems (CPS), IP and other sensitive data, and proprietary trained AI models at risk. SanQtum is a direct response to that need.
Created through a rigorous development process in collaboration with major enterprise tech partners and government agencies, SanQtum pre-integrates a best-in-breed tech stack in a micro edge data center form factor, ready for deployment anywhere — from near-prem urban sites to telecom towers to austere environments. A first cohort of initial sites is already under construction in Northern Virginia and expected to come online later this year.
SanQtum’s cybersecurity protections include zero trust permissions architecture, quantum-resilient data encryption, and are aligned to DHS, CISA, and other US federal cybersecurity standards. Sovereign AI models with ultra-low-latency computing enable secure decision-making at machine speed when milliseconds matter, wrapped in cyber protections to prevent data theft and AI model poisoning.
The need for more sophisticated cybersecurity solutions is widespread and growing by the day. Globally, the cost of cybercrimes to corporations is forecasted to nearly triple, from $8 trillion in 2023 to $23 trillion by 2027. For government agencies and critical infrastructure, cybersecurity is literally a matter of life and death.
Daniel Gregory, CEO of Available
AI is now seemingly everywhere. So are cyber threats, from nation-state attacks to criminal enterprises. In this environment, decision-making without AI — and AI without cybersecurity protections — are no longer negotiable; they’re mandatory. As we head into the July 4th weekend, which has historically seen a surge in cyber attacks each year, security is top-of-mind for many Americans, businesses, and government agencies. We live in a digital world. And AI is now seemingly everywhere. So are cyber threats, from nation-state attacks to criminal enterprises. In this environment, decision-making without AI — and AI without cybersecurity protections — are no longer negotiable; they’re mandatory.
-
Funding & Business6 days ago
Kayak and Expedia race to build AI travel agents that turn social posts into itineraries
-
Jobs & Careers6 days ago
Mumbai-based Perplexity Alternative Has 60k+ Users Without Funding
-
Mergers & Acquisitions6 days ago
Donald Trump suggests US government review subsidies to Elon Musk’s companies
-
Funding & Business6 days ago
Rethinking Venture Capital’s Talent Pipeline
-
Jobs & Careers6 days ago
Why Agentic AI Isn’t Pure Hype (And What Skeptics Aren’t Seeing Yet)
-
Funding & Business3 days ago
Sakana AI’s TreeQuest: Deploy multi-model teams that outperform individual LLMs by 30%
-
Funding & Business6 days ago
From chatbots to collaborators: How AI agents are reshaping enterprise work
-
Jobs & Careers6 days ago
Telangana Launches TGDeX—India’s First State‑Led AI Public Infrastructure
-
Jobs & Careers3 days ago
Ilya Sutskever Takes Over as CEO of Safe Superintelligence After Daniel Gross’s Exit
-
Funding & Business3 days ago
Dust hits $6M ARR helping enterprises build AI agents that actually do stuff instead of just talking