 |
|
|
|||||
The Ford Mustang Mach-E cruises down a London road choked with traffic, using its onboard AI system to avoid jaywalkers and cyclists, and navigate roadwork as it drives to its destination.The autonomous vehicle from British startup Wayve Technologies is on a test run ahead of the U.K. government’s robotaxi trials set to launch in the spring. Tech companies including U.S. company Waymo and China’s Baidu also plan to take part in the pilot program, making London the latest arena in the global robotaxi competition.While self-driving cabs aren’t new, London’s ancient road layout and busy streetscapes could pose special challenges for the technology.There’s also skepticism from London’s famed black cab drivers, who must pass a grueling training course known as “The Knowledge,” which requires memorizing hundreds of routes and takes years to complete. They’ve previously opposed technology that’s disrupted their industry, and protested the arrival of Uber.Self-driving taxis are “a solution looking for a problem,” said Steven McNamara, general secretary of the Licensed Taxi Drivers’ Association, which represents black cabbies.He doubts that robotaxis would have any advantage on London’s road network, which is laid out in a convoluted spiderweb that dates back to Roman times unlike the grid layout in American cities like San Francisco and Phoenix where Waymo operates.The British capital is notorious for being one of the world’s most congested cities and its streets are already clogged with other modes of transport, including private cars, buses, motor scooters, bicycles and electric rental bikes.McNamara and many others have noted that robotaxis face another challenge from pedestrians crossing the streets. While jaywalking is illegal in the United States and many other countries, it’s not an offense in Britain.“It’s virtually impossible to drive anywhere (in London) without somebody walking in front of you,” McNamara said. In London, with a population of nearly 10 million, he wondered “how these cars are going to deal with those volumes of people?”The robotaxi companies say there’s room for the new technology.“I think Londoners are going to love autonomous driving. It’s going to be another choice alongside the Tube, cycling, walking, “said Wayve CEO Alex Kendall in a recent interview at the company’s workshop.Wayve is teaming up with Uber for the taxi trials, which are part of Britain’s move to adopt national regulations for self-driving vehicles. The nation is seeking to position itself as a world leader in the technology.Chinese tech company Baidu is also teaming up with Uber, as well as its ride-hailing rival Lyft, to operate its Apollo Go autonomous vehicle service in the London pilot.Waymo, owned by Google parent Alphabet, will also take part and plans to launch a London passenger service by the third quarter of 2026, company representatives told reporters last month.Waymo officials sought to ease concerns that the company would suddenly flood London streets with robotaxis, noting that it has operated 1,000 total vehicles in San Francisco since going into full service in 2024.“We’re not here to replace anyone,” Waymo spokesman Ethan Teicher said. “We’re here to add another option for people who will choose to take black cabs or other modes of transportation when it suits them and choose to take Waymo, when it makes sense.”Waymo’s self-driving Jaguar I-Pace sedans have been spotted doing test runs around London. Wayve’s Ford Mustang Mach-E vehicles have also been doing road tests with human backup drivers sitting behind the wheel, ready to intervene if needed.On a recent demo ride for The Associated Press, Wayve’s Ford steered automatically through a three-mile (five kilometer) loop in North London without any problems.Cruising down a straight and open stretch of road, the car maintained a steady pace of 19 miles (30 kilometers) per hour, a tick under the speed limit.A traffic light changed as the car approached, forcing it to brake firmly and lightly jolting the passengers forward the only moment that the driving was less than smooth.Kendall said Wayve takes a different approach from traditional self-driving technology. It doesn’t rely on “high definition” maps and “hand-coded” safety systems rules written by programmers anticipating every scenario.Instead, it uses an AI trained on millions of hours of data gathered by its cars to learn and understand how the world works.“This is the key thing for self-driving, because every time you drive on the road, you’re going to experience something different,” Kendall said. “You can’t rely on a self-driving car being told how to behave in every scenario it encounters.”He said Wayve is positioning itself as a technology company providing hardware and software that can be added to any vehicle to make it autonomous. It signed a deal with Nissan in December to build self-driving cars that will go on sale in Japan and North America by 2027.Kendall wouldn’t reveal any more specific details about the robotaxi service it will operate in collaboration with Uber, such as pricing.Waymo, which has its own app to hail rides, will have “competitive” prices and fares will be in line with the market, officials said last month, while adding that it is often able to “demand more premium pricing.”Experts say there’s a role for robotaxis in Britain, but it might be a niche one.They’re best poised to fill gaps in Britain’s public transport network, such as serving villages that have lost bus services connecting them to bigger towns and cities because of budget cuts, said Kevin Vincent, director of the Centre for Connected and Autonomous Automotive Research at Coventry University.There will still be demand for human drivers, especially from out-of-town visitors and tourists, he said.If you find a “cab driver who knows the area, you can ask him questions. You feel confident and comfortable you’re going where you need to go,” which is a service that won’t be easily replaced in the short term, Vincent said.Self-driving taxis can’t replicate the human touch, said Frank O’Beirne, who has been driving black cabs for 14 years.For example, one of his recent fares was a pair of blind passengers going to touristy Leicester Square. He ended up parking at a cab rank and walking them across the street to their destination, a Chinese restaurant that turned out to be in the basement of a casino.“They would never have found that, ever, (on their own),” said O’Beirne. “There’s nothing like us. I can’t see the space where autonomous taxis can operate, really.” Kelvin Chan, AP Business Writer
Category:
E-Commerce
For the last several years, enterprises have treated AI as something to test. A pilot here, a proof of concept there. That era is ending. According to new global DeepL research, a survey of 5,000 global executives on the impact of AI agents reshaping business, 69% expect AI agents to fundamentally change how their companies operate in 2026. Nearly half anticipate major transformation, while another quarter say that change is already underway. This moment didnt arrive overnight. While 2025 was the year agentic AI moved from theory to application, enterprises are making the shift structural this year. Leaders are no longer asking whether AI works but rather deciding where AI agents belong inside their operating model. As tools mature and agentic systems become capable of coordinating work across functions, AI agents are unlocking new opportunitiesnot just automating tasks. By eliminating manual coordination, AI agents enable organizations to move faster and smarter, enabling the creativity, problem-solving depth, and judgment that turn velocity into measurable value. Yet, when it comes to scaling agents and validating their investment, most organizations remain stuck in pilot mode. McKinsey reports that while 62% of companies are experimenting with agents, only 10% are scaling them across a single function, and just 32% of leaders report an impact on EBIT at the enterprise level. The gap between early adopters and those who hold out will widen in 2026not because of trial and error, but execution. Three shifts will define this gaphow enterprises automate core operations, deploy AI for growth, and build the communication infrastructure agents require. AUTOMATE THE CORE AI agents are no longer confined to experimentation or pilots. Enterprises are deploying them into operational workflows like processing returns in customer service, investigating customer complaints, automating approvals and ticketing, supporting prospect and competitor research in sales and marketing, and optimizing working capital in finance. Whats changing is continuity. Instead of accelerating individual tasks, organizations are increasingly making agents responsible for managing handoffs between themreducing friction. Looking at AI agent adoption more broadly, DeepLs research shows global executives cite proven ROI and efficiency (22%), workforce adaptability (18%), and enterprise readiness (18%) as the primary reasons they feel confident expanding agent deployment. Results, not optimism, are driving this shift. At the same time, known barriers are beginning to soften. Cost (16%), workforce preparedness (13%), and technology maturity (12%) remain challenges, but enterprises are actively addressing them as they gain experience operating agents in production environments. The real risk now is inaction. Organizations that fail to identify which workflows should be automated first keep valuable talent focused on low-leverage workwhile competitors redesign operations around intelligent systems. Customer service offers a clear example. Companies like Perk are deploying AI agents to take on routine operational work in customer support, while human agents focus on complex, relationship-driven scenarios. As Tom Davis, senior director of operational excellence at Perk, notes: When weve got travelers stuck in airports, we want our humans focused on those momentsand in the background, a machine of AI handling the grunt work. That division allows human agents to focus on high-stakes relationship work while AI agents manage operational tasks at scale. AI AS A GROWTH ENGINE AI is no longer confined to cost reduction. Its becoming a driver of growth. The broader AI landscape shows strong momentum: 67% of executives reported measurable impact from AI initiatives in 2025, and 52% expect AI to contribute more to company growth than any other technology in 2026. Enterprises seeing the strongest returns are applying AI across revenue-generating functionscustomer service, marketing, sales, finance, legal, HR, and IT supportrather than limiting it to back-office automation. The competitive advantage comes from scale and integration, not just isolated use cases. But the real gain isnt just efficiencyits faster, higher-quality decision making. As b2ventures noted in their work with AI agents, the technology helps them make higher-quality investment decisions faster because agents excel at evaluating companies and surfacing insights that inform critical choices. According to DeepLs research, leaders in the UK (80%), Germany (78%), and the U.S. (71%) are seeing measurable performance gains from AI initiatives. This underscores that execution and organizational readiness are just as important as access to technology when it comes to turning AI into a strategic advantage. Ignoring AI in growth-critical areas is no longer conservative. Its a strategic risk, particularly in sectors where margins and customer expectations are shifting fast. LANGUAGE AND VOICE AI As AI agents move deeper into enterprise workflows, theyre changing how people interact with software itself. Instead of clicking through dashboards or submitting forms, employees increasingly instruct systems through natural language. In an agent-driven operating model, language becomes the primary user interface. This is the mechanism through which work gets done. That shift raises the stakes for fluency and accuracy. When language is the interface, a mistranslated prompt or misunderstood instruction doesnt just slow down communication; it can derail an entire workflow. For enterprises scaling agents across teams and regions, language precision becomes a requirement, not a nice-to-have. This is reflected in enterprise priorities where 64% of companies plan to increase investment in language AI in 2026, while organizations expect adoption of real-time voice translation to rise to 54%. These investments arent standalone initiatives. They are foundational to making AI agents reliable, scalable, and effective. EXECUTION, NOT EXPERIMENTATION This year, the organizations experiencing the biggest impacts will stop experimenting with AI and start embedding AI agents into core operations and applying them across growth-critical functions. By turning AI into a strategic advantage, these companies will streamline operations, make better decisions, and unlock measurable business value. Those who delay will watch the gap grow as early adopters accelerate ahead. Jarek Kutylowski is the CEO and founder of DeepL.
Category:
E-Commerce
Before the holidays, Adam Conner began vibe coding. Like everyone else in the know, he was using Claude Code. Compared to popular chatbots, Anthropics advanced AI agent speaks the language of computers: code. Normally, you click buttons in browsers, open folders, and drag files. But you can also do so by codinginteracting with software by typing commands into a terminal, a text-based app. Claude Code goes beyond such primitive tasks, though: an AI that can code can effectively do nearly anything on a computer. We expected developers to use Claude Code for coding, but then something unexpected happened, an Anthropic spokesperson tells Fast Company. We started seeing the discovery arc where people would approach Claude Code to tackle a coding task, then have an ‘aha moment’ when they realized it could help with other tasks. The result of that aha moment is a vibe coding phenomenon that lets developersand, crucially, non-developers such as Connerharness the AI agent to write code and create projects that could grow tomato plants, knit sweaters, and build fully fledged iOS apps in hours. Conners vibe coding wasnt as dramatic as keeping living organisms alive, but no less impressive. He used it for his work: an AI labor market simulator, built from Bureau of Labor Statistics and Federal Reserve data. It projects the potential impact that AI could have on the economy. (Headline: its pretty big.) I got it running in a day, says Conner, whos also the vice president for technology policy at the Center of American Progress (CAP), a think-tank. It shows how this AI is more accessible and powerful. You can have limited programming experience, and still build something quickly. But the real revelation, says Conner, came using Claude Code to form his own policy council: 21 AI agents with competing ideologies and political agendas. Within minutes, it had generated 24-page proposal papers, 12-page draft legislations, and hundreds of policy ideas. You can now direct a small army of bots to do the tasks of humans relatively quickly, says Conner. Its not yet the game changer that can fully automate someones job, but you can begin to see how AI could be transformational. Were now in the fourth year of widely available generative AI tools. As adoption has ticked up23% of U.S. workers are now frequent AI users, nearly doubling year-over-yearso too have AI-related layoffs. Despite many companies citing AI as the cause for layoffs, though, actual mass job displacement due to AI has yet to materialize. But the capabilities of Anthropics Claude Codeand in particular its advanced AI workplace tool, Claude Coworkcould change that. The new knowledge worker So far, generative AIs use case has been split fairly equally between work and personal needs, such as generating ideas, editing, and even companionship. In 2025, bots were more often leveraged for therapy, life coaching, and, increasingly, codingthe latter likely driven by the popularity of Claude Code: generating more than $1 billion in revenue, just six months after general availability, according to Anthropic. On January 12, the San Francisco-headquartered company launched Claude Cowork. Its effectively a UI update to Code for the mass market. Whereas Code can only be utilized through a terminals command linethe barebones essence of human-to-computer interactionCowork adds the friendly, compliant chatbot layer to the Claude desktop app. Currently, anyone with a $100-per-month Claude Max subscription can prompt the AI to complete nearly any computer-related task. Compared to popular chatbots like ChatGPT, Perplexity, or the original Claude.ai, Cowork can, in theory, leverage users hard drivestheir digital livesin its working memory. In practice, as a workplace tool, Cowork can organize haystacks of files into neat, delineated folders; turn screenshots of invoices into actionable spreadsheets; pull material from multiple websites, synthesized and analyzed in a single document; and even action slide comments. In short, its a general AI assistant, built for knowledge work. Its like having a junior researcher thats proactive, makes few mistakes, and solves problems before you realize there even was one, says Tomas Chamorro-Premuzic, professor of business psychology at University College London and Fast Company contributor whos written books about AI. Compared to other AI platforms, Cowork is smarter and more autonomous, he adds. In contrast to how people typically use ChatGPToften for their personal lives, such as travel plans, self-therapy, or recipesAnthropics latest agent can be more entrusted with their work. I use Cowork to digest research: Upload this folder and create a review of these articles main points, then look online for everything Ive written on this topic. So you can give it multi-layered tasks and receive highly accurate outputs, says Chamorro-Premuzic. These sorts of tasksbite-sized chunks of reading, researching, and writingare traditionally delegated to entry-level employees, says Conner. Its how knowledge workers judgment and expertise is honed on their gradual climb up the career ladder. If AI takes over some of those junior-level processes, it could pull the ladder up from them, he adds. The risk is that the effectiveness of AI agents like Cowork, and their widespread adoption, could impact entry-level hiring, continues Conner. This has long-term consequences for the economy. And in some ways, its already impacted how many entry-level jobs are available right now; some data suggests a 35% plummet in entry-level openings in the U.S. since 2023. Fewer junior roles eventually leads to fewer qualified candidates for mid-level and senior roles, Conner says. So, while short-term financial focus means some companies may want to adopt AI more and hire fewer junior folks, its possible we look back years from now and realize it hurt the talent pipeline. Coding the future Cowork has already deeply impacted the economy, just over a month after launch. p>On February 3, $300 billion was wiped from software and data stocks. It swiftly followed Anthropics release of plug-ins that can tailor Cowork to specific roles across sales, legal, financial analysis, and other industries. The worry is that if in-house lawyers can just use Cowork to do their work, they no longer need legal software, says Chamorro-Premuzic. He believes Cowork is an incremental upgrade, rather than exponential. But the next iteration will likely come soon. On February 5, Anthropic released Claude Opus 4.6. It has four times the effective recall of Opus 4.5, its predecessor that made Claude Code go virallaunched just 73 days before. The pace of release is accelerating (Sonnet 4.6, Anthropics free model, was released on February 17). Thats not just because of the AI raceClaude Code already writes 90% of its teams code. On December 27, its lead developer said all 40,000 lines of new code hed contributed in the past 30 days were written by the agentic tool. An autonomous, self-perpetuating feedback loop means the next upgrade is always imminent. We’re going to see AI involved in improving and building itself more and more, says an Anthropic spokesperson. In the closer term, Claude Code helps people build faster, with fewer resources, and the result is innovation across entire industries. We’re in the early innings of what’s possible. Claude Cowork isnt perfect. It requires oversight, particularly with high-stakes tasks. Like all AI, it can also make mistakesand given it can have access to personal files, apps, and tools, those errors can be more widespread than a chatbots text output. Its also not compliant first: conversation histories are stored locally, rather than within tightly regulated workflows. However, mass adoption of advanced AI tools like Cowork will likely be inevitable. So will labor market displacementits only the extent thats unclear. There are already signs that junior workers and college graduates are disproportionately affected by the onward march of AI. A November 2025 Stanford University study found a 16% relative decline in employment for early-career workers in occupations most exposed to the technology since 2022, such as software developers and customer service representatives. Not all entry-level jobs will be decimated by AI. Companies will always want to hire the best talent, and train them over time to accumulate institutional knowledge, workplace culture, and project history in ways technology cant. But there are concerns over how roles may be reshaped. If knowledge work merely becomes entering and fine-tuning prompts, it risks workers automating themselves, hampering their development and soft skills. Generative AI can tell you what to say in giving critical feedback, but its not the same as having a hard conversation and learning from it, says Chamorro-Premuzic. It becomes judgment without experience. The power of Claude Cowork means were one step closer to that future. Its like seeing the Wright brothers fly for the first time, says Conner. You wouldnt have understood the concept of a Boeing 767 flying across the ocean, but youd have grasped the idea of aviation. With Claude Cowork, you can see the future more clearly, and how this technology will have a major impact.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||