欢迎进入AI的产品时代

Predicting the trajectory of AI itself is pointless.Nobody outside the walls of the biggest and best-fundedarti昀椀cial intelligence labs really knowswhat’scoming down the pipe in terms ofAsevidence:we’ve continued to get periodic“future shock” moments despite analysts’ bestguesses. And we’ve been seeing them faster, withfewer gaps in between, when, by rights, we shouldbe getting better at forecasting them over time.Before late 2022, who saw the original launch ofChatGPT coming? In early 2023, who expected itto hit100 million users in two monthsof generalavailability? This spring, who thought the worldwould go temporarily mad for Studio Ghibli-styleimages, as the biggest manifestation of GPT 4o’snative image generation model? Ahead of lastmonth, May 2025, who expected Google’s Veo 3video generation model, which combined audioand video into single prompts, to potentiallyblowthe doors off advertising?You only have to glance overthe OpenAI blog postthat accompanied the initial ChatGPT researchpreviewto see just how quickly technologypredictions can end up looking rather humble. Andyou only need to look back at the articles written atthe peak of the Metaverse furore for a reminder ofthe opposite: how quickly the air can get suckedout of a room it sure looked like everyone wasinevitably rushing into.So am I going to start this year’s AI Report with apredictionofwhattechnology segment will look like, capability-wise,this time in 2026? No.Thereis,though,a line of more productiveprediction I’m much happier putting my name nextto. I’m con昀椀dent that the commercial and culturaluptake of AI is going to generate a whole new classof products, services, devices, and experiences,irrespectiveof how many more future shocks2025/26 brings us.I’malso comfortable saying that those newintroductions will change how we think about notjust fashion software and hardware, but the natureof work, the scope of automation, the value ofcreativity and potentially much more. Again: theseshifts are not going to come about because ofvague possibilities or vibes, but thanks to concreteapplications.Thereason I’m happy standing by this bigassessment is that it’s not an especially unusual one,and I’m by no means alone in making it. In fact, theevidence for it is simply everywhere in the spring of2025 in a way it simply wasn’t this time last year.BY BEN HANSONEDITOR�IN�CHIEF, THE INTERLINEWELCOME TO AI’SPRODUCT ERAWHAT IT MEANS FOR FASHION TO LOOK FORWARDS AT TWO VERY DIFFERENTHORIZONS - ONE WE CAN SEE FAIRLY CLEARLY, AND ONE WE DEFINITELY CAN’T. capabilities.thefastest-developing*Image created using AI. 3In my introduction toThe AI Report 2024, I said that the marketexpectation for brands, suppliers, technology vendors, and serviceproviders was to “insert AI into whatever they do”.Now that work is extremely well underway.According to the 2025 edition ofStanford University’sindustry-de昀椀ning AI Index(all 400+ pages of which arewell worth reading) the last year has seen a more than20% increase in the number of organisations thatself-report using AI in at least one area of theirbusiness - to where that 昀椀gure is now knocking onthe door of 80%.The same report also gets more speci昀椀c: theshareof companies,across industries,usinggenerative AI (not just AI in general) in one ormore business functions doubled in the sameperiod, to where it now stands at more than 70%.So it’s no longer accurate to say that ‘traditionalAI,’i.e.the kind of analytics,segmentation,classi昀椀cationand so on that would havepreviously been labelled as machine learning, isdoing most of the heavy lifting.In terms of pure scale and ubiquity, we’re now along way from the pilot phase of AI - and theattitudes of business leaders have evolved tomatch.Ina now-infamous memo, which circulated inAprilof this year,Shopify CEO Tobi Lutkeinstructed the eCommerce juggernaut’s internalteams to “demonstrate why they cannot getwhat they want done using AI” before they werepermitted to request more staff or additionalresources.In the same memo, Lutke called AI usage a“baseline expectation” for not just the incomingwaveof professionals,but throughout thecompany’s existing workforce.And even though it feels like a high bar to set,less than three years after generative AI landedas a class of professional and consumer tools,that expectation isn’t wildly out of step withwhat’s happening in the world. This May, SundarPichai, CEO of Alphabet / Google announcedthat, across enterprise and personal applications,the number of tokens - raw output - generated bythe company’s Gemini model family hadincreased50x year-over-year from 2024 to 2025.Everyone is, in other words, already using AI. The world ofwork is simply mirroring the world at large. Makeno bones about it:there has been notechnology adoption curve quite like this outside ofthe internet. And even though one company -OpenAI - is currently swallowing up most of thegeneral-purpose AI business, witha reported 80%market share amongst corporates,we are not