OpenAI is prioritizing Agentic commerce, that is, the use of AI tools to do shopping research and make purchases on a userâs behalf. This year, OpenAIâs ChatGPT has launched integrations with apps such as Booking.com, Canva, Coursera, Expedia, Figma, Spotify, and Zillow. It has also announced further partnerships with Target, Intuit, and others. With Instacartâs built-in search app, OpenAI is launching a grocery shopping experience inside of ChatGPT, allowing customers to brainstorm meal ideas, make a grocery list, and check out, all without leaving the chat interface. These agentic commerce tools can give OpenAI another way of making money, since itâll take an undisclosed âsmall feeâ when it helps merchants make a sale. Adobe has predicted that AI-assisted online shopping will grow by 520% this holiday season.
Harvey.ai, an AI platform for law firms, closed a $150 million funding round last month, valuing the company at $8 billion, double its valuation from the previous year. Andreessen Horowitz led the round. Harvey is backed by the investment arm of RELX, a $85 billion publishing group with a market cap, which owns the legal database LexisNexis. Harvey signed a deal with LexisNexis in June. The San Francisco-based three-year-old startup has raised more than $1 billion, including the new round, outstripping its rivals in valuation and fundraising. Annually, it generates more than $100 million in recurring revenue, or ARR. Harvey was founded in 2022 by Weinberg, then a junior lawyer with O'Melveny & Myers, and his friend and former DeepMind researcher Gabe Pereyra. The legal AI tools market is crowded: ⢠Harvey is head-to-head with the Swedish startup Legora, which has a $1.8 billion valuation. ⢠Other tech startups are Luminace, Clio, and Ironclad. ⢠Chasing legal niches is EvenUpâs tool for personal injury lawyers, and Finch for paralegals. ⢠Thomson Reuters, which owns case law database Westlaw, last year snapped up Casetext, another Harvey rival, in a $650 million deal.
AI can replace 11.7% of the U.S. labor force, amounting to $1.2 trillion in wages across finance, health care, and professional services. This is the main finding of an MIT study conducted using its simulation tool, the Iceberg Index, that simulates how 151 million U.S. workers interact across the country and how they are affected by AI. Researchers found that layoffs and role shifts in tech, computing, and information technology account for just 2.2% of total wage exposure, or about $211 billion. The Iceberg Index also challenges a common assumption about AI risk â that it will stay confined to tech roles in coastal hubs. It runs population-level experiments, revealing how AI reshapes tasks, skills, and labor flows long before those changes show up in the real economy. "Basically, we are creating a digital twin for the U.S. labor market," said Prasanna Balaprakash, ORNL director and co-leader of the research.
Anthropic announced Claude Opus 4.5, the latest version of its flagship model, following the launch of Sonnet 4.5 in September and Haiku 4.5 in October. It includes improvements to memory for long-context operations. "Claude Opus 4.5 is state-of-the-art on tests of real-world software engineering," and "is the best model in the world for coding, agents, and computer use," said the company. Opus 4.5 is already available in the apps, API, and on all three major cloud platforms. For developers, pricing is $5/$25 per million tokens. This model will face competition from other recently released frontier models, most notably OpenAIâs GPT 5.1 (released on November 12) and Googleâs Gemini 3 (released on November 18). Alongside Opus, Anthropic released updates to the Claude Developer Platform, Claude Code, desktop app, and consumer apps.
OpenAI introduced shopping research to all users this month. Released for the holiday season, the feature has been post-trained on GPT-5-Thinking mini with reinforcement learning for shopping tasks. Essentially, ChatGPT conducts deep research for users across the Internet, including reviews, comments, and high-quality sources, helping them find the right products without sifting through multiple sites. The model might ask clarifying questions. The users just describe what they are looking for, i.e., "Find an affordable electric bike that can be folded and stored in a small apartment." The chatbot considers users' past conversations and their memory graph to deliver a personalized buyerâs guide in minutes. The company ensured that this tool "performs especially well in detail-heavy categories like electronics, beauty, home and garden, kitchen and appliances, and sports and outdoor." For simple shopping questions, such as checking a price or confirming a feature, a regular ChatGPT response is quick. When the user wants in-depth shopping research, such as comparisons, constraints, and trade-offs, the model takes a few minutes to provide a more detailed, well-researched answer. The company trained the model to read trusted sites, cite reliable sources, and synthesize information to produce high-quality product research. It is also designed to be an interactive experience that can update and refine its research in real time, adjusting to user product preferences. To purchase an item, they can click through to the retailer's site. In the future, OpenAI plans to offer merchants in Instant Checkout the option to purchase directly through ChatGPT for merchants who are part of Instant Checkoutâ .
