Inside Invisible

🧭 How embedding engineers speeds up AI time-to-value

AI’s promise is huge, but with 95% of projects failing to deliver ROI, the gap between ambition and execution is widening. Our latest report unpacks the forward-deployed model: how embedding engineers directly into business operations turns AI from a pilot into performance. Inside, you’ll see how proximity accelerates delivery, why FDEs outperform traditional consulting, when not to use the model, and how enterprises like the Charlotte Hornets are cutting build times from months to days.

DOWN TO BUSINESS

🛒 Amazon’s AI shopping assistant drives $10 billion in new sales

Amazon says its AI shopping assistant, Rufus, embedded directly into Amazon’s mobile app and website, is on track to generate an additional $10 billion in annual sales. CEO Andy Jassy shared that 250 million customers have used Rufus this year, and shoppers who engage with it are 60% more likely to complete a purchase.

⚖️ Federal judges used AI to draft filings with fake quotes and names

Two U.S. federal judges admitted their offices used generative AI to draft court documents that included false quotes, fabricated cases, and fictional litigants. The filings, later retracted, prompted a Senate Judiciary Committee inquiry and renewed calls for formal rules on AI use in the judicial system. The Administrative Office of the U.S. Courts has since formed a task force to develop guidance as AI tools become more common in legal drafting.

☁️ Google and Anthropic sign multi-billion-dollar cloud deal

Google and Anthropic have signed a cloud partnership worth tens of billions of dollars to expand Anthropic’s computing capacity as demand for its AI tools surges. The deal will help Anthropic scale to serve hundreds of thousands of enterprise clients and sustain its $7 billion annual revenue run rate. Meanwhile, OpenAI also signed a seven-year, $38 billion agreement to buy cloud services from Amazon.

💳 AI agents are reshaping how payments work

A new report forecasts that “agentic commerce,” where AI agents shop and pay on behalf of consumers, could create a $5 trillion global opportunity by 2030. In this model, payments aren’t a simple handoff between buyer and seller but a continuous exchange between AI systems. Stripe and PayPal are already building agent-ready protocols, signaling a shift where payment infrastructure must evolve to manage agent identity, delegated authorization, and real-time fraud detection.

🤖 Allianz launches agentic AI to settle insurance claims in hours

Allianz has unveiled its first agentic AI system designed to automate food spoilage claims, cutting processing times from days to hours while keeping humans in control. Built in under 100 days, the system uses seven specialized AI agents to handle coverage checks, fraud detection, and payout recommendations, with a claims professional making the final call.

🤝 Anthropic signs Cognizant in one of its biggest enterprise AI deals

Anthropic has secured professional-services firm Cognizant as one of its largest enterprise customers. The partnership will see Cognizant roll out Claude to 350,000 employees and help the company “transition from a system integrator to AI builder”, according to Cognizant’s chief executive.

HOT MODELS

🧬 Breakthrough brings AI drug discovery closer to reality

Scientists have released a new AI model that can predict how proteins interact with drugs and other molecules, a crucial step in designing new treatments. The model builds on previous breakthroughs in protein folding but is now available for commercial use, allowing companies to accelerate drug discovery without giving up proprietary data.

🧠 AI platform unifies 40+ models in one workspace

ChatPlayground AI has launched a professional platform that consolidates more than 40 leading AI systems into a single workspace for testing and comparison. The tool lets users run the same prompt across models, analyze PDFs and images, and export structured results for audits or presentations.

🎥 Zoom and Nvidia partner on enterprise AI assistant

Zoom has partnered with Nvidia to launch AI Companion 3.0, its new workplace assistant designed to automate tasks across Microsoft 365, Google Workspace, Slack, and Salesforce. The system uses a hybrid approach that routes simple tasks to smaller models and complex ones to larger ones, improving speed and efficiency.

PLOT TWIST

🚨 AI-generated misinformation spreads through Google search results

False claims about Australian road rules created by an AI-generated website recently appeared in Google search results, warning drivers they could face a $250 fine for not keeping headlights on at all times. Transport for NSW clarified the information was false and urged the public to rely on official sources.

⚖️ UK court rules in favor of Stability AI in landmark copyright case

A London High Court has ruled that Stability AI did not infringe Getty Images’ copyright by training its image-generation model on Getty’s photos. The judge found that because Stability’s model does not store or reproduce original works, it is not an “infringing copy.” The ruling narrowed Getty’s claims after parts of the case were withdrawn and comes as the UK government consults on whether to introduce a text and data mining exception to copyright law.

🎬 The number one sign you’re watching an AI video

Digital forensics experts say the easiest way to spot an AI-generated video isn’t six fingers or melting faces—it’s bad quality. Blurry, low-quality clips often hide the subtle distortions that give synthetic content away, making them more convincing than polished fakes. As AI tools advance, those visual clues are vanishing fast, forcing a shift from “does it look real?” to “where did it come from?”

😄 Study finds people emotionally adapt to AI humor

New research using brainwave analysis found that people respond to jokes told by AI almost as positively as those told by humans. While participants initially showed surprise at AI humor, their cognitive and emotional responses grew stronger over time as they adjusted to the technology’s style. The findings suggest that as AI becomes more socially capable, user trust and emotional engagement with AI companions could deepen.

♻️ Subscribe on LinkedIn and share it with your network!