AI in Internet Software & Services

Telepass (Italy — 75% of motorway tolls, 10M+ customers, 18 European countries) deployed Salesforce Agentforce in 6 weeks to autonomously handle 40,000 weekly customer conversations. The FAQ agent now resolves 87% of incoming inquiries on its own and is backed by Data 360 for knowledge grounding plus MuleSoft for SAP integration. A second internal Agentforce deployment has already cut service-rep average call handle time by 50%.

Paris-based multiplayer AI platform Dust deployed Claude via Anthropic’s API to power its no-code AI agent infrastructure, saving roughly $10,000 per day on model spend (18-19% reduction) after optimizing prompt caching. The deployment raised autonomous tool-calling depth from 4-5 to 8+ steps per agent run with no engineering changes and now serves more than 300,000 agents deployed across customers including Datadog, Vanta, and 1Password.

Austin-based SaaS company SupportFlow deployed the ECOA AI customer-support platform and cut first response time 73% (14.2 hours → 3.8 hours), raised CSAT from 4.1 to 4.6, and nearly halved annualized agent turnover (32% → 14%) within 90 days by automating 58% of incoming tickets end-to-end.

Manipal Hospitals, one of India's largest healthcare providers serving nearly 7 million patients annually across 37 hospitals and 10,500 beds, deployed Google Cloud generative AI for nurse handoffs, cutting handover time by 90 minutes per shift.

Baichuan AI released Baichuan-M3, a new-generation medical-enhanced LLM that beats OpenAI's GPT-5.2 on HealthBench (+28 over M2 on HealthBench-Hard at 44.4), SCAN-bench (74.9 on Clinical Inquiry, +12.4 over GPT-5.2-High), and posts a lower hallucination rate. The model is trained on the SCAN framework to explicitly model the clinical decision-making process.

Microsoft and NVIDIA are partnering to bring AI across the nuclear energy lifecycle — from permitting and design to construction and operations. Aalo Atomics cut permitting time 92% (saving ~$80M/year); Southern Nuclear, Idaho National Lab, Everstar, and Atomic Canyon are also deploying Microsoft Copilot and Azure-based AI tools.

Google DeepMind and Commonwealth Fusion Systems (CFS) are partnering to apply AI to SPARC, a compact tokamak aiming to be the first magnetic fusion machine to generate net energy. The collaboration uses TORAX (a JAX-based plasma simulator) plus reinforcement learning for plasma simulation, optimal-pulse discovery, and real-time control.

Google DeepMind's Graph Networks for Materials Exploration (GNoME) used deep learning and active learning to discover 2.2 million new crystal structures stable enough for synthesis, with 380,000 on the convex hull and 736 already independently realized in labs worldwide.