The wire, auto-curated daily.

Agentic AI takes enterprise with Google's ARDS standard and Alteryx's Agent Studio

Major tech players are launching open standards and tools that let business users build autonomous agents without heavy IT involvement.

Google launched the Agentic Resource Discovery Specification (ARDS), an open standard backed by Microsoft, NVIDIA, and Salesforce, enabling AI agents to autonomously discover and interact with web resources. Alteryx introduced Agent Studio and an MCP Server, allowing business analysts to convert existing data workflows into autonomous agents without relying on centralized IT teams. Warner Bros. Discovery deployed agentic AI on AWS with autonomous agents that self-optimize ad campaigns across linear and digital channels. These moves signal a shift from chatbots to systems that can act independently across enterprise workflows.

AI-designed universal vaccine enters human trials as photonic chips cut diagnostics time to 0.8 milliseconds

Breakthroughs in healthcare AI span drug design and ultrafast imaging, with new models promising faster, more efficient diagnostics.

Researchers at the University of Cambridge announced the first human trial of an AI-designed universal vaccine, marking the first time a vaccine's key component was designed entirely by AI. In diagnostics, Shenzhen University developed an all-fiber photonic AI platform using black phosphorus-based tunable modulators that processes liver CT scans in 0.8 milliseconds, compared to 85 milliseconds for electronic processors, with 246 times greater efficiency than GPUs. These advances highlight how AI is accelerating both the discovery of new therapeutics and the speed of clinical imaging workflows.

US and Chinese labs close the gap as open models match proprietary performance and costs fall 10–100x yearly

The AI landscape is tightening, with open-weight models catching up to proprietary systems while inference prices drop sharply.

US labs like OpenAI, Anthropic, and Google still lead many benchmarks, but Chinese labs such as DeepSeek, Alibaba, and ByteDance are closing in fast, especially on reasoning and coding tasks. The gap between open and closed models is shrinking, with Llama, Mistral, and Qwen now matching or beating GPT-4 on several benchmarks. Inference costs continue to fall, with 10–100x reductions each year, and smaller models are achieving what 70B+ parameters did last year. This trend points to increasingly accessible, efficient, and globally competitive AI capabilities.

AI consumes over 10% of US electricity as space-based data centers and new chips aim to ease the load

Surging AI compute demand is driving energy use higher, spurring innovations in infrastructure and efficiency.

AI is now consuming over 10% of US electricity, a figure that underscores the growing energy footprint of large-scale models and training runs. In response, data centers in space are gaining momentum to meet computing demand, leveraging solar energy and avoiding Earth's environmental constraints. Meanwhile, new chip designs, such as one from UC San Diego using piezoelectric components, aim to slash data center energy waste by rethinking how power is converted for GPUs. These developments reflect the industry's push to sustain AI growth while managing its environmental impact.

ServiceNow unveils Project Arc, an autonomous desktop agent governed by AI Control Tower and secured by NVIDIA OpenShell

ServiceNow and NVIDIA have expanded their partnership to introduce Project Arc, a new enterprise autonomous desktop agent that resides on employee desktops and completes complex, multi-step work without pre-built workflows. Secured by the NVIDIA OpenShell runtime and governed by ServiceNow’s AI Control Tower, Project Arc leverages the Action Fabric and is grounded in the ServiceNow CMDB to execute tasks across enterprise tools autonomously. The agent is currently in Early Preview, marking a significant step in extending agentic AI governance from desktops to data centers.

AI Control Tower expands into a full enterprise command center with five dimensions: Discover, Observe, Govern, Secure, and Measure

At Knowledge 2026, ServiceNow announced a major expansion of its AI Control Tower, transforming it from a governance layer into a comprehensive enterprise AI command center. The five dimensions now include 30 new integrations across major clouds and ERPs, continuous runtime monitoring via the Traceloop acquisition, AI-driven risk assessment aligned with NIST and EU AI Act standards, identity access governance via Veza, and ROI dashboards to track model spending and value. This release, part of the Australia update available in August, aims to bring order to "agentic chaos" by governing every AI agent, model, and asset across the enterprise.

Autonomous Workforce grows to cover every major enterprise function with role-specific AI specialists

ServiceNow has broadened its Autonomous Workforce from IT into all major business functions, introducing role-specific AI specialists for IT, CRM, Employee Service, Security & Risk, and more. These specialists share a common foundation of data, workflows, context, and governance, allowing them to execute complete workflows from intent to outcome within a customer's guardrails. While L1 IT Service Desk and CRM specialists are generally available, IT Ops, SRE, and Security & Risk specialists entered preview or are scheduled for general availability in June or September 2026, with further functions like HR, Finance, and Legal on the roadmap.