Eighty-five percent of enterprises run AI agent pilots, but just 5% ship them in production, according to Cisco’s Jeetu Patel. The blocker isn’t “rogue agents” but the lack of a trust architecture that covers delegation, identity, and telemetry for action risk. Patel also outlined Cisco’s rapid security tooling and a push for AI-built products.
OpenAI has introduced Workspace Agents, a business-focused successor to custom GPTs that teams can build or select from templates and deploy across tools like Slack, Salesforce, Notion, Google Drive, and Microsoft apps. Instead of pausing when you stop chatting, these agents run on a cloud coding backbone, can schedule long workflows, and operate under granular admin permissions.
Your news, in seconds
Get the Beige app — every story in 60 words, updated hourly. Free on iOS & Android.
Enterprises are moving AI agents from experiments to production, and the key question is how to manage them. Google and AWS are taking opposite paths: Google emphasizes a Kubernetes-style control plane for governance and identity, while AWS pushes config-based harnesses in the execution layer for faster deployment. Both aim to reduce new risks like state drift in long-running agents.
Tata Steel and Google Cloud say they’ve scaled agentic AI across the company in just nine months, deploying 300+ specialised AI agents. The rollout spans employee-facing tools, back-office automation, and manufacturing safety—using platforms like Zen AI, a Digital Assistant, and shop-floor “Safety EyeQ” for real-time alerts—cutting routine ticket resolution and customer response times.
Swiggy is opening its AI commerce infrastructure through Builders Club, a developer programme for external builders, startups, and enterprises. Backed by AWS and powered by Amazon Bedrock and AgentCore, it grants approved teams access to multiple MCP servers and 18+ APIs across Swiggy Food, Instamart, and Dineout. Builders can create AI agents and copilots that take real actions, with invite-led access, rate limits, and engineering support.
Enterprise AI platform Vibrium has raised a $1 million seed round to strengthen its product development and AI agent framework. The company says its goal is to close the gap between what AI can do in tests and what enterprises can actually deploy, helping move customers from experimentation toward real business execution.
Never miss a story
Set alerts for the topics and sources you care about. Download Beige for free.
After 18 months of “agent building,” BAND (Thenvoi AI Ltd.) targets the tougher next step: letting AI agents built on different frameworks actually collaborate. Exiting stealth with $17 million, it offers a deterministic “Slack for agents” interaction layer with multi-peer comms, identity-based permissions, and an enterprise control plane for auditability—positioned as infrastructure for a universal orchestrator.
Backed by SoftBank, Emergent has released Wingman, an AI agent built to automate routine work across popular tools. The twist: unlike many rivals that act on their own, Wingman seeks user confirmation for significant actions and gradually learns personal preferences over time. The launch targets demand for agents that manage workflows independently, not just respond to prompts.
Swipe through stories, personalise your feed, and save articles for later — all on the app.