Asana buys Stack AI — adds a no-code agent builder into its workplace stack

What changed: Asana announced the acquisition of Stack AI, a no-code workflow‑automation company that builds agents that work inside business systems (Salesforce, Slack, G Suite), with Stack AI’s founders joining Asana and the product folded into Asana’s AI tooling roadmap.

Why it matters: If you run ops or product teams, this makes it easier to agentify end‑to‑end business processes without heavy engineering — Asana is positioning itself to deliver agents inside the same place teams already plan and run work, which reduces integration friction.

Try/watch: Pilot small, repeatable automations (approvals, status collection, CRM lookups) inside Asana’s AI Studio / AI Teammates to judge whether Stack AI’s no‑code approach reduces implementation time versus building custom automations; watch how Asana prices deeper agent integrations and data‑access controls.

Workday + Google Cloud broaden partnership — Workday agents now in Gemini Enterprise

What changed: Workday and Google Cloud expanded their partnership so Workday’s Sana self‑service agent is available in Gemini Enterprise, with Gemini now the default model for Sana and deeper data integrations to let agents act on HR and finance workflows while honoring Workday policies.

Why it matters: For HR/finance operators, this means employees can get answers and complete common tasks (time‑off, payslips, approvals) from a conversational agent inside the model interface while Workday enforces permissions — reducing context switching and manual ticketing.

Try/watch: Run a controlled pilot for a narrow use case (e.g., time‑off queries and manager approvals) to validate accuracy and audit logs, and monitor data residency and permission boundaries as Gemini becomes the default model behind those interactions.

CoreWeave launches unified agentic capabilities — close the training→inference feedback loop

What changed: CoreWeave announced a suite of agentic features — serverless RL for post‑training, production‑ready inference, multi‑agent observability, and automated improvement tooling — that it says closes the loop so agents can learn from real‑world runs and be retrained more quickly. The release highlights serverless RL, built‑in monitoring, a Weights & Biases integration for experiment tracking, and claimed cost/time improvements versus local GPU setups.

Why it matters: Builders and platform teams struggling with long agent iteration cycles should treat this as an infra option to shorten dev→production feedback, surface multi‑agent failure modes, and embed continuous improvement — which can materially reduce the time and risk of shipping agentic features.

Try/watch: Test a non‑critical agent workflow against CoreWeave’s stack to measure actual iteration speedups and end‑to‑end cost; validate the observability signals that detect multi‑turn failures before widening rollouts.

Hystax releases OptScale AI — FinOps and governance for LLMs and agents

What changed: Hystax launched OptScale AI, expanding its FinOps product to include an AI Gateway, security/guardrails, analytics and tracing, and agent controls (cost/recursion limits, anomaly detection), and positions the product to cut LLM spend and centralize agent governance.

Why it matters: Operators deploying many small agents now face rising model costs and blind spots; a platform that consolidates routing, cost optimization, and audit logging can make multi‑agent deployments operationally manageable and auditable.

Try/watch: Try the free tier on a dev environment to measure routing savings and audit completeness, and evaluate the anomaly detection for false positives/negatives before relying on it for blocking production agent behavior.