Meta’s Muse AI agent starts attacking subscription bloat

What changed: Meta rolled out its Muse AI personal agent as a broad consumer assistant that can cancel subscriptions, request refunds, and manage everyday admin across services. Meta’s launch has become a pillar of its AI strategy and is associated with about a 30% share price jump in the past month despite no disclosed usage or revenue figures.

Why it matters: Consumer agents that can act on billing and account data move AI from chatty helper to financial control center, creating room for startups to build vertical agents around bills, health, or housing. Founders should expect users to benchmark new tools against Muse’s ability to take actions, not just generate text.

Try/watch: If you offer subscriptions or consumer services, audit how easy it is for an agent like Muse to cancel, downgrade, or renegotiate—then decide whether to lean into agent-friendly workflows or build new retention playbooks.

KT’s Agentic On lets enterprises wire AI agents into legacy systems

What changed: KT unveiled Agentic On, a platform that lets companies plug AI agents into existing business systems and internal documents without rebuilding their stack. Businesses can assemble functions without complex programming and orchestrate multiple agents to review documents, request approvals, and write back to core systems, all under granular permission controls for data, models, and agents.

Why it matters: This is a template for how incumbents will add agents on top of legacy IT rather than ripping and replacing, which favors startups that can slot into existing ERP and CRM environments. Operators should see Agentic On-style platforms as proof that agentic automation is moving from experiments to standard tooling in regulated corporate settings.

Try/watch: Map one or two high-friction workflows—like contract intake or compliance reviews—and prototype agents that read your docs, route approvals, and update systems while respecting your current access-control scheme.

AWS and Salesforce embed production AI agents into Slack, CRM, and voice

What changed: AWS and Salesforce announced a suite of AI agent integrations that put AWS DevOps Agent directly into Slack workspaces, with Security, FinOps, and Partner Central agents scheduled for later in the fall. The partnership also connects Amazon Quick to Salesforce CRM data and links Salesforce Agentforce Voice with Amazon Connect, allowing agents to act on business context across chat and voice without manual context switching.

Why it matters: Embedding agents inside tools teams already use—Slack, CRM, call centers—shortens adoption time and makes it easier to automate incident response, cost optimization, and partner workflows with minimal training. Buyers should expect their cloud and CRM providers to ship opinionated agents for common tasks, and plan how to blend these with any in-house agent frameworks.

Try/watch: Start small by piloting AWS DevOps Agent in a limited Slack channel or project, tracking impact on deployment speed and incident handling before expanding to security or finance agents.