Perplexity launches always-on Teammates for ongoing work

What changed: Perplexity introduced Teammates, persistent AI agents designed for complex work rather than one-off questions. Users can assign separate roles, run agents simultaneously, monitor them in dedicated threads, and interact with them through Slack; the feature is limited to the $200-per-month Perplexity Max plan, with mobile availability planned later.

Why it matters: This moves agent software closer to a standing digital workforce: agents retain project context, keep working when the user is away, and can coordinate across tasks. Founders and operators should evaluate whether a narrowly defined, repeatable workflow can justify the high subscription cost before attempting broad delegation.

Try/watch: Test one low-risk workflow—such as research monitoring or issue triage—with explicit approval requirements, and measure completed work, errors, and credit consumption before giving an agent broader access.

Amazon lets selected customers shop with Alexa by text

What changed: Amazon began rolling out a text-message interface for its Alexa AI agent to selected U.S. customers. Eligible users can use natural-language messages to search for products, add items to their cart, and make purchases without opening Amazon’s app or website.

Why it matters: Shopping agents are becoming a new purchasing surface, which could reduce the importance of traditional website navigation, search results, and sponsored placements. Retailers and consumer brands should make product information, pricing, availability, and fulfillment policies easy for software agents to interpret accurately.

Try/watch: Audit the machine-readable quality of your catalog and test whether an independent assistant can correctly identify products, compare options, and complete a purchase using ordinary language.

Meta adds Spaces for collaborating with Muse agents

What changed: Meta launched Spaces, a new way to communicate and collaborate with its Muse AI agents. The feature is intended to extend Muse beyond a simple chat interaction by giving users customized, app-like environments for working with their AI helpers.

Why it matters: The product direction suggests that consumer agents will increasingly be organized around persistent workspaces rather than isolated conversations. Builders should think about the agent’s shared context, tools, and user interface as one product—not just the underlying model.

Try/watch: Create a small workspace around one recurring activity and define what information the agent may retain, what actions require confirmation, and how a user can review or undo its work.