Perplexity Teammates: always-on AI agents for work

1. What was announced

On November 4, 2026, Perplexity launched "Teammates", AI agents meant to own ongoing work. According to its official blog, a teammate plans and coordinates projects, switches approach when priorities change, and presents results for review. It works inside the company's agent platform, Perplexity Computer, using the user's tools and files, and remembers what it learns across areas such as marketing, sales, support, operations and finance (Perplexity blog).

Its defining trait is that it acts without waiting for a prompt. In the official example, a sales teammate notices that a regular customer has stopped ordering, reviews the order history and support issues, then either drafts an offer for that customer or flags problems to fix before reaching out. Add a teammate to a Slack channel and colleagues can hand it work, and teammates can pass work to one another. Permissions follow the existing Computer settings, each teammate runs in a separate environment, and saved credentials are encrypted (Perplexity blog).

On availability, GIGAZINE reports that it comes first as a web version for subscribers of the paid Perplexity Max plan, with an iOS version coming soon (GIGAZINE). The Verge notes that, like OpenAI's Dots and Claude Duet, it runs around the clock, but says it is much more expensive because it requires the $200-a-month Max plan (The Verge). CEO Aravind Srinivas wrote on X that he envisions every organization having hundreds of teammates working for each person over time (Srinivas on X).

2. Why this matters to small businesses

What changes with this launch is that AI use is moving from "ask and get an answer" to "hand it over and have it keep watching". Similar products have appeared from several companies in a short time, and The Verge at least presents this as a move following OpenAI and Anthropic. In sales and operations at larger companies, having AI run daily jobs such as spotting signs of a lost account, summarizing inquiries and drafting quotes or proposals is becoming real.

Our view We think the $200-a-month price is not something small businesses will hand to every employee right now. Still, within a few months, a company whose sales and inquiries are run by always-on agents will follow up with existing customers faster than one that acts only when a person notices. If a lost account goes unnoticed, a competitor gets its proposal in first. This month, check whether your existing customers' order and inquiry history sits in one place. If it stays scattered, no agent can make use of it.

3. Points to check before using it

The official blog gives examples of teammates handling work that touches customers and internal tools, but it also says they send approval requests as they go, that completed work comes back for review, and that users can take over the desktop or redirect them (Perplexity blog). In other words, deciding how much to delegate and where a person approves remains the user's job. The scope of customer data you hand over also has to be set as an internal rule before any teammate is introduced.

Check whether customer order history, inquiries and quotes are scattered across email and paper
Write down which jobs AI may do alone (drafts, summaries, status checks) and which a person must always approve (sending, discounts, refunds)
Decide what customer data may be given to AI and put personal-data handling into your internal rules
Start with one job, such as checking on lapsed regular customers, and set a period in which a person reviews the results
Decide who administers the channels the AI connects to, such as Slack and email

Our outlook

Over the next six months to two years, we expect scenes in which each company's always-on agent watches a customer's ordering pattern, sends proposals on its own, and negotiates terms with the agent on the other side. In the same period, consumers' personal agents will also search for and compare shops and suppliers. A company whose products, prices, stock and lead times are not published in machine-readable form will be dropped at the comparison stage. Companies that first sort out their customer data and build a workflow around human approval, so that an agent can safely handle follow-ups, quotes and routine replies, will keep being chosen in AI-to-AI dealings too, while those still relying on someone to notice by chance will be slower to respond and easier for buyers to skip. Now is the time to lay that groundwork.

Questions this article answers

What are Perplexity's Teammates?

It is a group of AI agents Perplexity launched on November 4, 2026 to own ongoing work. They plan and move projects forward, investigate and act when circumstances change without waiting for a prompt, and present results for a person to review. They are first offered to subscribers of the paid Max plan.

What should a small business prepare before using always-on AI like Teammates?

Gather customer order and inquiry history in one place, and write down which jobs AI may do alone and which a person must always approve. Set the scope of customer data given to AI in your internal rules, and start with a single job, with a period in which a person reviews the results.

Sources

Related Q&A

Related services

If you want to try this in your own business

To see how the topics in this article would apply to your own operations, and what it would cost, send us a message through the contact form.