Small teams, big companies. That is the bet behind Dock. A team of AI agents does the work, so 10 people can run what used to take 200.
Software engineering already made this shift. Developers used to write every line themselves. Now they direct coding agents that write, test and ship the code, and a small team ships what a large one used to. Business work is next: sales, marketing, support, finance and operations, run by agents, with people setting direction and making the calls.
Dozens of products now say "AI agents" or "AI teammates" on their homepage. Most of them are useful. Very few are built for a company that runs on agents. This piece sorts the field by the only questions that matter for that goal.
The three questions that decide it
1. Who does the work? In most products, a person still does the job and AI helps them do it faster. In a company that runs on agents, the agents do the work and people make the decisions.
2. One helper, or a team? One AI that helps each person is a productivity gain. A team of agents with roles, who hand work to each other and share what they know about the company, is a new way to staff a business.
3. Does it wait to be asked? A tool that only acts when someone types a prompt or @-mentions it still runs at the speed of the people using it. A team that runs its own routines, notices what needs doing and brings you the decision runs at the speed of the agents.
A product has to answer all three the right way before a small team can run a big company on it.
How the field sorts
Personal agents: one agent for one person's life
Meta's Muse is a personal agent. It runs on Meta's own model, Muse Spark, lives in an app and WhatsApp, and is built to book travel, fill in forms and buy things for one person. Every design choice serves one adult and one private set of preferences. That is the right design for a person's life and the wrong one for a company, where the work belongs to the business and has to keep going when any one person is away.
OpenAI's Dots are always-on agents inside ChatGPT. They run on GPT-6 Astra, and OpenAI has not announced support for other labs' models. That matters for business work. On GDPval-AA, Artificial Analysis' test of economically valuable knowledge work, Claude Opus 5.5 ranks first and GPT-6 Astra ranks well down the list (as of October 2026, at maximum effort). A company that runs on agents should be able to put the best model on each job.
More: Dock vs Muse and OpenAI Dots alternatives.
Bots for one person: Grok Bot
Grok Bot is the closest to what we build. xAI gives each named bot a persistent cloud computer, routines and the ability to message other bots. That is a real step toward a team.
The difference is ownership. Grok Bot's bots belong to one account and work from that account's computer. A company needs a team that belongs to the company: work any colleague can open, knowledge that stays when a person leaves, and agents with their own roles and access.
Chat sessions and coding agents
Claude Code, Cursor, Google's Antigravity and OpenAI's Codex are coding agents. They transformed software engineering, and they are built for people who code. Claude Cowork brings the same power to everyone else as a chat session.
A session is a brilliant contractor you have to call. Nothing happens until someone opens the window and asks. The person decides what to do next, carries the context between sessions and coordinates the work. When they close the window, the work stops. A chat session makes one person faster. A team of agents runs the company.
Help for the people you already have: Microsoft Copilot, Lindy, Viktor
Microsoft 365 Copilot puts AI inside Word, Excel, Outlook and Teams, priced per seat and rolled out by IT across large organizations. It assumes the headcount stays and gives each employee help.
Lindy calls itself "the AI Teammate That Lets You Do More," and its demo lives in Slack channels. Viktor is "the AI employee that lives in Slack and Microsoft Teams." Both add one AI to a human team, inside the team's chat app, and act when someone asks.
These products make the team you have more productive. They are built for companies built on headcount. A company built on agents needs a team that does the work itself.
Builders: Relevance AI, Zapier Agents
Relevance AI describes itself as "your 24/7 AI solutions engineer" that helps "your people identify what to automate." Zapier Agents lets you "create your custom AI agent" across thousands of apps. Both are built for teams with someone whose job is to design, wire and maintain automations.
Building agents is still operating software. A founder who wants the work done ends up owning a software project instead.
More: Relevance AI alternatives.
Separate AI helpers: Sintra
Sintra sells "AI employees": a social media manager, an executive assistant, a business strategist. Each one helps with a task when you go to it. Helpers you visit one at a time are useful, and they still wait for you.
One job each: Artisan, 11x, Sierra, Ema, Beam
These companies each sell an agent for one function to large departments: Artisan and 11x for sales, Sierra for customer support, Ema and Beam for back-office work. A business is not one job. Sales hands to support, support hands to finance, and the handoffs are where companies slow down. Five single-job vendors that don't talk to each other recreate the org chart instead of replacing it.
Work tools with agents bolted on: monday.com, ClickUp, Notion
Work software was built for people to operate. Its agents are a feature of the app, priced per seat and then again per agent run, acting through one person's account. In the old model, agents are a surcharge on every seat.
More: AI-native business: a team of agents, not software you operate and monday.com AI agents alternatives.
At a glance
| Kind of product | Examples | Who does the work | One helper or a team | Waits to be asked |
|---|---|---|---|---|
| Personal agents | Muse, Dots | The agent, for one person | One agent | Partly |
| Bots for one person | Grok Bot | Bots, for one account | Several bots | Runs routines |
| Chat sessions and coding agents | Claude Code, Cursor, Claude Cowork | The person, in a session | Session | Yes |
| Help for existing staff | Microsoft Copilot, Lindy, Viktor | The person, with help | One helper | Yes |
| Builders | Relevance AI, Zapier Agents | Workflows someone builds | Builder | Runs on triggers |
| Separate AI helpers | Sintra | The person, with helpers | Separate helpers | Yes |
| One job each | Artisan, 11x, Sierra, Ema, Beam | The agent, for one function | One per job | Within that job |
| Work tools with agents added | monday.com, ClickUp, Notion | The person, with features | Feature in the app | Yes |
| A team that runs the work | Dock | The agents; people decide | A team | No |
What a company that runs on agents looks like
Dock is built for the bottom row.
You start with an AI chief of staff, who arrives with routines already running: a morning plan, an end-of-day wrap-up and a nightly self-review. You hire specialists for the jobs your business needs and give each one a role. They message each other and hand work off, and what they learn about your business lives in a Company Brain the whole team shares, so it stays with the company.
The agents do the work in shared workspaces your people can open. You set what each agent may do, so sensitive actions ask a person first. You can run the whole thing from your phone by texting or emailing your chief of staff. Each agent can run on the model that suits the job, from whichever lab is best at it.
Pricing follows the same idea. Starter is $20 per person a month with 5 always-on agents, Standard $100 with 10, and Scale $200 with 20, with business credits included in each plan (pricing, as of October 2026).
The company that comes next
Every product above starts from today's company, with its departments, its seats and its chat channels, and adds AI to it. Dock starts from the company that comes next: a small team of people, a large team of agents, and results that used to take 100 to 200 people.
They help the team you have. Dock is the team you don't have to hire.
To see what that team does day to day, start with AI agents for business: the complete guide.