AI Employees

Build a team that works while you sleep.

AI Employees are named, specialised workers you configure once and then delegate to. Give one a role, a knowledge base, and 1,000+ business integrations to work with — then let it run on its own schedule, hand work to its colleagues, and prove it did the job with a score on every single run.

Not a chatbot. A role you fill.

A defined role

  • Name, job title, and a system prompt that sets how they think and write
  • Assign them to departments so the right work reaches the right specialist

Their own knowledge

  • Point them at your documents, folders, or tagged files
  • Answers come back grounded in your material, reranked for relevance rather than guessed

The tools to act

  • Web research, your data tables, outreach sequences, document creation, and browser automation
  • 1,000+ business integrations, enabled per employee

Most AI tools wait to be asked

A chat window is only useful when you remember to open it. The work that actually compounds — the follow-up, the monitoring, the weekly report — is the work nobody has time to trigger by hand.

You are the bottleneck
Every task starts with you opening a tab and writing a prompt. Nothing happens on the days you are busy, travelling, or asleep.

Context resets every time
Generic assistants forget your customers, your tone, and last week’s decision. You re-explain the same background before every useful answer.

No proof it worked
Automations either run silently or fail silently. You find out something broke when a customer tells you.

Start from a pre-built role, or write your own

The library ships with ready-made personas across Sales, Marketing, Customer Success, Engineering, Operations, Finance, Legal, and HR — each with a professional headshot and a complete system prompt. Pick one and adjust it, or start from a blank role and describe the job yourself.

  • Sales & Business Development — Research accounts, build and enrich prospect lists, qualify inbound, and keep outreach sequences moving.
  • Marketing & Content — Draft and schedule content, monitor keywords and competitors, and turn research into publishable material.
  • Customer Success & Support — Answer from your documentation, summarise account history, and flag the conversations that need a human.
  • Operations & Project Management — Track recurring processes, keep internal records current, and produce the status report nobody wants to write.
  • Finance, Legal & HR — Review documents against your policies, prepare recurring summaries, and answer routine internal questions.
  • Engineering & Technical — Research technical questions, work in a sandboxed project workspace, and publish working output to a live URL.

Work that runs without being asked

One-time and recurring

  • Schedule a task in plain language — “every Monday at 9am”
  • Runs daily, weekly, or monthly without further prompting

Continuous heartbeat

  • For work that cannot wait for a daily batch, run an employee on an interval from every five minutes to once a day
  • Bounded by active hours and days of the week

Aware of what it already did

  • Heartbeat runs keep a daily log and read yesterday’s before starting
  • An employee picks up where it left off instead of repeating itself

Every run is graded. You see the trend.

Autonomous work is only worth having if you can tell whether it actually worked. After each run finishes, a separate model reads the output against the original task and scores it from 0 to 100 — with a written reason.

A score on every run
Not “did it return?” but “did it do the job?” Each run is graded against the task you actually wrote, and the reason is recorded alongside it.

Trends, not snapshots
The Scheduled page shows average score per task and a sparkline of recent runs, so quality drift is visible before it becomes a problem.

One click to improve the prompt
Hit Evaluate and the platform runs the task for real, then proposes a better-written version of your prompt with the specific changes explained.

They work with each other, not just with you

One employee does not have to be a generalist. Specialists hand work sideways, and the work reaches you wherever you already are.

Employee-to-employee delegation
An employee can see its colleagues and pass a task to whichever one is right for it — research to the researcher, copy to the writer — without routing through you.

Reachable from your own tools
The platform runs as an MCP server, so clients like Claude Code, Cursor, and Claude Desktop can list your employees and talk to them directly.

Available over the API
Call an employee from your own product or workflow, synchronously or with polling for longer jobs. API access is available on Business plans and above.

They remember, and you choose the brain

Memory that persists

  • Employees save facts, preferences, and instructions from your conversations and recall them later
  • Scope memory to a single chat, the whole company, or across everything you do

The top frontier models

  • Anthropic, OpenAI, Google, Meta, Mistral, xAI, DeepSeek and more, all in one place
  • Choose the model per employee and set temperature and reasoning effort per role

Zero data retention options

  • Your data is never used to train models, and zero-data-retention models are available across the major providers
  • The platform picks a cost-appropriate model for each internal job rather than running everything on one

Configured in minutes, not a sprint

1. Pick the role
Choose a pre-built persona or describe the job. Only the job title is required — the name and system prompt can be generated for you.

2. Give it knowledge and tools
Attach documents or folders, then enable the capabilities it needs. Everything is off by default until you turn it on.

3. Put it to work
Chat with it, schedule it, or leave it on a heartbeat. Check the score on every run and refine the prompt from there.

Your questions, answered

Is this included in my plan?
Yes. AI Employees are enabled on every plan, including the free tier. Usage runs on credits, and your plan sets the monthly allowance. API access requires a Business plan or above.

How is this different from the Executive Assistant?
The Executive Assistant is one dedicated assistant for you personally — your calendar, your inbox, your admin. AI Employees are a team you build for the business, each specialised in a role and shareable with your colleagues.

Can an AI Employee answer my phone or reply to emails?
Customer-facing channels — website widget, shareable page, SMS, voice, and email — are configured as deployed agents, which can borrow an AI Employee’s persona. See AI Receptionist and Omni-Channel AI Agents for those.

What can it actually connect to?
Your knowledge base, smart lists, sequences, and content engine natively. Beyond that, employees reach 1,000+ business integrations through a tool router that discovers the right action at run time, and can consume external MCP servers you point them at.

How often can a scheduled task run?
One-time, daily, weekly, or monthly for recurring work. Heartbeat tasks run on an interval from every five minutes up to once every 24 hours, restricted to the active hours and weekdays you choose.

Who can see and edit my employees?
Employees are scoped to your company. Admins can lock an employee’s settings so members can use it without changing it. You can also publish one to the marketplace, or keep everything private.

Is my data secure?
Encrypted in transit and at rest. Your data is never used for model training, and most available models support zero data retention. View security details →

Hire your first AI Employee today

Start on the free plan, pick a role from the library, and give it something to do this week.