Onboard your first AI Employee
A new Genosyn account opens a five-step first-run guide at /c/<company-slug>/onboarding: it explains what an AI Employee is, hires one, connects the AI Model it thinks with, turns the best recurring work into Routines, offers Gmail access, and ends on a summary of what is now running.
Name your company
Signup lands on Name your company — "Genosyn runs a company with AI Employees working alongside your team. First, the company they will work for — we explain the rest on the next screen." Fill in Company name, and optionally Mission (optional) and Vision (optional), then choose Create company and continue.
The mission field carries a hint explaining why it is worth writing: it is used to pick the recurring work Genosyn suggests for your first AI Employee. Both fields are optional and can be written later — from the launch plan step itself, or from Settings → Company.
Genosyn creates the company, adds you as its owner, and opens the guide automatically. The company is the home for its human Members, AI Employees, Connections, and work. To create another one later, open the company picker in the top bar and select + New company.
The five steps
The guide is headed Set up your first AI Employee and shows a rail of five steps, each with a one-line hint. On a narrow screen the rail collapses to Step N of 5 and the step name. Completed steps stay clickable, so going back to change something is always allowed — every step re-reads its state from the server rather than trusting what the guide thinks it did.
- How it works
- What an AI Employee is.
- AI Employee
- Hire one and connect a model.
- Launch plan
- Recurring work for the role.
- Gmail
- Optional mailbox access.
- First request
- Watch them work.
Steps 1 to 4 run in order and end on a summary headed You are set up. The first request is reached from that summary rather than before it, and its back button reads Back to summary.
Step 1 — How it works
The guide explains the product before asking you to configure it. This step defines every word the rest of the flow leans on, at first use: Soul (their written constitution), Skills (markdown playbooks, one per piece of work they know how to do), and Routines (work that runs on a schedule instead of waiting to be asked, where each firing becomes a Run you can read line by line).
How a working day goes walks the same loop the product actually runs: a Routine comes due on its schedule; the AI Employee reads its Soul and the Skills for that job; it does the work using only the Connections you granted it; it writes up the Run — and if you gated that Routine, it waits for your approval before acting.
The step closes with a preview of what is left — hire an AI Employee and connect an AI Model, review their launch plan, connect Gmail, give them a first request — and the button Hire my first AI Employee.
Step 2 — AI Employee
- Pick a starting role under Who should join the team? — Executive Assistant, Sales Development Rep, Research Analyst, or Operations Coordinator. Each card names the Skills and the Routines that template ships with, and states each schedule in plain English, so both words are taught by example.
- Review Name and Role, then choose Hire AI Employee. Selecting a template fills these in but never overwrites a name you already typed.
- Connect an AI Model in the panel that follows. It links straight out to Get an Anthropic key and Get an OpenAI key. The key is encrypted before it is stored, and this one model powers both chat and every Routine.
- Choose Build the launch plan. Genosyn checks for an active, connected model before advancing.
Step 3 — Launch plan
The Launch plan is the same surface you get after hiring anyone from AI → Employees. Genosyn combines the employee's role, chosen template, and the company mission and vision to recommend useful first work. The same inputs always produce a stable plan, so refreshing or returning to the guide does not reshuffle it.
If mission or vision is missing, the plan opens with Give the recommendations your company's direction. Fill either field and choose Save and refresh plan to rebuild the suggestions against real context.
Suggested Routines
Each card is a checkbox tile with its schedule in plain English and one of three states:
- Scheduled — the Routine came with the role and is already running. It stays in the plan so you can see what is live, and it cannot be selected again.
- Suggested — not added yet. Select up to five at a time.
- Selected — chosen, and waiting to be created.
A note above the list is explicit about what selecting means: anything you add starts running on its own schedule straight away, using the AI Employee's AI Model and so the credit on the key you registered, and any of them can be turned off at any time from Routines. Two controls create them:
- Add selected Routines — inside that note. Creates the selection and stays on the step, so you can keep working through the Integrations below.
- Schedule N Routines and continue — the primary button, which creates the selection and then advances. With nothing selected it reads Continue to Gmail instead. If the write fails you stay on the step with the error rather than moving past it.
When something is selected, Continue without adding them appears beside the primary button as the way to move on and create nothing. Either way, Genosyn confirms with N Routines are now scheduled for {name} and re-checks the employee before creating the batch: it never overwrites an edited Routine, and a matching name or slug comes back as already scheduled rather than being duplicated.
Recommended Integrations
The plan also recommends enabled Integrations that fit the work, and defines the three words in one sentence: an Integration is a connector type; a Connection is one account your company links through it; a Grant is this employee's access to that one Connection. A recommendation never creates access by itself — its status tells you the next safe action:
- Connect — the company has no Connection for this Integration yet.
- Grant needed — a healthy company Connection exists, but this employee does not have a Grant to it.
- Attention — a Connection exists, but it must be repaired or reconnected before the employee can use it.
- Ready — the employee already has a Grant to a healthy Connection.
Step 4 — Connect Gmail
The step is called Connect Gmail, not Email, because it only connects Gmail. Other mailboxes connect later from Email → Integrations. What you see depends on what already exists:
- A Google Connection with Gmail access but no linked mailbox offers Connect mailbox, which links it and grants access in one step.
- A linked mailbox with no Grant for this employee offers Grant draft access.
- Nothing connected yet shows Connect Google Workspace and a Connect Gmail button that walks the OAuth client setup and Google approval.
A note names all three access levels: Read browses threads; Draft — what this step grants — also writes replies, applies labels, archives, and marks read, so the employee can clear an inbox and leave a finished reply in the thread while a human presses Send; Send is never granted here. Change the level any time at Email → Settings → AI access. See Email.
You are set up
The guide ends on a summary rather than dropping you into a chat box. It reads the company's onboarding status back from the server — derived from real state, never a stored flag — so it reports what is genuinely true. Under What is set up it confirms four things:
- The AI Employee and their role, with a count of the Skills they arrived with. Hiring without a template creates none, and the row says so rather than claiming otherwise.
- AI Model — connected, or not connected yet with a link to connect one. When it is missing, the heading also warns that scheduled Runs will be skipped until it is there.
- N Routines scheduled — counting only the ones that will actually fire, with the time of the next Run and a link to turn any of them off.
- Email access — the level actually granted (read, draft, or send), or a pointer to add a mailbox later.
Where everything lives then links the three places this work continues: AI → AI Employees to read and edit the Soul and Skills, AI → Routines for every schedule, its on/off switch, and the transcript of each Run, and Approvals for the work you have gated. The two exits are Go to Home and Give {name} a first request.
Step 5 — First request
Pick a starter request or write your own. Genosyn opens the AI Employee's chat with that request filled in, but does not send it — review it, add company context, and send when ready. Every example asks the employee to check with you before creating or sending anything.
- Create a Routine — turn a recurring responsibility into scheduled work. The employee asks about the outcome, inputs, and time before creating anything.
- Find potential Contacts — define an ideal profile, then research a focused list. It never invents contact details and sends no outreach.
- Triage my inbox — group unread mail into needs a reply, needs a decision, and FYI, and draft replies without sending.
- Plan your first week — turn the role and Soul into concrete outcomes and the work that should become a Routine.
Or write your own takes any text and Open chat carries it across. Back to summary returns to the completed-setup screen.
Leave and come back
Every step after the first is skippable, and leaving is not a one-way door. While setup is unfinished, Home shows a Finish setting up {name} banner — or, before anyone is hired, that the company has no AI Employees yet — with Finish setup or Open the guide leading back to the right step. Because the status is derived from the company's real state (an AI Employee with a connected AI Model), the banner disappears by itself when setup is finished outside the guide.
You can also open /c/<company-slug>/onboarding directly. The URL keeps the current step and employee, and each step re-checks the employee's Routines, Connections, mailbox, and Grants so completed setup is never repeated.
When hiring from AI → Employees instead, the same Launch plan appears after the Soul review and can be skipped the same way. Skipping creates no Routines, Connections, or Grants. All of that work can be done later from AI → Routines, the employee's Connections page, and Settings → Integrations. See AI Employees for the full lifecycle.