Your first thirty minutes
Setup is a conversation, not a form. The assistant asks one question at a time, and your workspace fills in as you answer — goals, then a business model, then roles, projects, processes and integrations. Half an hour is typical; a complicated business can take forty minutes.
It is worth not rushing. Everything the agents do afterwards is built on what you say here.
Signing in
You can continue with Google or with your email address. With email there is no password: you get a link, you click it, you are in. The link is single-use and expires, so if you come back to it an hour later, ask for a fresh one.
BossForce is in private beta, so you will also be asked for an invite code. A code has a limited number of seats; if yours is full, use another one or leave your email and you will hear when seats open.
What you will be asked
The conversation moves through a fixed set of steps, and each one produces something real in your workspace before the next begins.
Who you are and what you do. Your name, your work, how you prefer to be addressed. This is short.
How things are today, and what hurts. The raw material for everything else. Be specific and be honest about the mess — "I have no idea where my money goes" is a more useful answer than a tidy one.
Your root goals. One to three, no more. Expect the assistant to push back here: most people name an instrument ("run my Instagram", "hire a marketer") rather than an end. It will ask why, more than once, until it has the actual objective — and then keep the instrument as a means to it.
Your business model. How the endeavour makes money or creates value, and which approaches are off the table.
The goal tree. The drivers underneath each root goal — the handful of numbers that actually move it — with targets and how they will be measured.
Roles. Which agents your goal tree needs, what each one owns, and where the red lines are. This is also where you say what you are and are not willing to hand over.
The work plan. Two to five workstreams, turned into projects (one-off outcomes) and playbooks (processes that repeat).
Integrations. Which external services those processes need. The assistant reads the tool catalogue and offers what is actually available — see connecting tools.
At the end you get a short recap, a saved checkpoint of the whole setup, and an offer to run the first heartbeat.
How to answer well
Answer in whatever language you prefer — the assistant follows your lead, and the interface itself is available in six languages.
Say the awkward parts. Constraints, dislikes, the client you refuse to work with, the tone you cannot stand: an agent that knows them produces work you do not have to rewrite. An agent that does not know them produces generic work, and generic is the main way this fails.
When you are asked something you have not thought about, say so. The assistant will offer two or three concrete options rather than leave you staring at the question.
If you change your mind about something from an earlier step, just say it. Nothing is locked; the assistant updates the entity and checks what downstream depends on it.
One workspace, one endeavour
If you bring two unrelated things — grow the consultancy and also get fit — they belong in separate workspaces. Goals, agents and memories are shared inside a workspace, and mixing two lives there makes every recommendation worse. How many workspaces you can run at once depends on your plan; see credits and plans.
If you have to stop halfway
You can, but finish soon. Until setup completes, the workspace is half-built: some goals exist, the roles may not, and the agents have no full picture to work from. If you tell the assistant you need to get to work right away, it will create a minimum viable workspace and pick the rest up later — but the sooner you finish, the less generic everything is.