Your first week with an AI agent team: a day-by-day walkthrough

Day one takes one sitting. You type your goal in plain language — "more clients for my web-design studio in Rotterdam" is a complete, working brief. Bro responds with one to three short clarifying questions: who exactly you want to reach, in what region, through which channels, and what success looks like. Answer in single sentences; precision beats length. What you say here shapes every agent that gets built, so name the niche and the area rather than keeping it broad.

Bro then proposes a team — typically two to seven agents, each with one sharp responsibility. For a client-finding goal that usually means a Finder that sources local businesses, a Qualifier that scores each one, and an Outreacher that drafts first messages. Read the proposal like a hiring plan: you can adjust an agent's brief or drop one before saying yes. Approving the team design is deliberately your first approval — the pattern of agents proposing and you deciding starts before any work does.

Once you approve, the team appears on a living canvas — one island per agent — and you can watch each agent's actual reasoning show up on its timeline as it works. Nothing else is required from you today: the Finder runs on OpenStreetMap out of the box, so there are no API keys to hunt down, and the 14-day trial starts you with 500 credits and no credit card.

Come back on day two and there is work to react to: the Finder has surfaced a batch of businesses in your niche and area, and the Qualifier has read each website and given it a 0–100 outdated score with a written reason. Resist the urge to accept everything that looks plausible. Your job today is judgment, not throughput.

Dismissing a bad fit is not cleanup — it is teaching. Every lead you reject tells the Qualifier something concrete about what "worth pursuing" means for you: too big to need a freelancer, already redesigned last year, a category you don't serve, a location that's technically in range but practically isn't. Ten deliberate dismissals on day two make day ten's list noticeably cleaner. For each lead, a quick three-part check is enough:

Read the written reasons, not just the scores. Because the score is explainable, you can disagree with the reasoning instead of the number — and that kind of specific correction is exactly what sharpens the next qualification pass.

By day three the Outreacher has drafts waiting in your Approvals queue. Each one opens with a real, specific finding about the recipient's website — that's the agent's core discipline — and every draft has already passed a second, stricter review that strips out invented specifics, hype, and filler before you ever see it. So what reaches you is decent. Decent is not the goal.

Pick the draft closest to something you'd actually send and rewrite it properly: your greeting, your sentence rhythm, the phrases you'd never use, the one detail you'd lead with instead. When you approve an edited draft, the agent distills your changes into a lasting lesson that shapes every draft after it. One thorough edit on day three is worth more than twenty light touch-ups in week three — front-load the teaching.

Nothing has been sent yet, and nothing can be: outbound email goes through your own sender, which you haven't connected. Approved drafts simply wait. That's not a limitation to work around — it's the design. The queue is where speed and control stop being a trade-off.

By day four, reviewing stops feeling like a task and starts feeling like a habit. Block ten minutes at the same time each day and work the Approvals queue in one pass: approve what's right, edit what's close, dismiss what's wrong. Reviewing is free in the literal sense — approvals, edits, and briefs cost zero credits; credits only move when an agent does actual work — so being a demanding reviewer costs you nothing but those ten minutes.

This is also the day to connect your sender: a Resend API key or your Gmail account via OAuth, stored in an encrypted credentials vault. Brohns never sends from its own domain on your behalf — every email leaves from your address, under your name. Before the first send, set your guardrails: a daily send limit, a send window (business hours reads far better than 2 a.m.), and your do-not-contact list.

Then approve your first sends. The approval step is enforced on the server — the recipient and content come from the database, and every approved action lands in an audit log — so "nothing goes out without you" is architecture, not a settings toggle. Hard bounces are automatically marked never-contact-again, which protects your sender reputation from the very first email.

Agents don't observe weekends. While you're away, the Finder keeps sourcing, the Qualifier keeps scoring, and new drafts stack up in the queue — but nothing outward-facing happens without you. Monday's queue will be fuller, not scarier, because everything in it still waits for your yes.

The weekend is also when the autonomy question naturally comes up. After five days of reviewing, you know exactly where the team is reliably right and where it still needs your eye. Every team starts at approve-everything, and the autonomy ladder lets you grant routine autonomy per team when you're ready — independent within limits you set. Honest advice: don't grant it in week one. Let the lessons from your edits and dismissals accumulate first; you can step up later, step back down anytime, and there's a kill switch if you ever want everything stopped at once.

Here's the part most onboarding guides skip: replies take time. Local business owners answer cold email in days or weeks, not hours, and even a sharply targeted first message lands in a busy inbox. Judging your agent team by week-one reply counts is like judging a new hire by their first Friday. Measure the pipeline instead: leads found, qualified, and contacted with messages you genuinely stand behind.

What does show up in week one is compounding quality. The Qualifier's lists get cleaner with every dismissal, the Outreacher's drafts drift steadily toward your voice with every edit, and your ten minutes of review buys more finished work each day. The effort you put in this week isn't overhead — it's the training that makes week four mostly approvals.

If you're starting today: write the goal one level more specific than feels natural, answer Bro's questions in full sentences, and plan to be a tough editor on day three. The trial's 500 credits comfortably cover a first week like this one — finding and scoring a batch of leads costs 5 credits, a qualification pass costs 2, a draft costs 3, and all of your reviewing is free.