Reading your whole market every morning is a full-time job. So hand that job to something else.
A groundbreaking, a funding round, a new leader taking over, a brand opening in a market it has never operated in. Every one of those signals sits in public news for weeks while everyone waits for a job posting or buys the same list from the same vendor. The people who move early are reading. Covering an entire market every morning is what stops the rest of us.
The Starter builds the piece that does the reading. One agent, one job, pointed at one market you name.
Three blocks. That is the entire thing.
Each block gets an explanation before you paste it, so you know what you are building rather than copying magic words. Total time is around twelve minutes, and most of that is the agent working while you watch.
Block one · about one minute
You build the specialist
One file, written in plain English, that defines the agent: what it hunts for, which sources it reads, what it is forbidden to invent, and the tools it is allowed to touch. Open it afterward. Every rule the scout follows is a sentence you can edit.
Block two · five to seven minutes
The scout reads your market
You give it two lines: what you do and the market to watch. It searches the trade press, the regional business journal, local outlets, and the chamber news, then returns the five strongest signals with the company, the link, the outlet, the date, and every person named in the story.
Block three · about four minutes
You tune it and go again
The scout grades its own run, marks which sources produced findings and which produced nothing, researches five better ones for your market, and reruns against the list you approve. Skipping this step is the difference between a party trick and something worth opening on a Monday.
FINDING 3 of 5 Company: [name of a developer in your market] What: broke ground on a 312-room convention hotel, opening 2028 Source: [your regional business journal], July 14, 2026 Link: [live URL, intact] People named: [name], SVP Development · [name], GM designate Why this matters to you: the leadership team for a 2028 opening gets assembled through 2026 and 2027. No posting exists yet. The GM designate is named in this article.
Shape of one finding from a real run. The contents come from your market, your sources, and this week's news.
What you need before you start
- A laptop and about twelve minutes.
- The Claude Desktop app, free from Anthropic.
- A paid Claude plan, currently $20 a month. Here it is plainly rather than buried in a footnote further down the page.
- One market you actually care about, named in a single line.
That $20 is why the Starter can be free. The scout runs on your subscription rather than on mine, so nothing about this costs me anything per run and nothing about it expires. Anyone telling you a free tool has no cost behind it is hiding where the cost went.
If you have been on the fence about paying for an AI plan, this is a decent way to find out whether it earns the money. Twelve minutes on your own market answers that better than any review.
Send me the three blocks
One email with the Starter attached, arriving in a couple of minutes. Three days later I send one more asking whether it ran, because the answer tells me what to fix. That is the whole sequence.
What you learn is worth more than the agent you keep
The Starter teaches one pattern, all the way through, on your own machine. It is called Specialists, and it is the piece most people get wrong first.
One agent, one job, a narrow brief, and a named list of tools it may use. Look at your scout file afterward and every sentence in it is about finding signals. Nothing in there is about writing, or about remembering you, or about deciding what happens next. That narrowness is the reason the output is specific.
The common mistake goes the other way: one enormous instruction that asks a single agent to research, decide, write, and remember, which produces four kinds of vague. Narrow the job and the quality climbs immediately.
What one agent cannot do, said plainly
You will hit all four of these within about a week of running it. None is a defect in what you built. Each is a piece the Starter leaves out on purpose, and each one has a name.
It forgets you
Every run, you retype who you are and which market to read. The fix is a memory file the agents read first.
It forgets what it found
Run it twice and Tuesday's story returns on Thursday looking new. A ledger fixes that, and it is also how you avoid contacting the same person twice.
It stops at findings
Nobody identifies the right person inside the company, nothing carries an honest confidence label, and nothing gets drafted in your voice. Those are more specialists, written recipes, and a conductor deciding who runs when.
It waits for you
Nothing arrives at seven on Monday morning without you asking. Scheduling is what turns a chat session into a system.
And if you want the rest of it
The Agent Builder Sessions are two hours, live, seventeen numbered blocks, and all five patterns explained as you paste them: memory, several specialists working together, written recipes, a conductor, and a ledger. Your first real run happens in the room, on your market, around minute 75.
You leave with the whole system running on your own laptop, a handout that rebuilds it from scratch, and the recording. Nothing it produces sends without you, which is the same promise the Starter makes.
No rush on this. Run the Starter first. Someone who has pasted three blocks and watched them work knows exactly what they would be buying, and that is a much better place to decide from.
See the sessions and the dates Ask me anythingQuestions people ask
Do I need to know how to code?
No. Three blocks get copied and pasted, and the agent itself is a plain English file describing a job. If you can make a folder on your desktop and paste text into an app, you can finish this.
Will it send emails or post anything for me?
Never. The scout writes files and shows them on screen. That rule is written into the file you paste, in plain language, and you can open it and read it. Every system I teach carries the same rule.
Does this work outside hospitality?
Yes. The scout cares about signals and sources, neither of which belongs to one industry. Hospitality is where I built it and where my examples come from. Construction, professional services, healthcare, niche B2B, and local markets have all worked, because block three researches your market rather than assuming mine.
Is this a stripped-down version of the paid thing?
It is one real specialist out of a system that runs five patterns. What you build works on its own and keeps working after you close the page. The parts it leaves out are named on this page rather than hidden, because a crippled demo would tell you nothing useful about whether the full build is worth two hours.
What happens to my email address?
Two emails. The Starter arrives in one, and three days later I ask whether it ran. Nothing else goes out unless you reply. Your address never gets sold or shared.
What if it does not work on my machine?
Reply to either email and tell me what happened. Almost every problem is one of three things, all covered in the troubleshooting section of the Starter itself. I read every reply personally.
Who wrote this
Michael J. Goldrich is the founder of Vivander Advisors and teaches agentic workflows to industry leadership groups, including the AHLA Global Technology 100. He runs these same patterns every morning on his own business, and the scout in this Starter is a trimmed version of the one that reads his market.