Leads without a sales team
Istvan projects light onto buildings on four continents and never had a sales department. A plain ask gave shallow research. With his own knowledge as the direction, the agent found over three thousand lines of leads, then organised, filtered and scored them.
- Where
- In the terminal, with Claude Code
- Size
- A longer one
- Short link
- piy.show/s124
Twenty years of light, and no sales team
Istvan projects onto buildings. His studio has done it since 2006, on four continents, and the work has always arrived by being found. They never had a sales department. The question he brought to the show was the one they had never solved: who actually commissions a thing like this, and how do you reach them.
It became the longest session so far, four hours. He left with around three thousand organisations, a survey on his own site to score them, and a domain bought on air to run it from.
Start with the problem you have never solved, not with the tool. The tool is the same for everybody. The problem is yours, and that is where the agent needs you.
Why start with the problem and not the tool? Prompt it yourself
We have never done sales. Our work has always found us. Who commissions work like ours, and how would we reach them? Ask me what you need to know first.
The plain ask gives plain research
First we simply asked the agent to find some leads. Istvan said wow. I said it was shallow, and it was. A plain ask gets the average of what everybody already knows.
The clearest test I know came out of that session. The agent is good at what everyone does and bad at what only you do. Finding customers is something thousands of people do, so it had plenty to say. Knowing which building, which festival and which city buys a light show is the part only his studio knows.
So the first answer is not the result. It is the starting point that shows you what the agent does not know about your business.
Why was the first research shallow? Prompt it yourself
Before you research, tell me what you assumed about my business and my customers. I will correct it, then you start.
Give it your knowledge, then let it dig
This time we used his knowledge to give the agent actual research direction. He named two competitors. The agent collected their projects, their clients and the locations, and built a library of who buys this work. Then the events: a lead library is not a list of events, it is a calendar of decision windows, because the supplier is chosen long before the opening night.
Four research streams ran at once, in the terminal. It came back with about three thousand three hundred sourced lines. Too much to read, and exactly what we wanted.
You bring the knowledge. The agent brings the hours. Name the competitors, name the clients you lost to them, name the cities, and let it chase every line.
What knowledge should I give the agent first? Prompt it yourself
Here are two competitors. Collect their projects, their clients and the locations. Then find the events and venues that commission this kind of work in the next years, and when the supplier gets chosen.
Too big to read? Ask for a filter
Three thousand lines is a wall. The mistake is to ask for a shorter list, because the agent would cut it by its own rules, not yours. Istvan asked for a better way to see it instead.
His real prompt, typed in the moment: good start, three thousand names is a huge list, so create me an easy filter where I can find and make initial scoring with locations, and whatever the sales teams usually use for filtering leads. That is the whole secret. He did not know how sales teams filter. He told the agent to use what they use.
When a result is too big, ask for the tool that lets you read it. A filter, a score, a sort by the date the decision is made.
What do I do with a list too big to read? Prompt it yourself
This list is too big to read. Make me an easy filter where I can score every lead by location, by budget and by when they choose a supplier.
I can help you build yours
I build things like this with an AI agent, on my own computer. In a private session I help you do the same with your idea. Or bring it to the show, free.
Ten versions, and a plain verdict
The filter came back as a page with score bars, and his reaction was the one you see in the Short. It is crazy. The rest of the session was shaping it, and two habits made the difference.
Ask for ten versions, because you can. You would never ask a designer for ten goes at something. The agent does not mind, and the asking is free.
Stop apologising to it and say what is wrong. Four hours in, Istvan talked to it the way you talk to a contractor, with a plain verdict on the work. The blunt version got better results every time.
How should I talk to the agent when the work is wrong? Prompt it yourself
Give me ten versions of this filter page, side by side. Then I tell you which one is right and what is wrong with the others.
Your knowledge, its hours
You bring your knowledge, and your agents help to amplify it. That is why I do this show. If you do not learn how to use it, your competitors will.
For your own business the shape is the same. The thing you have never done, the knowledge only you have, and an afternoon.
What would this look like for my work? Prompt it yourself
Here is what I know about my customers that nobody else knows. Use it as the direction for the research, and tell me where my knowledge runs out.
Start with your own leads
Name two competitors. Ask for their clients and the events they won. Then ask for the filter, not the shorter list, and score every lead by the date the decision is made.
What do I do next? Prompt it yourself
Here are two competitors. Collect their projects and clients, find the events that commission this work, then make me a filter where I can score every lead.
Do it in the terminal
I do this with Claude Code, the AI agent in the terminal. It works in a folder on your own computer, so it reads your files every time and can look things up for you.
A web chat works too, but there you paste everything in again in every new chat.
-
Get Claude Code
It comes with Claude Pro, US$20 a month, paid to Anthropic. Claude's plans
-
Make a folder for it, and open it
mkdir -p ~/projects/my-leads && cd ~/projects/my-leads
-
Start the agent
claude
-
Ask it
Here are two competitors. Collect their projects, their clients and the locations. Then find the events and venues that commission this kind of work in the next years, and when the supplier gets chosen. Run the research streams at once and bring it all into one list.
Never opened a terminal? In a private session I put Claude Code on your computer with you, and you start your first idea.
Prompty