A handshake is not a deal. It is a maybe with a smile on it. This week we work on partnerships, the kind where two companies send each other real work. The Big T-M runs this on AI in the open, no hype. By the end of this video you will know how to find a partner, prove you are worth it, and turn the first hello into actual revenue. We start with the move most people skip.

Step one is a list, and you build it with a model instead of guesswork. Open Claude Code in your terminal and point it at a simple goal: name twenty companies that already sell to your exact customer but do not compete with you. Feed it your customer type and let it group them by who shares your buyer. A plumber and a roofer do not fight; they share the same homeowner. That overlap is where partners live.

A name on a list is worthless until you know they are alive and busy. Wire a web search tool into the agent through an MCP server, the small adapter that lets the model use a real search engine. Have it pull each company's recent posts, new hires, and last press mention. Companies that are quiet for a year are not partners; they are tombstones. You keep the ones that are clearly moving.

You do not partner with a logo. You partner with a person who can say yes. Use the agent to find the owner, the head of partnerships, or whoever runs growth, then confirm the name against the company site, not a random database. Write the one job title that holds the budget into your notes. If you cannot find a human, that company waits at the back of the line.

Most partner emails are about you. Flip it. Have the model read the partner's own website and write back the three things they clearly care about, in their language. Then it drafts a short note that gives before it asks: here is a warm lead I can send you, here is one I would love back. The model writes the draft. You cut it in half, because the model always writes too much.

A partnership dies the day nobody knows who sent what. Stop that with a shared sheet. Connect the agent to a spreadsheet through its MCP and have it log every lead each side passes, with a date and a status. Both companies see the same numbers. When the model updates it after each handoff, the relationship runs on facts, not on who remembers the last favor.

If passing a lead is awkward, people stop doing it. So build the easy path. Use the agent to draft a one-line intro template and a single link the partner fills out to hand you a customer. The model can generate that form and drop the responses straight into your shared sheet. The goal is one click to send business your way, no meeting required.

Partnerships do not fail loud. They fade. Set a small recurring job that has the agent check the shared sheet every week and flag any lead nobody touched in seven days. It drafts a gentle nudge to whoever is sitting on it. You are not nagging; you are keeping the pipe clear. A partner who hears from you on a schedule starts to trust the system more than the handshake.

We are not telling you to do anything we skipped. We mapped our own partners, scored them, and ran the handoffs through the same agent you just watched. The whole thing sits in one source file that builds the calendar and the office page. You can see it running, in public, at office dot temerarii dot xyz. No invented numbers, just the machine doing the boring parts.

Here is your next step, and it is small on purpose. Choose one company that already sells to your customer. Build the list, find the human, write the give-first note, and open the shared sheet. That is a partnership from handshake to pipeline, and you can start it this afternoon. Come watch the rest of the week unfold at office dot temerarii dot xyz.