My First Week with ChatGPT Dots.

A conference is a firehose. Golf pairings, a lookbook of attendees, a pocket full of business cards, and notes I can barely read by Friday. Most of the value is in the follow-up, and most of the follow-up never happens.

This time I ran it with ChatGPT Dots. It had just come out, so this was my first week using it, learning as I went. I called it Jarvis, gave it a British voice, and talked to it more than I typed. Here's how it went.

Before: know who you're playing with

Jarvis researched my golf partners ahead of time. For each one I got a few memory cues and a couple of conversation starters.

Four hours on a golf course is the best networking at any conference. Walking to the first tee already knowing something real about the group changes the whole round.

During: one place for everything

Business cards, the attendee lookbook and my notes all went into a networking tracker in SharePoint. For each person: contact details, LinkedIn profile, my priority, the follow-up, and where each detail came from.

That last column sounds fussy. It isn't. Three weeks later, "where did I get this number" has an answer, and I can trust what's in the sheet.

After: the follow-up actually happens

• LinkedIn. I chose who to connect with. Jarvis sent the requests and tracked who accepted. When people replied with contact info, it pulled that into the tracker. Recurring checks keep it current. Outreach is still going.
• Email. Jarvis drafted each note from what I'd written down about the conversation. I rewrote what needed rewriting and sent every one myself, with my own signature. If you got one, I read it before you did.
• CRM. I added Event as a deal lead source in HubSpot, and made the event name required when it's picked. Next conference, the pipeline will show where the deals came from.

Two things I learned

I kept saying thank you. Every "thanks, Jarvis" burned tokens on a polite reply nobody needed. I knew that, and I did it anyway. That's the real test. You don't thank a tool. You thank someone who just took work off your plate. By midweek it felt like working with a sharp assistant, not operating software.

Use each tool for what it's good at. Claude handles context better: long threads, messy notes, keeping track of who said what. Dots is better at doing things. So I did the thinking in Claude and copied the results over. When Claude wouldn't do something I asked, I had it write step-by-step instructions and handed those to Dots. The tool that understood the problem wrote the plan. The tool that could act on it did the work.

The takeaway

I went home with more real follow-ups from this conference than any before it, and I didn't spend my evenings typing up cards.

The agent didn't replace my judgment. It made sure my judgment got applied to every contact instead of the first ten. That's the same thing we build for clients at Cay Digital, and it was good to feel it from the user's side.