What event pros actually do with AI, when nobody's selling anything

A few weeks ago, I got a group of event pros on two calls, from Sydney to Dubai, and kept it off the record on purpose. No vendors in the room, nobody performing on a stage. Just planners and agency owners comparing notes on how they use AI day to day. Not the demos. The real workflows that are saving people hours this month.

Four of them are worth passing along.

The planner who fired her timesheets. One person in the room runs more than 1,200 events a year, and she stopped filling out timesheets entirely. AI reads her activity from a time tracker and her calendar, then writes the weekly hours-per-project sheet on its own. The report she used to dread now arrives done.

The hundred-dollar speaker scout. Another needed a niche expert on a budget that wouldn't cover a sourcing agency. So someone suggested pulling competitor conference lineups and running them through a few Sales Navigator filters. One person had landed a major bank's head of blockchain security for an event in Amsterdam, on a tool stack that runs him about $100 a month.

The business wiki that updates itself. One founder built an internal agent that stitches together his documents and meeting transcripts so the knowledge doesn't live in his head. Drop a single vendor invoice into it and it updates the budget behind the scenes. New hires stop re-asking the same questions because the answers are already connected.

The inbox that reads itself first. An event planner in Dubai gets more than a hundred messages a week across her channels, and maybe twenty are real events. I suggested she use AI to read her entire inbox using Unipile and draft replies to the few that are real. I run a version of this myself. The rule everyone in the room agreed on was identical. AI drafts, and a human approves before anything is sent.

A quick word on tools, since it came up. The room leaned toward Claude over ChatGPT for the heavier work, and a few people pointed newcomers to Cowork as the gentler on-ramp before Claude Code. Nobody was dogmatic about it. Use whatever you will actually open every day.

If there was one throughline, it was this. You build the automation with AI, then keep a person on top of it to check the work. The wins were never the flashy demos. They were the boring, repetitive tasks that eat up a Friday afternoon.

One caution worth repeating, because it got real in the room. The risk with all of this is not the model, it is what you connect it to. A few people had stories about new tools requesting access they had no business having, and one pointed to a recent breach in which a fake compliance badge opened the door. Wire AI into systems you already trust, and be slow to hand a brand-new tool the keys. Someone even suggested having a privacy lawyer glance at your setup once a year, which is cheaper than the alternative.

I run these roundtables every few months, off the record, so people can be honest about what is and isn't working. If you want a seat at the next one, just reply, and I’ll make sure you’re on the list.

Till next time,

Noah

Do More With Less Using AI