Everyone Is About to Have a Gatekeeper

OpenAI put out a demo this week for dots, always-on agents living inside ChatGPT. It's the same week I've been running Meta's Muse, so the race for the personal assistant is officially on with real money behind both sides.

Most of the reaction I saw was about what these things do for the person using them. That's the small half of the story. The bigger one is what happens to everybody trying to sell that person something.

The demo is worth watching for about two minutes. The agent names itself Alfred, builds out a website for a fall launch, updates a board deck with live numbers, catches an app migration out of somebody's Google Meet notes, reschedules a finance review that collided with a daughter's recital, books the kids' after-school lessons the morning registration opened, and finds a replacement cake vendor for a wedding before the person even hears the first one canceled.

Read that list again and notice how many of those are purchases.

A vendor got replaced. Lessons got booked. A calendar slot got taken. In the version of the world that demo is selling, a business somewhere lost a booking and never found out it was in play, and another business won one without ever talking to a human being.

Now the usual caveat. This is a scripted product video, everything works perfectly, and the real thing will be clumsier than that for a while. I've watched enough launch videos to know the gap between the reel and Tuesday.

But the direction is not in question, and the direction is the part that should get your attention.

Here's the shift underneath it. For the last ten years the whole game in lead generation has been speed to lead. First one to reach the customer while the problem is hot wins the job. Every CRM, every autoresponder, every text-back-in-30-seconds tool is built on that one idea, and it worked because there was a person on the other end with a phone in their hand.

Put an agent in the middle and that assumption breaks. The customer isn't sitting there waiting for your callback. Their assistant is out collecting options, filtering them, and coming back with two. Speed still matters, but it's speed to the agent, not to the person, and the agent isn't in a hurry the way a homeowner with a leak is.

If you've ever sold B2B, you already know exactly what this is. It's a gatekeeper. You spent years learning how to get past the executive assistant to reach the person who could actually say yes, and you learned that the gatekeeper doesn't care how good your pitch is if you can't get through the door.

That's every consumer in about ten years. Maybe sooner.

If you sell anything to anyone, the question stops being how fast you can respond and starts being whether you're one of the two options the agent brings back.

The uncomfortable part is that this isn't a marketing problem, it's an operations one. An agent evaluating you can't be charmed. It can only read what's actually there. Your hours, your service area, your prices, your response time, your reviews, whether your booking flow can be completed without a phone call. Everything you've been meaning to clean up and haven't is about to be the whole pitch.

Three things worth doing about it this month.

Find out whether you're already visible. Ask an assistant not just to recommend your category in your town, but to actually complete the task the way a customer would. Book me a detailer this week. Find me a mortgage broker who works with self-employed buyers. See what it says and see whether you're in it. That's your ranking now.

Make yourself answerable. Every question a customer asks a human should exist in writing somewhere public. Pricing ranges, what you do and don't do, turnaround times, what happens after they book. Not because anyone reads your FAQ page, but because the thing deciding whether to recommend you does.

Then close the loop a machine has to walk through. If booking you requires a phone call during business hours, an agent will route around you to someone whose calendar it can reach directly. This is the one most people will be late on, and it's the one with actual revenue attached.

And if you're not selling anything, the move is simpler. Go pick one of these things, dots or Muse or whatever your platform gives you, and run a real week through it. Not a test. Real work, where you'll find out fast what it's bad at.

I've been running Muse for a couple of weeks and the useful parts were not the ones I expected.

The Facebook data is the headline for me. I pulled my follower data out and into my own database, which is the kind of thing that normally means an export request and an afternoon. It also gives me genuinely useful feedback on posts, the sort of insight that's technically in the analytics but buried where nobody digs. Having something tell me "next time try this instead" is worth more than another dashboard.

The part that surprised me is Marketplace. The agents are workhorses there. If you have ever thought about flipping anything, the grunt work that made it not worth your time is mostly gone.

None of that is why I'm writing about it though. I'm writing about it because two weeks of using one made the commerce implication obvious in a way that reading about it never did, which is the general lesson. You don't learn what these things change by watching demos. You learn it by giving one something real and watching where it goes.

Paint Protection Pros is fully indexed and Google is throwing a little test traffic at it, which is the phase where you want to declare victory and there is nothing yet to declare. Too early for numbers. I'll report them when there are real ones, including if they're bad.

The other thing I've started is cinematic AI video. I've been interested for a while and never given it the time to get good, which is exactly the kind of thing I write about other people avoiding. The bet is that being able to produce footage that holds attention, without a crew and a shoot day, becomes a real advantage for ads and social over the next couple of years.

No Scoreboard this week.

Go ask an assistant to actually hire someone in your category, in your town, this week. Not to describe the category. To do it. Then look at whether you were an option, and if you weren't, you know what you're working on this month.

See you next Wednesday.

Nick

If you're a business owner wondering where AI actually fits in your sales process, or you're building a team and trying to figure out what AI fluency should look like, I'm happy to talk. Book time at nicksakkis.com.