Jordan Metzner on why bots like Grok Bot and Muse differ: their own computer, email and LinkedIn access, and an email that got 33% off a sold out hotel.
The new wave of AI bots like Grok Bot and Muse is different because each one runs on its own browser and its own computer, using your accounts to do work while you are away. On Episode 57 of Built This Week, a hosts-only episode, Jordan Metzner explains how his bot emailed a sold out hotel and got him 33% off within about fifteen minutes. Sam Nadler, three weeks into using Grok Bot himself, pushes him for practical examples.
Jordan walks through the steps that got us here. First came pasting code into ChatGPT and copying it back. Then models moved into the sidebar and edited your code directly. Then came the Claude Code model, where you just talk to it and it edits your code. More recently, Claude Code and Codex learned to use your browser and the tools in it.
Jordan sees Muse, Dot from OpenAI and his favorite, Grok Bot, as a step change on top of that, and earlier tools like OpenClaw as the first iteration of bots.
But I think the big step function is with Grok Bot and Muse and some of these other ones is its ability that it has its own browser and its own computer.
— Jordan Metzner, Episode 57
The bot uses your accesses to build workflows. We have our email connected to Grok Bot, so it can prepare, send or reply to emails on our behalf, and Jordan has connected his LinkedIn. He can tell it what kind of people to reach out to and how many per day, and it targets that audience and makes the connections. Older automation software did similar things, he says, but it always ran in your own browser. Grok Bot uses its own computer.
His bigger point is about work. He expects these bots to come to the enterprise quickly as real copilots, for example a sales agent told to find an audience, customize messages, test what converts and respond to everyone in real time. Instead of chaining multi-step workflows together, you describe the outcome and let the agent do its work.
Watch the sold out hotel story in Episode 57 (starts at 5:43) · Episode 57 notes and transcript
Jordan wanted to stay at a specific hotel across the street from the office where he was working, and its website said it was sold out. He had Grok Bot handle it. The bot found the right person to contact, generated the email and wrote to the hotel asking whether there was a room and whether he could get a discount.
And I don't know, maybe fifteen minutes later, someone from the hotel called me up, offered me 33% off of the room rate
— Jordan Metzner, Episode 57
He was able to extend the rate to another guest as well, and says it saved about $200 on a one night stay. He called it "pretty light lifting." As he points out, there is not much to show on screen: it is a chat-based experience, and the work happens on the bot's side.
Jordan thinks agents are the beginning of a mass disruption of aggregators. Sites like Booking.com and Expedia exist because nobody wants to visit 20 hotel websites to compare prices for the same dates. An agent can find the best hotels in an area, check each website for the best price, and, if you have the time, contact each hotel directly.
He listed other uses where a bot searching the market for you pays off: big purchases like a used boat or a house, finding a specific used car and alerting you when new ones come online, generating and auto publishing social media content, and market research turned into a landing page. He also said he believes phone calls are coming soon. Who wins is still open in his view: Grok Bot, a product from Claude, or companies like Google and Microsoft that already own the enterprise stack. We covered Muse and the race for personal agents with Ivan Lee in Episode 56.
Sam asked why a bot on its own virtual machine feels so different from a model taking over your browser. Jordan compares tools like Claude Cowork to an employee who sits next to you: you tell them what to do and they do it, and they now run for a long time, but if you stop sitting next to them, they stop working.
A bot with its own server can run cron jobs and long, complex workflows, stay logged in and keep state. You don't need your laptop open for it to reach LinkedIn. You tell it to connect with 30 people a day and it uses its own browser to do that. We talked about agents that keep shipping overnight in Episode 51.
One was like you had to guide it, the other is you can just let it go do its thing and it'll take care of it for you.
— Jordan Metzner, Episode 57
Sam asked about how we have used agents to update all of our websites. Jordan says agents can go into Webflow and not only make edits but make improvements and optimize the site. He built his own JavaScript tracker so we can track traffic, make our own heat maps and understand user behavior on our sites better. He sees constant room to improve on design, SEO and ads.
On ads, he described agents generating AI content with voiceovers, text to speech, music and sound effects to make what he calls beautiful ads that he says are highly converting. He expects another step function improvement as agents get more power and control. Our earlier take on automated site content is in Episode 2.
This is a short, hosts-only episode with the hotel story, the LinkedIn workflow and the website agents told in Jordan's own words. Watch Episode 57 of Built This Week, where the full transcript is also on the page.
