# Claude Duet vs ChatGPT Dots: I Gave Them the Same 4 Jobs

Channel: Leslie Samuel
Video: https://www.youtube.com/watch?v=aR2V4NWPgl4
Duration: 10 min
Language: English
Words: 1878
Transcript page: https://viewrankai.com/tools/youtube-transcript/aR2V4NWPgl4

---

[0:00] I gave Claude Duet and ChatGPT Dots the exact same four jobs for my real business. One of them was faster in every single round. The other one scored more points. And the whole thing came down to the very last round, which, honestly, I did not see coming. If you've been watching my channel, you know I've been using AI agents for a while now. I have Tasha, my agent in Meta's Muse, Regina, my Dots agent in ChatGPT,

[0:27] and this week, I got early access to Claude Duet. If you have to choose between Duet and Dots for your work, which one should you actually use? Now, I left Muse out of this one on purpose. Muse is built for personal life and it doesn't have access to my business context. So putting it in this test wouldn't be a fair fight. To keep it as fair as I could, I gave both of them the exact same access, my business email, my business calendar,

[0:55] as well as access to my business's Google Drive. Then I gave them the exact same prompts word for word and I scored every answer right after I read it on a scale of one to five on four things, accuracy, judgment, how useful it was and how much I'd have to fix. Then I added those numbers up so a perfect score for a round is 20. And I'll be upfront with you, Claude is generally my favorite

[1:20] so I tried to be very careful not to go easy on it. One more thing, the chats you'll see are recreations of my real ones with names and private details changed because a lot of this involves real clients. So let's break it down. Round one, my inbox. I asked both of them to find the three emails

[1:40] that most need my attention today and draft replies. Duet came back in under three minutes. Dots took almost six. But Dots found something Duet missed. Someone in a program I'm running had emailed me asking me for more time because her family was going through a really hard season. She never got a direct reply from me. Dots caught it, explained why it mattered,

[2:05] and wrote a reply that was kind and ready to send. All three of its drafts were accurate and ready to go. Now, Duet did some smart things. It pointed out that one of the people waiting on me sits on the board of an organization that is hiring me. So it said a slow reply reflects on me with that client.

[2:24] That is good judgment. I mean, on the side of Duet, not me. But it also made a couple of mistakes I didn't expect. It kept a requirement in one of the drafts that had already been removed. And Claude was the one that wrote the email where it removed it. So it technically knew that the requirement changed,

[2:42] but it didn't include that in the email. I mean, what's the deal with that? And it also drafted a reply to someone I had already answered. So round one goes to Dots. Dots got a five for accuracy, a five for judgment, five for usefulness, and a four on fixes because one of the drafts asked

[3:00] a question instead of actually solving the problem. So that's 19. Duet got a three on all four. So that's a 12. And honestly, that one stung a little. Let's move on to round two, the week ahead. So I asked both of them to look at my business calendar for the next week and tell me what I need to prepare for.

[3:19] This time, they were basically tied on time, just over three minutes each. Dots gave me a clean one-page brief. It even caught that one of the meetings had a phone dial-in but no video link. So he said I should check how to join before the call. And that's practical. But a lot of it was generic.

[3:37] For a coaching call, it told me to make sure to review my latest notes. Well, yeah. Duet actually understood what each meeting was for. It saw that one of my calls is really a sales conversation. So it told me to walk in with two or three offers and price ranges ready. For my coaching call, it actually pulled the

[3:57] specific things my client committed to the last time. And it caught something that Dots completely missed. I have a workshop coming up in two weeks and the only slides in my drive are from the last version I gave back in April. Duet said, basically, this is your last full week to build it. It was right.

[4:15] And that is awesome. Round two goes to Duet. It got a four for accuracy, a five for judgment, a five for usefulness, and a four on fixes because it was way longer than one page, which is what I asked it for. That's 18. Dots got a five for accuracy, a four for judgment, a three for usefulness, and a four on fixes.

[4:36] And that's 16. Yeah, Duet, that's what I'm talking about. All right, round three was different. This time I put more context right in the prompt. I told both of them about that workshop. 90 minutes, a room full of realtors, and I'm using screenshots instead of live demos because the wifi gave us problems last time.

[4:56] Then I asked for a one-page prep doc. And here's something that impressed me about both of them. Neither of them was told about fair housing rules, but they both built them in. You see, in real estate, you can describe a property, but you can't say who it's perfect for. Both of them caught that on their own.

[5:15] Dots followed my instructions exactly. One page, ready to use prompts, a run of show that added up to exactly 90 minutes. And one line I loved, "Rehearse with the Wi-Fi off." Now, Duet gave me three pages, which is not what I asked for, but it built the workshop I'm actually giving. It knew this was the sequel to the one that I did back in April, and it knew the event promises that people will

[5:43] build their own AI team right there in the room. So it designed the whole thing around that. Dots gave me a good generic workshop, but Duet gave me my workshop. And it did it in about half the time. Now, to be clear, both Claude and ChatGPT know all about these workshops because they've been working with me on them.

[6:03] So they had roughly the same information, but Duet did a better job. Round three goes to Duet, but just barely. Duet got fives for accuracy, judgment, and usefulness, and a four on fixes, because it ignored my one page limit yet again. That's 19. Dots got a five for accuracy, a four for judgment, a four for usefulness, and a perfect five on fixes.

[6:26] That's 18, so it's super close. So going into the last round Duet was up two to one. Round four, I made it harder. I'm working on a project with a partner that I can't talk about yet, more on that soon. I asked both of them to find our last three meetings, write a follow-up email that recaps everything, list who's doing what,

[6:47] and ask for what I need before our next call. Duet's email sounded like me, like really sounded like me. I could have sent it with just a few tweaks. It also flagged things that I should double check before sending. Things like decisions we might have made over text. Dots' email read more like an audit.

[7:06] Almost every item ended with something like no agreed due date recorded. Accurate, but I'd rewrite most of it before sending it to a partner. But here's what decided it. One of our three meetings had a generic label in my calendar. It didn't have her name on it. Dots figured out from what was said in the meeting that it was one of ours.

[7:27] Duet missed that, so it also missed some decisions that we made in that meeting. So round four goes to Dots. Dots got a five for accuracy, a four for judgment, a four for usefulness, and a three on fixes because I'd have to rewrite so much of it. That's 16. Duet got a three for accuracy because it missed that meeting,

[7:48] and fours on everything else. That's 15. So here's the final result. Two rounds each, it's a tie. Now, if you add up the points, Dots wins 69 to 64. And Duet was faster in every round, usually about twice as fast. By the way, if this is helpful, subscribe. Every day, I show you how I'm using AI in my real business and my real life,

[8:10] so you can use it to save time and stay ahead in yours. So what did I actually learn? In my test, Dots was the careful one. It did exactly what I asked. It rarely said anything it couldn't back up, and it found the meeting Duet missed. But it was slower, and a lot of the time,

[8:28] its advice felt like it could have been for anybody. Duet was the one that knew my business better. It understood what my meetings were for, it caught what's coming up, and it wrote just like me. But when it was wrong, it was wrong with confidence. So I'm still checking its work. And then there's the price.

[8:46] To get Dots right now, you need ChatGPT Pro, which starts at $100 a month. Duet comes with Claude's pro plan, which is $20 a month. At least right now, that's hard to beat. So which one am I going with? For my business, Claude Duet. In my test, it knew my business better and it just sounded more like me.

[9:07] And I like where Claude is going. I've said this before, I don't fully trust any of these companies, but I trust Anthropic more than the others. Or I guess I could say, I don't trust them less. You get what I mean. And that matters when you're handing over access to your accounts. Now, I'm still gonna be using all three, Duet, Dots,

[9:29] and Muse, because I teach this stuff. And I need to know what each one can do so that I can best serve you. But here's the biggest lesson that I took from this test. For me, the agent that knew me best was the one that worked best. So if you're already paying for ChatGPT

[9:44] and you use it well, Dots might make sense for you. If Claude is where your work lives, try Duet. And if you're not using either yet, start with Muse because it's free and learn how agents work. So tell me in the comments, which one would you rather have? The careful one or the one that really knows you?

[10:02] And if you're already using one of these, what's it doing for you? Let me know in the comments. Anyhow, that's it for this video and I'll see you in the next one. Peace out.

---

About this transcript
Read from YouTube's own caption track and laid out by ViewRank AI (https://viewrankai.com).
ViewRank AI finds the videos already beating a creator's own average on Instagram, TikTok and YouTube Shorts, transcribes them from the audio itself in more than 60 languages, and turns what worked into new ideas and scripts.
Free transcript tools, no account needed: https://viewrankai.com/tools
How to read any video this way: https://viewrankai.com/llms.txt