Friday Quote Missed? This AI Workflow Drafts It From Email (Shawn Hain)

# Shawn at AIO IRL [00:00:00] **Shawn:** All right. Hi, everyone. My name's Sean. Um, I guess, I guess the question I've got for you, who here's had to get a quote Friday afternoon and you didn't get to it, and you wind up doing it on a Sunday morning? [00:00:11] I have. Probably everyone else. Yeah. A few of us. Keep that in mind. So I build automation systems for small businesses. Um, the boring back office stuff [00:00:26] that you don't necessarily really wanna do. It's not necessarily what you're good at. Done some work with trades and other businesses where you're a tradie. You're out there on the tools doing work. You don't wanna spend your whole afternoon just doing quotes, right? As an example. So this is the kind of stuff I'm seeing out that people are wasting time on that AI can help you with. [00:00:44] Um- [00:00:48] Sorry, guys, forgotten where I was gonna go with this. Um, and one of the things I have seen is if you don't get to that quote until a Sunday, you send it to them on Sunday evening, they get back to you Monday, they went with someone who got back to them on Saturday. [00:01:00] So people are actually losing business 'cause they're too busy. [00:01:03] So what I've built here is a tool that will essentially read your email as it comes in and build the quote for you. The-- One of the key things I've got within my business that I have seen a number of people make mistakes with, and I think, um, we've spoke about it before, AI shouldn't take away the judgment. [00:01:26] It shouldn't be sending things to our customers for us. So I do a quote, do a draft, and then get you to look at it and make sure it's correct, 'cause it does hallucinate, it does make mistakes. Is it the kind of work you wanna do? Is the pricing right? All that needs to be checked. So I will just pull up an email I drafted earlier. [00:01:48] This is just a nice, simple, straightforward quote... Oh, wow, sorry, guys. For a, uh, Sparky, three new power outlets. Yeah, garage upgraded to a thirty amp circuit. Now, I don't know if anyone knows electrical [00:02:00] trades, thirty amp circuits don't exist. Got a thirty-two amp, so we'll see if it picks that up. And I've also got two new, um, light fan combos where I'll supply the fan, so we'll see if it picks that up as well. [00:02:12] Um, and my name's just buried down the bottom, so hopefully it'll pick that. There are the three things I guess we'd look for. Just send that I'm just gonna wait for it to land before I run it. Now, normally this would actually pick it up on its own, but just for demo purposes, I'll manually trigger it so we're not waiting on it. [00:02:32] But now I'm waiting on my internet. And this is why I have a recording in case this breaks. There we go. Came through, so we'll trigger it. So I guess before I trigger it, I should probably walk through what it's gonna do. So it's basically six steps. First is reading the email, the second is the AI actually assessing what's in the email and breaking it down into what are they after, who is it, where are they? [00:02:57] The next part is loading in our [00:03:00] pricing and service sheet, then comparing what the customer asked for to our service sheet and matching them, so we can do fuzzy matching like the 32 amp circuit breaker. Then doing the math. No AI here. I don't let the AI do math. I don't trust it. It has done horrible things. [00:03:18] Um, and then generating a PDF and saving it to drafts and sending me an alert going, "Hey," ping the salesperson, ping the, the plumber going, "Hey, you've got a quote to look at," so they know that when they get off the tools to have a check. So we'll just run it and see what happens. So it's pulled in the email. [00:03:39] It's now reading the email and trying to understand what's happening. The AI parts can take a little bit. Sorry, can't make it go faster [00:03:53] Yeah. It, it can be, um, hypnotizing, I guess. Now it's pulled in the pricing information. [00:04:00] One thing I probably should show you, sorry. One of the things as well is as it's going, it's actually logging all of this to a spreadsheet so that you can actually see what's happened [00:04:12] And as it goes along, thinking again, now it's doing the... Sorry, I can't actually see that. My eyes are shot. Oh, no. I'm gonna leave it alone. So that's the, um... It's now comparing what the customer's asked for to our service sheet and trying to work out how to match those together. Now it's building the quote. [00:04:32] Well, it's pricing, I should say. Mm, and going through generating my email, and logging away as we go. So it took about a minute and a half to do that. And of course, my Slack doesn't work Swear I tested this earlier. Oh, that's[00:05:00] [00:05:00] Okay. Well, Slack is broken, so I'm just gonna leave that alone. What it should do when it's working properly is send you an alert. It'll come up on your phone or on your computer so that you know that that's happened. And I'll jump over back to my Google. And you can see there it's now got a draft. So it's ins- saved it for you so you don't have to write it. [00:05:22] And [00:05:33] And so that's the quote that it's pulled up there. So it's captured my name It's caught the 32-amp circuit breaker, which is a big one. You can't tell them with, uh, something that doesn't exist, and it's not charging them for the fans 'cause they're picking those up. Now, the Slack alert that didn't work also includes a warning that, "Hey, I made an assumption," which is I went, "There's no 30-amp breaker, so I'll go to a 32-amp," so that [00:06:00] whoever's looking at it knows to double-check that that's correct. [00:06:03] So that's the kind of stuff that I've built for customers as part of larger systems rather than just a one-off thing, more comprehensive. Um, that's basically it from me. Um, if you don't have any questions. [00:06:30] Yeah, so I'll spend time talking with the customer and I'm trying to understand their business 'cause I'm, I'm not a tradie, right? I build software. I don't know the last thing about installing a circuit breaker. I wouldn't try. Um, but I'll spend the time talking to them and work out what do they do, what kind of questions they get off customers where the customers get it wrong, they don't understand. [00:06:49] And if we've got access to previous data, like what was mentioned in the previous talk, um, we'll use that to help train the model to know what to do. If it's not available, we [00:07:00] can capture that through ongoing monitoring of the systems as well [00:07:26] Yeah, that's it. Exactly. So we can build fairly complex things here. If you can describe it as a system, we can describe it in some way that the AI can understand as an algorithm or a process [00:07:41] That's it [00:07:49] **Dave:** Thanks, Sean. Live demos are always sketchy, especially when it's not your home territory and we're at a pub. This is someone's home territory, but, uh, not a demo's. [00:08:00] Ser- Sergio [00:08:03] **Speaker 7:** Uh, so how, how you charge the customer? You, you charge for a consultancy in the beginning and then you build it, or you charge for a subscription that you are doing frequently? [00:08:14] How, how, how this is working? [00:08:18] **Shawn:** I generally do project-based pricing, so I'll sit down and work out what the customer needs, often break it into stages. So rather than like doing everything all at once, we'll build systems that then build on top of each other. 'Cause I find that as we build these solutions, it's often we start with one idea, build some of it, and go, "Well, hang on a minute. [00:08:35] The reality actually looks like this. We need to shift focus." Where I might charge an ongoing fee is if they need technical support to maintain the system. So that can be maintaining the prompts or the online services that are needed to make it continue working [00:08:52]