SUMMARY
In this episode of The Check with Joseki Tech, President Adam Steidley discusses the advancements in generative AI for email personalization with Dan Bush, VP of Development, and Jasmin Salas, Director of Marketing. The team delves into refining the in-house email writing assistant, demonstrating a shift from generic to highly tailored messages. Key highlights include:
Generative AI Integration: Utilizing AI to craft personalized sales emails by incorporating customer-specific information such as skills and previous positions.
Improvements Demonstrated: A comparison of baseline and enhanced emails showcases significant progress in personalization, directly addressing the recipient’s professional background and potential interests.
Technical Adjustments and Feedback: Adjustments in AI settings to fine-tune the balance between personalized content and generality to avoid overly scripted messages.
I: GENERATIVE AI EMAIL PERSONALIZATION
Adam Steidley
Hey everybody, welcome back to The Check with Joseki Tech, where we're looking at our work in generative AI. My name is Adam. Today, joining me, I have Dan and Jasmin.
Dan Bush
Hello. Hello.
Adam Steidley
So Jasmin's our Director of Marketing. Dan's our VP of Development, and he's been our main architect of what we've been building here. So last week, we looked at the fact that we were taking extra information to pass into the generative AI to build that sales email. This revolved around the template and today's message. Now, we will add more information about the customer and see how that makes for a richer experience. So, Dan, do you want to show us what the baseline looks like here?
Dan Bush
Sure. So, let me share my screen. Let me know.
Adam Steidley
Yep.
Dan Bush
Now, the standard approach. We're going to go to send an email. Our plugin is going to kick in for HubBoost here. We'll hit hub boost to do generation. Position this window over here so we can scroll down through and look at these contact skills and positions.
Adam Steidley
And right now, we're sending the title, the name, and his About text.
Dan Bush
That's correct. So what we're seeing here is this contact. If we roll down here, you'll notice they have skills in the hospitality industry, customer satisfaction, and a few positions at different companies. Although we have all this good information, you'll notice that none of it's included here. We'll only see if the contact included some of that stuff in there. So it's way less personalized than it could be.
Adam Steidley
Absolutely.
Dan Bush
We'll look at the generation loop when considering skills and positions. Okay, here's a contact: Andrew Somebody. As usual, we will click the email plugin to kick in. We'll click Hub Boost to start our generation. As before, it's doing its magic by contacting the OpenAI. Here we have our generation.
Jasmin, do you want to take it from here?
II: REVIEW OF IMPROVED EMAIL
Jasmin Salas
Let's read this guy out loud and see if it sounds vaguely human:
Hello, Andrew. Your track record of turning big ideas into measurable results at ACCESSALL and beyond is nothing short of impressive. Your role as a Strategic Growth & Partnerships Executive, coupled with your rich experience in sales, strategic partnerships, and marketing, showcases your innovative approach to driving growth and engagement. At Joseki Tech, we're excited about the possibility of supporting your mission to provide customer-centric, emotion-driven solutions.
Our suite of technology solutions is designed to complement your expertise in hospitality, customer satisfaction, and your knack for crafting qualitative, quantitative solutions focused on sustainable growth. We believe our tools can help streamline your operations, enhance customer engagement, and facilitate innovative marketing strategies. Together, we can push the boundaries of what is possible and create even more significant impact in sectors you're passionate about, especially in promoting charitable programs that benefit children's education and health. Let's explore how Joseki Tech can be part of your journey towards operational efficiency and strategic innovation.
This is about as personalized as it gets. Someone took the time to read this guy's resume, understand his profile, and understand the kind of people and work he does. And I am involved in what we do with what they are doing. So it's not just us who can help companies like yours; we can help you achieve your specific goals. From what we've seen on your profile: this is what we think your goals are, what we think would work for you, and how we can work together in those ways.
So this is awesome, frankly, for the growth we've seen from the beginning with the most generic kind of email you're sending out to anyone, to now, this. Andrew Somebody is getting exactly what they want from this message. They understand what we are trying to sell them and why we think we are the best choice. So what does everybody else think about it?
Adam Steidley
Yeah, so it's interesting. And this is one of the things I feel about generative AI — it's about getting the right data in there. So we started with just LinkedIn, which is as good a place as any because it's a good free text of what people think about themselves. But now we're taking it to the next step, right? We're looking at the specific skills that give us more detail about what this guy is doing and their path from job to job. So, it lets us get to that next level. For other applications, there will be whatever is unique to your application; that will be the important point to put in there.
So whether that's that customer's past sales or if we look at that customer, if they've been to your website, what pages are they looking at it, that'll point us toward what products they're interested in. But it's really about getting that richest set of data into the AI to make appropriate decisions about what's going on.
III: TECHNICAL ADJUSTMENTS & FEEDBACK
Dan Bush
Yeah. I want to check and see visually and verify what it's doing here. If we look at this contact and scroll down, you will see that their skills are listed as hospitality in the industry. I am just trying to see if we can see this position — the strategic growth and partnership. I wanted to see some connectivity here. Right. Talking about startups here, did we have any language about startups?
Adam Steidley
What do we have as their roles and jobs? Strategic growth and partnerships. There's that. Let's loosen things up a little bit so we see the less verbatim text of what it's got, and there's. Isn't there a dial we can tune for that, Dan?
Dan Bush
I believe so. There's a temperature setting that we can adjust.
Adam Steidley
So, tweak that. So it's taking the idea from there, but not necessarily the verbatim words, because that's what the generative AI also offers, spreading it out. We don't want to look at what makes it look like we're cutting and pasting from the list of data.
Dan Bush
Yeah.
Adam Steidley
Cool. So it's a really good result.
Jasmin Salas
Another thing that I wouldn't necessarily include is the promotion of charitable programs and the benefit of children's education. It has nothing to do with the rest when I'm reading it. And it is just pulling in information to make it sound more personal, but it doesn't align with anything we are trying to say. So, in the final section, which says you're passionate about, especially in promoting charitable programs and the benefit of children's education and health, it's sprinkling in some information that isn't necessarily needed. Returning to what you said about turning down the exact verbatim from their profile would help ensure that we don't look like we're just spitting out anything that comes to mind.
Dan Bush
We can try this quickly in a feedback loop.
Adam Steidley
Right.
Dan Bush
Let's tell it to do that, giving us at least a second sampling.
Jasmin Salas
That sounds better. It sounds more professional. To reread that last paragraph, "Our technology solutions are designed to complement your expertise in hospitality, customer satisfaction, and your ability to craft qualitative and quantitative solutions focused on sustainable growth." I also noticed that the language changed slightly to craft qualitative and quantitative before there was a comma in between, which made it look awkward. So I like that it did update that.
Adam Steidley
Is it listening to the call? Because that's how you read it the first time.
Jasmin Salas
Maybe it's listening to us. That's the best part of AI. We can use information. Our tools can help streamline your operations, boost customer engagement, and facilitate innovative marketing strategies. We're eager to explore how Joseki Tech can support your mission toward operational excellence and strategic innovation.
Dan Bush
So, for next week, I'm hearing you guys want an Alexa skill that listens to the podcast and then trains the model, right?
Jasmin Salas
Ideally.
Dan Bush
All right, maybe we should start.
Jasmin Salas
Feeding in all of the transcripts from it.
IV: NEXT STEPS & CONCLUSIONS
Adam Steidley
That's a reasonable idea, but we can explore some privacy concerns between now and then. This is a great step around what we're seeing, and we're reaching critical mass on getting data in from all different angles. So we've talked about the knowledge base, the starting instructions, the feedback, and the information we have about the customer. So we've now got good, rich data for all those pieces, and now it will be about tuning and tuning from what we have there. But our plumbing is pretty well set at this point. That's a good landmark.
So, in our next game, we will look at some tools we've been building to help speed along the tuning process. So we can start looking at batches of data instead of onesie twosies.
So with that, thanks, everybody, and we will see you next time.