Opening Keynote

Watch to learn more about what to expect with the summit and hear the perspective of OpenAI's Head of Sales.

Speakers

Aliisa Rosenthal - Head of Sales at OpenAI

LinkedIn

Aliisa Rosenthal currently leads sales and partnerships at OpenAI, supporting enterprise customers building AI integrations into their products.

Aliisa started her career journey in the field of Artificial Intelligence at Quid, where she held the role of Director of Strategic Partnerships. She then moved on to become the first sales hire at Mixpanel, run Enterprise Sales at InVisionApp, and serve as VP of Sales at WalkMe. Aliisa graduated from Brown University and lives in San Francisco with her husband, two children, and chocolate lab.

Nik Mijic - Founder & CEO at Matik

LinkedIn

Nikola (Nik) Mijic is the CEO & co-founder of Matik. Prior to Matik, Nik worked in various roles and companies that focused on helping customer success teams retain and grow customers. At LinkedIn, he built internal tools that helped them leverage LinkedIn data in their presentations and be more efficient while doing so. Nik was the first non-engineering hire at Bluenose Analytics, where they built a platform that used predictive analytics to engage at-risk customers and identify drivers of churn.

Watch Opening Keynote

‍Transcript

This transcript was created by AI. If you see any mistakes, please let us know at marketing@matik.io.

Matik MC: Hi, everyone. Welcome to Matik AI and customer success summit. You are currently watching the opening keynote for which I'm super excited to introduce our speakers. We have here, Nik Mijic, cofounder and CEO of Matik, as well as Aliisa Rosenthal, head of sales at OpenAI.

Nik Mijic: Awesome. Well, welcome everybody to our very first AI and Customer Success Summit as Bex just mentioned, my name is Nik Mijic and I'm one of the co-founders of Matik. And if you don't know what we do, we automate the creation and sending of data driven content within PowerPoint and Google Slides. And we do that by personalizing text, charts, images and tables that we insert into these presentations. So, I, as a result, we are very plugged into the customer success community and have spoken to a lot of executives as well as individual contributors that are on the front lines and CSM and account managers. And a recurring topic that continues to come up with no surprise is the role of artificial intelligence and how AI can impact the customer success function. So that is the primary reason why we're doing the summit today is to get not only thought leaders within customer success, but also thought leaders in AI all under one roof to be able to chat about and have a dialogue around the impact of AI in customer success. So, I do wanna thank all of our speakers who are attending the summit and giving their time and sharing their insights and perspective. Really, appreciate that as well as all the attendees. I know that you all have busy calendars, and appreciate you taking the time to be here with us today. So if you can't attend all the sessions, do not worry. We will be recording all of the sessions. And anybody who's registered for the summit will be getting a link to the recording after the summit. So, awesome. Well, to kick things off, I am joined here with Aliisa on who is the head of sales at OpenAI, and they've been at the forefront of this AI movement. So, Aliisa, thank you for joining us today. I'll love to pass it over to you to introduce yourself a little bit more, and also maybe give folks who aren't as familiar with OpenAI a little bit of context as to what you guys do.

Aliisa Rosenthal: Sure. Yeah, thanks, Nik. Great to be here today. So my name's Aliisa. I am the head of sales at OpenAI. I've been here a little over a year and that year, things have changed rather dramatically. OpenAI was founded as a research lab about seven, a little over seven years ago, pretty small research lab that probably not a lot of people had heard up until recently. And, and OpenAI’s mission ultimately is something called ag, I, and this means essentially a general intelligence that can perform work as well as human. So an autonomous system that can perform work and that's our goal. We're not there yet, but all of our efforts and research are really with the goal of getting towards API. Along the way. We are producing a lot of different models modalities. So different modalities like speech, vision, text, et cetera. And I think the world got their first taste of our models with the launch of ChatGPT, which was last December and that was taking one of our text models and basically putting it into a really easy to use user interface. No we list and that went a bit viral, I think you can say, and people all over the world really got their first exposure to a large language model and the power of AI, and generative AI and where we are today. So, I OpenAI is a research lab that then takes those models and turns them into products. So that's really where I come in and my team comes in. The, the main product that we sell right now is our API. So we make our models available via API so companies can come and build on those models, integrate them directly into their products for their own end users?

Nik Mijic: Awesome. Appreciate the context. What was it about opening a product that really got you excited, to join the company in the beginning?

Aliisa Rosenthal: Yeah. You know, it's funny because I think originally I wasn't necessarily, I didn't really know who OpenAI was. I knew someone who worked here. I was digging around actually, one of open is customers documentation, and discovered GPT and reached out to my friend here and said, hey, can I learn a little bit more about what you guys are building? And he said, hey, why don't you come meet with us? We're thinking about building a sales team. So I met with, you know, Sam and Mira and most of the leaders here about a year and a half ago and, it was a bit of a gamble. I mean, there really wasn't much of a product yet, but it was the coolest thing I've ever seen. You know, this was pre dolly, which is our age model, but we had it in a form where I could start to test with it. And the first time I use dolly and I generated that image, I mean, now, I know, these text image models feel like they're ubiquitous. But this was a, really mind blowing experience for me interacting with dolly, interacting with GP T. At the time. It was GP T3, a slightly older model. But even then, you know, the magic, of sending my first API call and getting a completion and it really felt like magic. It's the right word for it and it's funny. I think once you experience, the AI, the generative AI technology where it is now, you can't see it, you know? And once.

Nik Mijic: I…

Aliisa Rosenthal: Once I had felt the impact of it, suddenly everything else felt irrelevant. It just felt like I need to be here. I need to be at this company. I don't know how we're going to productize it. I don't know how we're going to go to market. I don't know if they need a sales person, but I just want to be part of.

Nik Mijic: That's awesome. Yeah, I remember my first time you think same back experience was blown away. And it was just, I think the prompt, what I ended up inputting was just write an e-mail back to responding to a customer and I was just so impressed at the detail and how good the writing was. That really just blew me away. So I definitely had a, very, similar experience. So there's obviously a lot of application on how people are leveraging your guys with API. What are some of the misconceptions about AI that kind of bother you the most? And that you guys see in the market?

Aliisa Rosenthal: Yeah. I think I'll answer that in two ways. I think there's misconceptions about AI in general and then maybe misconceptions about open. I specifically. So I think in general, the big misconception we hear is that I need to take the model and I need to train it on my own data. I need to finetune it. It's not gonna really be helpful for me unless I can finetune it on my company's data. And the reality is these models are so all purpose and so intelligent. You don't really need a ton of data to make them useful. And there's also, there's different ways to give them access to data or to make them speak in the voice you want. So if it's a voice branding issue, you want the model to respond in a certain way that's where prompting can get you really far. I just putting in a prompt that says, speak with these words, don't use these words, your gens, your millennial, your, in your French, you know, basically, you can tell the model, you can give the model personality without having to really use any data to train it. And similarly, a lot of a lot of…companies or, you know, builders working with our API, assume that the only way to give it access to additional data is by fine tuning it. And the reality is there are a lot of other ways to get access to data. So embedding as a common approach retrieval. And basically, this means that you are giving get access to an embedded set of information that it can search across in its responses without having to retrain a whole model. So you don't need and also says you don't need an ML team. You, you don't need an AI team to start to work with our models. You just need basically a developer who can work with python to integrate these models directly into your end user product. The second misconception I'll say, and this is more about OpenAI than AI is kind of similarly going back to the data. I think there's a really big fear that people have that anything that, you know, they use OpenAI for will then be used to train our models. And we do not train on any data sent via API. So any companies building on our API that data is, it's a totally different infrastructure than the data we used to train our models. So that's the other misconception, I hear a lot of fear and uncertainty and nervousness about using our API because people are afraid if I put it in the model, then someone else can come along and see the same thing that I put in which the training doesn't work. That way. It's different than, you know, like putting it in, a search engine or any other technology we're used to, you know, anything sent via API, is never trained on.

Nik Mijic: Makes sense. And I guess going back to the first point about the prompt engineering, are you seeing certain applications of prompt engineering? That, can you maybe give me some examples of how people are doing that prompt engineering as better find in those models?

Aliisa Rosenthal: Yeah. So we see a lot of companies say putting in their brand guidelines. We see, you know, like fund examples. If, you know, I have a customer who's built sort of a shopping assistant. And the prompt is, you know, your age speak using hash tags, use this sort of, you know, language, be brief, be concise, you know, those sorts of prompts can help really give, the chat bought if you're building a chat bought experience a lot of personality.

Nik Mijic: Awesome. Very cool. And I guess what are some of the applications of AI outside of just like B to B tech that you're most excited about? Maybe the team at OpenAI, you guys are most excited about?

Aliisa Rosenthal: Yeah. I mean, I think healthcare is a really interesting space and I think we're really just scratching the surface with healthcare. You know, there's certainly applications starting to build with us doing things like, you know, summarizing doctor's notes, writing soap notes, making it easier for pharmacists to understand the doctor's notes and relay instructions to patients. I think there's going to be an explosion and innovation and healthcare. And there's so many really fascinating applications of AI both on vision and on sort of the, you know, CT scans and radiology, on the tech. So understanding doctor's notes translating them from pharmacist and for patients. And then you start to extrapolate out to the future. So there's where we work today and what the models can do now. And then there is a Gi and maybe even super intelligence, which I know I start to get a little meadow when I talk about this. But this is what gets me really excited. You know, I think often when people think about opening, I get asked all the time, you know, what or how do you feel about the future? And I am extremely hopeful and optimistic that API will be a force for amazing scientific research and discovery. And you can imagine a world where we had, you know, autonomous systems where they focused on just curing cancer or solving carbon capture. And you were able to funnel compute into systems who just spent all day every day, almost imagine you had an endless list of intelligent humans just spending all of their time researching a problem. The odds that we cure cancer, the odds that we solve carbon can capture suddenly skyrocket. So, this is not where we are today. This is a more futuristic perspective. But this is what makes me extremely hopeful and optimistic and excited about Gi.

Nik Mijic: Yeah, healthcare, I think there's definitely, a lot of opportunity. I totally agree with you on that. And I think not just for the healthcare providers but also putting more control in the hands of the consumer and being able to have more inside there from a consumer perspective. So totally agree with that. And then, I guess, you know, in terms of a lot of the folks that are attending the summit are in the world of customer success, right? You're in sales. You've worked at other companies like WalkMe, where you've partnered up with customer success, and then it yourself, you know, where do you see the potential AI unlocking, you know, benefits for the CS team?

Aliisa Rosenthal: Yeah. I think where you know, ChatGPT is really great just out of the box is in understanding new industries, understanding new personas. So imagine you're propping for a meeting or, you know, you've got a meeting tomorrow with maybe a new type of role that you've never had to interact with before. Maybe it's a head of, you know, marketing or engineering or product or design and, you know, they're coming into a meeting and you want to understand, what do they care about? What is someone in this role? Typically focus on what, you know, what are some terms I should know and understand this is where that can be a really great teacher. Similarly if it's a, if it's a brand new industry, if you are suddenly selling to an auto manufacturer, and that's not really your wheelhouse teaching me everything I need to know about auto manufacturing, who the big players are, what are the trends? I should be aware of what are some things that this person might be stressed out about as the, you know, head, of an auto manufacturer on the supply chain side? Like how do I prepare myself for that? So it's really good. I think for both sales and marketing, and sales and customer success for understanding, you know, industries personas, people you're meeting with helping you be really prepared, helping you draft bullet points, talking points, etcetera. I think that's where we are today. I think when you extrapolate out and we start using, you know, in sort of like the customer success operations side. You've had a chance to play with code interpreter, which I think is one of the coolest tools that we've offered. But it basically is a data scientist. It's a it's a data scientists that you can upload files to and ask questions. So on the operational side, you know, here's all of my customers and their usage and create prioritization tiers for me. So I know who to focus on, you know, here's, all this data and organize it for me, and basically what code interpreter does is it's using python behind the scenes to perform like a data scientist. So if you have customer success, you know, leaders, or even I see is on the team who want to perform intricate analysis of their own books of business. And maybe don't have access to a data scientist. It's I mean, I use it all the time personally. I'll upload our leads or accounts and I'll say, help me organize, help me prioritize, help me tier, help me create territories so that's where they're I think, the analyst or operational part is really interesting as well.

Nik Mijic: Makes sense. So, what I guess a follow up question to that, you know, the models are obviously getting better and better that you guys are coming up with. I think there's been a lot of talk about hallucinations, right? Like what the model spits out, how accurate is it? How, how do you feel like teams are corresponding with that? And, and overcoming that, is it with the prompt engineering? They're just continuing to finetune the output, but we'd love to kinda get your thoughts on that.

Aliisa Rosenthal: Yeah. You know, the models certainly aren't perfect today. GPT4 is much better than three and three point five on the hallucinations, you know, as every GP T next comes out, that problem will eventually go away or be mitigated pretty substantially. That said, my general guidance to customers right now is don't have an open ended, you know, generate chat bot that you're building into your product for your end users, have some sort of guard rails in there. And there are different types of guardrails. You can build on the most conservative risk of our side of the spectrum. You could set up a chat bot that basically says only pull answers from my documentation. And that's using this tactic called embedding I referred to earlier where, you could embed your documentation and you could tell GPT to search across your documentation and it's responses. You're only returning data that you are absolutely 100 percent certain. Is that is like, the most strict, you can go what we typically see our customers doing right now is using GPT for things like data transfer transformation like summarization categorization. So like a common use case is taking support tickets and maybe not auto responding to them quite yet. I, or if you are auto responding, having a human just review those to make sure they're all accurate, but doing things like autosummarizing categorizing and routing. And that can really speed up the process responding to tickets. If you're not comfortable having, you know, completely open ended generative responses to all your support tickets. And I think that's where we are in this, the journey of this technology. We'll get there where you need fewer and fewer guard rails and you already need way few with GPT4 then FPT three point five. But yeah, we encourage have some human oversight and, or use this as a transformational role, transformational way or use embedding so that you're just pulling data from your documentation or sort of approved language.

Nik Mijic: Awesome. Very cool. Go ahead.

Aliisa Rosenthal: Yeah. Well, I was gonna say, I would love to turn one of these questions back on you, which is, what are you seeing on the customer success side with AI?

Nik Mijic: Yeah. I think there's obviously a lot of application within the CS workflow. And I think one of the things that I'm really most excited about is really bridging the gap in the last mile dilemma. When it comes to data enablement, you kinda mentioned code interpreter. I think it's something along those lines where I think in today's macro economic environment. They need to be able to articulate and demonstrate ROI is now more imperative than ever. And that usually lands on the side of the CSM to be able to go and take data and showcase the ROI and value that your product or service has delivered. And so for us, I think tangible examples that we see customer success managers and just CS function in general, putting together business reviews, right? Putting together ROI analysis, putting together benchmarking report. It's not just pulling data for the sake of pulling data and showcasing and doing a data dump, but it's being able to take that data, whether it lives in a dashboard, whether it lives in a CRM and then being able to analyze that data, spit out meaningful key insights and takeaways that hopefully tie back to value and type back to the objectives that your customers purchase your product for. And that is the piece that AI has such a huge potential to be able to not just pull the data but actually analyze the data on your behalf to, and give you at least a starting point on what the key takeaways are, right? Like data is never perfect, but it can get you to the one yard line. And so a few things that we've done on our side that I'm really excited about is today, our product has the ability to go in, pull data from a BI tool or database or CRM, the populated chart, a table, a text box within our PowerPoint or goal slide. But we now have the capabilities with AI to be able to say, hey, not only send the data to this chart that's in the presentation, but I analyze that data and spit out the key takeaways and bullet format off to the side, right? So this is what you're doing manually before now, you have the ability to work. I can go and do that for you. And really kinda instead of you being the data analyst, you can spend more time on building the relationship with your customer. And I think that is something early on that we've seen has provided a huge benefit to our customers. And as we're talking with the folks in the market, the other that I'm really excited about is that we've dedicated resources for, but we're calling presentation an answer. And this is an extension within Google slides that will spit out executive summaries and talking tracks for your slides. So again, you know, are not necessarily data analysts as you mentioned before. And our hope is with this extension that we can go ahead and do all the heavy lifting for you. Whether it is a presentation that came from Matik that we generated on your behalf or if it is a presentation that you put together manually extension will scan all of the data and the content that is on the slide and spit out an exact summary that you can put at the beginning. You could share in a follow up e-mail right? As well as even the talking points going back to what we call data enablement. How do I speak to this? How do I tell the story in the right way? And I think AI has a great way to be able, to bridge that gap and close, that last mile dilemma. And at the end of day uplevel the customer success managers to better do their job and better build a relationship, with their customers. So that is something that I think is really exciting. I think, from an AI standpoint, is to help bridge that gap when it comes to I enablement. So I know that we are pretty close to the time. Again, the whole goal of the summit today is to really dive a little bit deeper into AI and the impacts that I can have on customer success. So again, I wanna thank all the speakers that are going to be participating in today's summit, thank you for your time. Thank you for sharing the insight to our hope for all the attendees again is that you do have the ability to take something away from the summit that's tactical that you can bring back to your teams. We are going to be doing some workshops at the end of the day that's part of the agenda that will showcase how to leverage AI and ChatGPT within your day to day workflow. So really do appreciate everybody attending and thank you so much for the time today and spending with us and getting the summit kicked off and looking forward to continuing the conversation.

Aliisa Rosenthal: Yeah, my pleasure, Nik. Thank you.

Matik MC: Awesome. Thank you. Both Nik and Aliisa for helping kick off the summit. As Nik mentioned, the presentation enhancer will be available for free to all summit registrants and attendees. So just keep an eye out for an e-mail post summit that will share the link to the presentation enhancer as well as we'll share, let you know a little bit about when session reporting will be available and all the other good details. So, thank you everyone for attending the opening keynote and hope the rest of your day goes well.

---

Click here to see all sessions & speakers, and navigate to recordings.