Why Your Next Executive Assistant Will Be an AI — Deon Nicholas (Espa.ai)

19:21Deon Nicholas:That is usually when she likes to tackle this kind of issue. Let's put something on her calendar, and then she'll be able to do that. So I suppose able to do something like that. For ones that require a simple reply, like, hey, can we meet tomorrow? You know, something like that. It's going to look up your calendar. It's going to draft a reply in your voice, and then it's going to send that email back to the person. So there's this ability to execute. And then the second piece is your context. Again, you can have all the intelligence in the world, but if you don't have context on the person, then you're kind of screwed.

19:50Deon Nicholas:And so when it comes to context, we think about that in terms of both personalization and integration. And so one of the secret sauces of Aespa is that Aespa really rapidly learns you from, for example, your emails and your patterns. Like, oh, Ann likes to go for a run at 5 p.m. Dion has kiddo time from 7 to 8 a.m. And little things like that, it's able to learn from your communications and from your scheduling so that when it's scheduling for you or when it needs to answer a question or do whatever, it just knows how to do that. It learns your tone of voice so that if you hate emdashes, I'm actually an emdash guy.

20:31Deon Nicholas:I'm not going to lie. I'm so mad that everyone's trying to get rid of the emdash. But if you hate emdashes, it will learn those nuances. is learn how to talk like you. You speak formal to your LPs, and you're pretty colloquial to your founders or your one-line chat. One of my friends was literally, she talks in shorthand, very Gen Z. Instead of saying for real, she'll just say FR when talking to her friends over email or whatever. And her ESPA has learned that. And so you can see that.

21:01Anne Dwane:Okay, so this is different than prompting because I see in the product there's some custom instructions, but you're saying way beyond that. It's observing patterns. It's observing patterns, exactly.

21:10Deon Nicholas:And then actually that actually ties into the last piece, which is privacy, which is actually in so many ways a pillar. So the fundamental question is how can you get personalization and privacy at the same time? And we're actually doing a ton of really fun, innovative stuff around how to do that. But in short, like nobody wants their data being sucked up and used by, you know, open AI or whatever. But at the same time, you do want the personalization. And so you can do that either through prompting, prompt engineering, or you can do that through models, right? And over time, we're actually going to do both.

21:43Deon Nicholas:But with models, if you're training a model on somebody's data, then you have to train that only for them, right? And so I think in the long term, there's going to be the resurrection of this idea of small language models rather than large language models that are monolithic, trained on everything. Imagine a small language model that can either run on your phone or run on a local machine or just run in a way that only you have access to it, but it is constantly learning you. So that's one way to think about personalization with privacy. Another way of doing that is through the creation of quote-unquote memories, right?

22:12Deon Nicholas:Like this idea of which eventually become context and documents or whatever back into the model. But it's human readable. It's human editable. So imagine if you can, and you can do this with Aespa today, you can literally audit and see everything it's learned about you in plain English. And if you don't like something, like, you know what? I do want to bring back the em dash. Well, then let's just go and, you know, let's just go and edit that, right? Oh, that's awesome. So imagine being able to do all these things in ways that the user can control, right? Like no training on your data in some monolithic model.

22:41Deon Nicholas:And anything that the model is learning about you, you get control of. And then there's a lot of technology and a lot of algorithmic prowess that you kind of have to build into the model in order to be able to do that at scale. So very, very excited about these kind of pillars, right, from execution to like the context layer, but then also through to privacy.

23:00Anne Dwane:Well, and I saw an actual demo the other day. We were at a meeting and you spoke into your phone and said, introduce this person to this person. And almost instantaneously, the email was sent out. And so it's like the self-completing to-do list. Exactly. That was amazing.

23:19Deon Nicholas:Which is super cool. Yeah. And so that's such a fun use case. And again, I believe in and as part of the context pillar, complete integration into your life. The average person doesn't actually want to interact with AI by sitting in front of a text box. Right. The average person, again, if you're running a business, you're a founder, you're a VC, whatever, you're out and about, you're doing things. And so for me, one of my favorite things to do with Espa is literally that voice memo on the go. You know, I'm at a dinner party or I'm driving even like you can just voice voice speak. so it's hands-free, all of these things.

23:50Deon Nicholas:Yeah, and that was a really cool case. I had met someone who was starting a fund, and he wanted an intro to an LP. And I was like, oh, you should meet my guy Robbie over here. And, yeah, and then I just voiced. I was like, hey, introduce Robbie to Satish, and boom, boom, boom. And it worked out, right? It just happened. It just happened. Yeah, yeah, it was amazing. And that goes, again, to that execution pillar. So that's how we think about that flywheel with something like Aespa.

24:16Anne Dwane:Got it. And let's zoom ahead to the future. What does it look like? So obviously Apple wants to own the user, you know, Google and Gmail and G Suite and all the agentic frameworks. So what do you think in a few years the world looks like of individuals with agents?

24:36Deon Nicholas:Yeah, what I'm finding is there's like this really interesting kind of white space in a sense, because all of the people you would expect to be building something like this have perverse incentives on some level. So what I mean by that is like take, you know, the open AIs and the anthropics of the world, right? They, on the one hand, they're very positioned to build like a great agentic product. But on the other hand, A, they are incentivized to be building better and better models. And how do you do that? By having people's data and doing that sort of thing. So the idea of privacy, like you can't have zero data retention from like these bigger models unless you're a company or an enterprise or something like that, which is why like we, you know, as Team Aespa have negotiated that so that we can give that benefit to our users.

25:24Deon Nicholas:Right. And so I think that there's a little bit of like misaligned incentives. And then I also think on this at the same time, focus is a little bit different. Right. Like you saw OpenAI completely cancel Sora. Right. Which would have been a great consumer product. but they're focused on their revenue bottom line, which is like finding probably more and more enterprise use cases around, you know, GPT or cloud or whatever it is there. On the other hand, you have like the open clause of the world and they're effectively just built for developers. Very powerful technology. I think really, really cool stuff.

25:56Deon Nicholas:But like the average person is not that. Like, again, it's Silicon Valley. It's us building for ourselves. And so when you actually look at it, like all the people who kind of have potentially an in here end up kind of creating their own form of lowest common denominator. And so when you think about the average person who has the need for an agentic AI assistant to actually be able to take action in their lives, it's going to be quite a while. You know, in the absence of like aespa, it's going to be quite a while before you're going to be able to get access to that. I think like it's really five, 10 years away in most cases.

26:30Deon Nicholas:So that's why we're really building this and we're trying to accelerate that timeline so that everyone, again, small business owners, freelancers, creators, like we actually have music producers using Aespa right now and swearing by it, right? And it's kind of crazy to see that, right? So I think in five to 10 years, we are going to see a convergence. I think the best way to describe it is very similar to the cursor model. You know, AI coding should not really have been a category. Like it was just like, you know, there was a GitHub co-pilot here and all these other things there. And like there was never, there should never have been a now$60 billion company built on coding agents.

27:08Deon Nicholas:Or there should never have been a cognition, right? Like in all those things, my portfolio company. But what happened was the great builders started to focus and go all in on creating a 10x better experience. And in some levels, the model layer labs followed. And they've been chasing. And I think Cloud Code is actually very, very good. But again, you still have cursor, which many people are like, well, there's no moat. Should they die? Like, what's going on here? But you realize that once you have the breakout capacity and differentiation, you can actually create a moat, right? And in their case, they're either going to build the model or go with SpaceX, right?

27:44Deon Nicholas:Yeah, right. And so I think that's just fascinating. And so right now, for the next two, three years, a lot of these companies are probably focusing all their energy in chasing this market. I think the AI, personal, exec assistant, whatever you want to describe it market is going to be a fast fall. It's going to be the next domino where you're going to have a few companies who are going to be able to create breakout amazing products. And the model layer labs are going to start to kind of turn inward and be like, well, this is the next thing we have to either try to own or buy or build or whatever.

28:15Deon Nicholas:And you're going to see a similar race and probably an even bigger one. Like we've learned that the cap on coding is pretty infinite, like the amount of things we can build. But I think like the cap on human work is also infinite in terms of the things we can achieve, the things everyone wants to do, especially in post-COVID era. You have people who are now you've got work, you've got life, you know, founders are doubling as investors. You've got, you know what I mean? Like people are our work, their parents, like all these things going on. And we're going to see more of that. Right. In terms of this life.

28:47Deon Nicholas:And so I think this need for like, you know, I think the Uber moment for an AI executive assistant is arriving.

28:52Anne Dwane:And you have some early users. And so what are you learning about how they used it? Anything surprising?

29:00Deon Nicholas:Yeah. So first of all, some fun use cases that people are using it for. Like, it was really cool. We saw like a spike during tax season. Yeah, it was like a bunch of people like, hey, can you find all the like 1099s in my email or find all charitable receipts from the last year? Because I'm like, I'm good, you know. So we saw a bunch of really cool stuff. So what's fun about these use cases is people know what they need when they need it. So for me, for example, a few months ago, I actually was playing basketball and I tore my Achilles heel. I kid you not. It sucked. I'm still more or less recovering.

29:41Deon Nicholas:But I remember and I was like, this really sucked. And so I had a bunch of meetings that whole week. And I just asked Aespa to reschedule everything, reach out to people, say I'm so sorry, whatever. right like in that moment that that was what i needed as before um and then i was like you know like i had painkillers or whatever medication i was like i'm a forgetful guy as but can you just remind me like every like when i need to take my doses um and it would text me you know as what would technically be like all right did you take it how much when blah blah blah all right i'll remember for next time little things like that um i don't even know these aren't even like you you know, regulatoryly combined use statements, whatever.

30:17Deon Nicholas:That's prescribing. Yeah, like it's not, but yeah.

30:20Anne Dwane:So actually, that brings up an interesting question. There's a fine line between an agentic assistant and a companion. Do you think about that ever? Like personality or, you know, people have...

30:34Deon Nicholas:Yeah, yeah, yeah.

30:35Anne Dwane:Get engaged with...

30:38Deon Nicholas:Your AI personal companion. Damn. No, we do, actually. Like, as in, you do have to think about what kind of personality and, like, what kind of relationship and what kind of, like, relation to the world you want your AI to have. Yeah. So even when we were launching and thinking about, like, our website and our brand and all of those things, like, is Aespa your AI assistant? Is Aespa your AI companion? Is Aespa your AI agent? Like, how do you— Confidant. Confidant, right.

31:11Anne Dwane:Like, I mean, he really does know a lot about you. And the best, like, chief of staff or executive support people actually are quite, I don't know, they can think ahead and they're prescient and they know you so well. It's a really interesting thing.

31:24Deon Nicholas:It is a really interesting thing. And so we've erred on, again, like when we think about our pillars, right, like execution, context, and privacy. Like we've really been grounded in those things. And so the first one being, like, the ability to execute. And so Aespa has a personality you can actually customize. It's actually one of the memories that, you know, as you're talking as you can kind of like configure. Again, you have full control over. But at the same time, like it's really focused on the people who are trying to achieve great things and trying to get stuff done. And so, yeah, we see it as like this kind of personal executive assistance.

31:57Deon Nicholas:Why we kind of use that phrase. But, yeah, when it comes to just like AI these days, like, I mean, it's a crazy world.

32:04Anne Dwane:Yeah. Yeah. Well, it is interesting. And like reflection had that pie. Do you remember pie? I remember pie. And it would sometimes a couple days later ask you a question about a conversation you had. Right. Like, oh, how'd that turn out? And it was just like kind of amazing. Yeah. I get care.

32:23Deon Nicholas:You care about me. Yeah. Yeah. One of the surprising things in terms of user behavior that we found was there's this balance between proactivity in your AI and annoyance. Yeah, or creepy. Or creepy. Yeah, like proactive or creepy. Which one is it? You know? And so that's been an interesting product learning, I would say. For example, one of the features we have is when you get an important email, ESPA can decide if it's important enough to literally text you about it. And in the early days, we got it wrong. Yeah, I got a lot of texts. You remember that. You're like, I remember the text. And it's like, we got that wrong.

33:04Deon Nicholas:And we're like, well, it's being proactive. And then you realize, no, people don't actually want that level of proactivity. Now, it's so funny. I got texts from some of our users being like, this is the coolest thing ever, because we've dialed in on that nuance. And so, for example, it only texts you if, A, it's a very big decision that an assistant shouldn't be making, and it's not sure, it doesn't have enough context for the answer. And those are rare enough and important enough that when you get that text, you're like, oh, this actually deserves my attention. Yeah, it's working well. And that shift was super cool to see when we just nailed that.

33:39Deon Nicholas:And so I think there's a lot of nuance like that when building any AI agent.

33:42Anne Dwane:And how do you think about your interface is a lot about, to me, voice, text, and it's abstracting a lot of the inbox away. Correct. Do you feel like that's the future?

33:56Deon Nicholas:That's a great question. So I'm a bit biased in this sense because I never really wanted to build a workflow tool. Like I don't want to build another inbox or anything else precisely because of what you're saying there. I think like at the end of the day, when you take like a step up in kind of the Maslow's hierarchy of needs here, people don't really care about having another tool. They care about the job that they're trying to get done. And so when you if you were to hire an executive assistant, you don't hire someone who also has an app that they're like, well, you can only talk to me through this app.

34:34Deon Nicholas:Like, it's not how this works. It's another teammate who you Slack with or you WhatsApp or you email. And I think that is the paradigm we want here at Aespa is that you are focused. You're in your workflows. You're in your inbox. You're doing things. But then by having an AI assistant supporting you on the executive side, you're going to spend less time in that inbox and thinking about the mechanics of it. And you're going to be spending more time thinking about the high-level function or the job you're trying to get done.

35:05Anne Dwane:Yeah. I mean, that is exciting. And that seems like the TAM is humongous, right? Exactly, yeah. And will ESPA work with other ESPAs?

35:15Deon Nicholas:Yes. In fact, that is available literally today. It was so cool. Another one of those things where when you crack it, you're like, this is sick. Like, I just, like, the future is here. Yeah, you can actually do this today, right now. So, you know, you can CC your assistant at ESPA.ai or your ESPA assistant on a thread. and it will schedule meetings for you. So like literally, if you're meeting with a founder, meeting with a VC, hey, adding my AI assistant, I think it's pretty, you should be clear, it's an AI. But it just works, right? And it can have a full-on conversation. It has all your contacts.

35:50Deon Nicholas:It knows when you don't want to meet, like all of these things and can schedule. And when we think about it, maybe this is like an interesting topic for a broader conversation around defensibility and moats. I actually think the idea of connected context is actually in so many ways the future. Because imagine now, anytime you and I ever want to meet, let's add aespa to our text thread. Or let's add aespa to our email thread. My aespa talks to your aespa. Instantaneously, we now have a meeting. So that's calendar, shared calendar. What about shared collaboration? If you and Ben have both met with a founder maybe separately and you're trying to write the memo on this, well your aespa has your context and ben's aespa has his context imagine just being like hey aespa like write the first draft of this memo and it's going to have all of the nuances of what you all think and so when it comes to context and collaboration or documents or getting stuff done like aespa will be able to do that um this is a very bad example because this is not something i should have my assistant help me with my wife keeps reminding me um is but like let's say i'm trying to shop for a gift for my wife and, you know, she's been using aespa and maybe she's been shopping and maybe aespa has connections to perplexity or other AI for, for search and shopping.

37:06Deon Nicholas:And she, you know, she gives my aespa permission to, to her aespa. I like, I know maybe her wish list or whatever. And aespa can help me figure out what that is. And so that's again, shared context. Right. And so I actually do think like when you really extrapolate to the future, new paradigms for collaboration, calendaring, and context actually appear.

37:30Anne Dwane:Yeah.

37:31Deon Nicholas:And in so many ways, that becomes a weird network effect in the AI age, which I would not have expected for an AI company.

37:36Anne Dwane:Yeah. And it's interesting to think about your agent can be controlling your privacy, but letting you get things done. Correct. Exactly. And like, I don't have to share my calendar with everybody if my agent can one-off work with other agents.

37:49Deon Nicholas:Exactly. And it's almost like you're giving out tokens or, you know, like app authorizations for somebody else's ESPA or some other, you know, token. Again, like a shopping app may just want your contacts and you can like one-off be like, yeah, ESPA, you're allowed to give that to that shopping app. And then it will be able to do something for you.

38:05Anne Dwane:And what is state of the art in execution today? Like credit card purchases? What's happening with and without authorization? Yeah.

38:13Deon Nicholas:I mean, I think all of that is possible, if not going to be possible over the next few months, quite frankly. So today, we started with your kind of productivity suite. So your emails, calendars were also plugged into your granola notes and all of your meeting notes and things like that. We're building out integrations every single day. We're backed by Zapier and folks like that. But yeah, like the other day, someone was like, well, can I use ESPA to book a lift? Right? Because, you know, imagine doing that. I'm like, I look it up. I'm like, the API is there. Like, why can you not? Right? And so I think all of these things, whether you're trying to book a reservation for a hotel, whether you're trying to book a lift, book airport travel.

38:57Deon Nicholas:Again, these are all like booking use cases. But again, you're trying to get work done. Whatever that may be, you can use ESPA. We're going to enable over time things like payments and credit cards and stuff like that. I think there's a whole fascinating discussion on what is the payment economy for agents? Yes. Is it crypto? Who knows, right? But like, you know, the ability to at least give a, imagine like literally just an AESPA. You give your AESPA a small credit card and permissions over certain things. All of that's going to be possible. Browser use is also the other thing. There's a lot of companies working on that.

39:30Deon Nicholas:And we're also integrating that into AESPA as well. So being able to log into your, like one of your portals. Again, like for me, for this Achilles thing, like, you know, having Aespa log in and be able to see when all my physio is, put that on my calendar. Today, it can do that not through login, but I just sent it my, like I had literally downloaded a PDF and I was like, figure this out for me. But imagine like it just being able to have all your context.

39:54Anne Dwane:It's amazing. Okay. So where can people find Aespa and what's the call to action for them to try it?

40:01Deon Nicholas:Yeah. So Aespa is really a tool for everyone, which is really cool. Like if you're listening to this podcast, then you're probably trying to achieve amazing things. You're probably a founder. You're trying to raise funding, trying to launch your company or what have you. Right. And so I think everyone who is trying to achieve more should be using Aespa. So head over to www.aespa.ai and, yeah, sign up. Start a free trial. And, yeah, excited to see what you achieve.

40:31Anne Dwane:Great. And would you like to sing the tagline of Aespa?

40:37Deon Nicholas:I will not be karaoke-ing on this podcast. Okay, got it. But you still owe me. I want to know. You were supposed to tell me what your go-to karaoke song is.

40:44Anne Dwane:Yeah, I got to get back to you on that. Where's Claude? We're out. We're out. Time to go. Okay, well, thank you so much.

40:52Deon Nicholas:I think this was a lot of fun. Thank you so much for having me. Thank you.

40:58Deon Nicholas:Hey, this is Ben Kaznoka, co-founder of Village Global. Thanks so much for tuning in to the Village Global podcast, where we go deep on all of the biggest topics in tech. If you enjoyed this conversation, please subscribe to our YouTube channel. You can check us out on Spotify, Apple, wherever you get your podcasts. We'd love to see you for the next one.