Imagine what it would be like if executive assistants weren't just for C level people, but even your interns have an executive assistant.

Not only will your intern have it, but it'll outperform the C level person's real life executive assistant in the near future.

This is already able to do stuff with higher quality than team members.

I mean, this is mind blowing.

Nick, welcome to the show.

Thanks for having me, Aiden.

It's super excited to do this.

One of the really cool things about the the work that you do, Nick, is that you're constantly working with super innovative founders, super innovative companies, and then you're helping them innovate and now innovate with AI.

But you know, before we dig in, I'd love for you to just tell us quickly, like a little bit about your background, you know, how you got into the world that you're in today.

And like, what does your company do?

First of all, thanks for having me my entrepreneurial slash career journey.

I started off as a I'm a financial engineer actually by background.

So I was a high frequency trader on Wall Street for eight years.

So, you know, I've been thinking about automation.

I've been doing machine learning and AI for 20 years, basically from grad school to to that.

And then when I got out, I started getting into entrepreneurship and I got became really obsessed with, you know, what's the most efficient way to run a company.

And so my main company, I have a few, one is leverage and that's a operational efficiency training and consulting company.

And so before even this boom on the AI side, what I noticed was there's all these new ways of working and all these tools.

There's Slack, there's Asana, there's Fellow, there's Zapier.

How do you bring it all together and have a strategy that's cohesive and everyone's maximizing the use of these tools and, and they understand the purpose and best practices.

So I noticed a gap in knowledge with those things.

And there's a lot of companies that have been around for many, many, many years that have historically done really well and figured out, you know, some product or service that works and they've got customers, but they really didn't pay much attention to, you know, how, how do you keep up in this modern world and leverage all these fantastic tools that if you properly use it can fundamentally change how you operate?

So I became really obsessed with, you know, not just individual productivity, but how do, how do teams and organizations be productive?

And then I wrote a book called Come Up for Air which talks about my framework.

Yeah, I was a best seller too, right?

Like it did really well.

Oh, thank you.

It was not New York Times, but I was fortunate to get Wall Street Journal best seller.

So the book's done well.

And, you know, now, you know, we work with a lot of businesses.

AI is obviously a big part of it.

And, you know, we're starting to build AI agents and tools that support a lot of the philosophies and frameworks and best practices that go along with the book.

One of the biggest issues in business is e-mail.

You know, every business uses Outlook or Gmail pretty much.

And sometimes it's 10 or 20 hours a week, people spend an e-mail and they've just never been taught best practices.

So one of the key things that we teach at Leverage, as dry as this sounds, is how do you get to inbox zero?

And we have a framework called RAD.

You can read about it.

And it's one of the quickest wins that companies can do.

But taking it a step further, we're building an agent to just help automate that process.

And it takes into account a whole bunch of different factors.

And it's able to organize, categorize and draft emails way better than what you would see in some of the native solutions because we're just looking at so much more robust data to craft a good e-mail.

And then there's there's a whole bunch of other agents too that go along with the.

Book I'm super curious to dig in and so we're we're going to dig into to that stuff as well but I thought maybe a good place for us to start is just to jump straight into demo.

You have something really cool to show us and then we're going to talk about implications and then we're going to talk about agents and what kind of agents you help people deploy and and everything else.

This is clod.

So is this your go to AI?

Maybe like what do you use most of the time?

Is this your go to?

I'm using Clod a lot.

My typical philosophy with these things, Whichever 1 you want to use, you just have to make sure that you're using it right, you know?

I've got preferences with work management tools.

We promote Asana a lot.

If if someone wants to use Monday, they could be totally successful with Monday.

So there's personal preference.

As of today, you know, in June 2025, I'm liking Claude, but this stuff changes so fast.

Maybe next month, you know, Open AI comes out with some updates and now that one's better.

So at the end of the day, I'm not getting married to any of them.

I like Claude right now because on the desktop you can set up what's called MCP servers, and that was what I was going to demo for you right now if you want.

And I.

I really think MCP, it's something that I think most people aren't even aware of, stands for a model context protocol.

Essentially, it allows you to extend your internal knowledge and integrate it into one of these platforms, and then it could start answering questions, accessing data, but then also performing things with agents all through the one interface.

When you go to any of these platforms, you're just getting access to the general LLM, the general knowledge that it's got.

But once you start integrating it with your stuff, the Sky's the limit with how much more valuable these things can be.

Even starting to think, you know, what's the future of documenting knowledge?

And honestly, you know, like we promote Coda a lot or Notion's another popular one.

But I really think that, you know, a lot of the future of knowledge is going to be, I think you always want a knowledge base, But rather than going and reading, you know, 20 steps in some SOP in a knowledge base, you can make it a lot more dynamic and interactive and actionable and take all that knowledge, but stick it into something like clod or custom GPT.

And now the knowledge is not just stored there, but you can start asking it questions and it can start performing actions, right.

And so I think that it, I think that the future of knowledge capture is really going to be in these tools as well.

Let me just show you an example of what I mean with MCP.

My framework is called the CPR framework.

That's what my book is all about.

So you could see here, Claude's got these things natively integrated.

And then I've set up Asana separately.

There was a little bit of technical stuff and then I've set up my own agent, the CPR agent.

What CPR stands for is communicate plan resource.

That's the main framework in my book.

And so this agent has tools and it has access to tools and sub and its own sub agents.

So agents is a buzzword, basically an agent.

You code it some rules.

Think of it like a person, like you tell a person like, hey, you're responsible for these things, and then it has to go and figure out how to do it.

So at the top you have what's called an orchestrator agent, and in that case it's this executive assistant agent.

So that agent is programmed with certain contexts, certain rules, but then also the ability to access other agents.

So it's kind of like a manager, right?

And so this is the manager and it's going to oversee other agents.

So it might determine, oh, hey, we have to go to e-mail, so it might hit this one.

Or, oh, this is something related to tasks and projects.

We use a sauna, so maybe it has to go here, or hey, maybe it's related to a person.

Let's go and look them up in the CRM, right?

So by connecting and setting up this one thing here, it's extended access to knowledge as well as functionality and and all the different core tools that I use at the company.

Makes sense, right?

So I've got my knowledge base here.

I've got a whole bunch of stuff so.

It's interesting.

So through MCP, which is again, it's a protocol, so it's a way to talk to another system and it makes it easy for LLM to communicate with another system.

So through the MCP protocol, you've created a way to access all these agents that you've programmed.

So if someone, someone is working, I guess with you, your clients, they basically have access to this agent, yeah.

Right now I'm just doing this internally, but the idea is this is going to start, you know, we're going to have a suite of these agents and then we'll work with companies 'cause there's some customization, different people have different tools, different needs, etcetera.

So what the idea is, we'll go with a company, we'll set up some data pipeline stuff, some customization, but now they have the executive.

Imagine what it would be like if executive assistants, for example, weren't just for C level people, but every, everyone from the bottom up, even your interns have an executive assistant.

That's the future of these things.

I'm like, maybe, maybe they're going to have like 50 agents like an executive assistant that they all have access to and it's going to like 10X their productivity.

I mean this is mind blowing right?

Like this idea that every single person, even your interns, can have an executive assistant.

It can already happen and honestly it can do a better job than like your human executive.

Not only will your intern have it, but it'll outperform the C level person's, you know real life executive assistant in the near future like this is already able to do stuff, you know, with higher quality than team members.

The other thing which I think is pretty mind blowing, is that it's really cool that, you know, we've talked about agents before, but I love this idea of agent orchestration and how in this case you even have a manager agent that decides which one of the other agents should perform tasks.

Totally.

So CPRS at the top, you know, underneath CPR you have like a communications agent, a planning agent and a resources agent.

So the CPR one figures out what to do there.

And then within communication there's like sub, there's sub things even within communication, right?

Is it internal, external, personal?

There's a whole bunch.

There's like layers just like an org chart.

You've got AC level person, and then underneath that maybe a VP or manager, and then underneath that regular employees.

And you have this hierarchy.

You have the same idea with agents, right?

And just like a human, you have to check the work.

And if you notice that it produced a result you didn't like, you have to go back to what were the set of rules and instructions and what was not clear.

And it's just going to the future role of a lot of people at every level of any role is going to be doing QA, quality assurance on on the agents.

And when they find something that bugged out, either fixing it themself or reporting the bug.

And I think a big responsibility for every single person in every company is going to be QA and.

By the way, this is not all that different from being a manager of a team, right?

Part of The thing is you own the work that your team produces.

You raise the bar.

And what I like about it is it's almost when something doesn't go right, you don't blame, you know, you don't blame the agent, you blame you because you didn't do a good job of explaining it in a way or creating the rules in a way that it would produce the right outcome.

Totally.

In general, senior people are laggards for new technologies compared to lower level people like CEO or ACFO.

At like a traditional, you know, 8 figure type of company, they're usually the last people to use sauna or Coda or some of these tools.

For the most part, what's interesting right now with AI is it's actually the complete opposite.

And actually senior people more than twice as likely to adopt AI than lower level people.

Want to take a guess why?

Well, I guess like they're used to delegating.

I mean, this is the thing that they do.

They already think in terms of delegation and, and that's a skill in itself.

And I think that that's going to start becoming one of the most important skills for people to learn is not how to use these tools, but like just how to be a delegator and think about delegation.

It's definitely a new world we're entering.

So what you have here on the screen is it says tell me about Aiden.

So here, so if I hit this right now, so I'm showing you an example of MCP, but the agents that we've built behind the scenes here, they're utilized here in this MCP, but then they're also utilized with other triggers.

So like I call this a trigger, right?

Like I'm going to write a message here inside of Claude that's going to trigger my agent.

But you could trigger the agent from a type form.

It could trigger right now when I get a new e-mail that's triggering an automation to run through and you know, categorize and draft e-mail.

So like, I've automated my e-mail right now and it's producing draft emails way better than, you know, you're going to get out of like Gemini or Copilot.

And I can explain why that is, but part of the reason why is what you're going to see here.

So part of the reason why it's producing better results is because it's not just trying to write you an e-mail, making sure that it's formal and concise and stuff.

It's looking at a really, really robust set of data on you.

And it's taking into account emails that you've had with my team, past meetings, stuff from the CRM, all this stuff that you're not going to get in a lot of these other tools.

It's just going to help you kind of do some general cleanup, but it's not going to have the context.

And I I feel like really to solve this problem you need all the context.

This is why we keep saying context is king.

This is like the new phrase for this era.

So you you need the context.

So, you know, imagine a world where you can kind of almost do everything from this one portal, right?

Imagine all these agents behind the scenes that are running and getting triggered from a new e-mail, a new calendar event.

You know, every morning we have one that scans your calendar, sees all the upcoming meetings and gives you like an executive brief on, you know, each person, right?

That's something that only CE OS usually get.

I'm super curious to see what happens.

So here you know it's going to be smart enough to figure out what agent to even call.

So you notice there it knows I don't have to even tell it go to my CRM.

It just knows right now it's going to look up information on you.

So let's see what it finds.

All right, So then it finds all this information on you.

Right, so all this stuff was in your CRM.

It's ultimately in the CRM, but then we have another agent that's going out and looking at your LinkedIn, looking at news and and what's going on in news.

We use perplexity for that part.

Then it's looking at all of our emails together, all the emails with my team, every, all the data that we've got on the CRM.

And this is great, by the way.

Like this is literally my background.

It even has the This New Wave podcast, which we're we're recording right now and even talks to you about like, yeah, how we connected.

If I want to even go more granular, right?

So if I were to ask it right now, write Aidan an e-mail you know, about whatever.

It's got all this context, so it's not just going to write you a concise, formal e-mail.

It can reference details that might increase the quality of the e-mail and personalize it a lot more.

I mean this to me is crazy because this is literally what what you would typically do in working with someone who's managing your e-mail.

And like you said, it was a privilege previously only for C level folks at the company.

But this idea that it can go back, research all the history, look at the CRM, like do all these things and then create the most perfect thing that you want.

And I and I bet like if you wanted to, you could record a voice memo to say, hey, I want to send an Aiden a note to congratulate him on something.

Can you like go write something super comprehensive and it can take that, probably put it in your voice.

Look up all the relevant contacts.

I mean, this is super powerful stuff.

Totally look.

And then like what's the most recent emails, right?

And then like and then it also identifies patterns like are we getting closer and not closer?

Is there friction?

So it's got a whole bunch of stuff here, right?

Yeah, it's pointing out that we were trying to schedule this for a while.

Yeah, I know this is a you know, but here we are.

We find we did it.

Nick, this is really like pretty mind blowing stuff.

Thanks.

Yeah.

And like a lot of these agents, they either directly interact with each other or indirectly interact with each other, right?

So like we have one that we call Sniper.

And so when you e-mail, if you write an e-mail or the in the morning when we want to, if we want to do like the executive brief, you know, the sniper agent is creating a like 360 overview on you.

And then that overview is utilized for many other agents, right?

So it's going to be utilized for the morning briefing agent, but then it's also going to be utilized for the e-mail.

So by me writing in the in Claude in that MCP, it produced a summary card for you, then that summary card is then used in other things, right.

So if I really wanted to kind of go deeper, like I would have looked at this this morning, you know, and then like, who have you talked to?

Like what are we talking about?

Are we in what context are you a potential prospect?

Are we trying to get you to become a client of some sorts?

Or, you know, is there some type of partnership, like what's the context here?

Because that matters for what I need to know going into the call or how I'm going to e-mail you.

It's great.

Like you said, it outperforms humans because it's almost like the amount of work that it does.

And I mean, that card looks, looks amazing.

So maybe you would have even your EA create something like that, but they would do it if you were just about to meet who knows, the president of a country, right?

Like it would have to be a big deal to spend that much time.

But now you can have that for everybody.

And my team has access to it, right?

So if you know you need to do a call with Jessica, she sees she sees a contact card, right?

And if she needs to write you an e-mail, it has context of that it has context of.

And again, there's there's privacy that you need to take into account too.

So it's only going to help Jessica write an e-mail with data that she has access to.

So if I'm sharing all of my emails, syncing it with HubSpot, like she can see it anyway.

So that means the agent for her can see it.

This is amazing.

And you know, just the concept like this kind of stuff was impossible before and and now you're showing that everybody can have access to it.

What's crazy is like given this idea that you can have your e-mail auto parse in advance, that drafts can be based on like all the other contacts that exist in your company.

I mean, this kind of world, I mean, it sounds very futuristic, but it's here like you're doing this stuff like you're doing this with clients.

The, the devil's in the details with a lot of these things, right?

Because anyone, you could spend an hour and set up something to automate or, you know, draft an e-mail, but every micro optimization or edge case, right?

You know, some emails contain like weird links in it, right?

Or maybe there's an attachment in the e-mail.

Maybe there's like a link to a Google doc or a link to some website.

And so you have to be able to, if there's a loom video, transcribe and understand the context.

If there's a document under read the document, if there's a website.

So you have to do all that.

Then you also have to be mindful, you know what model is best used for different use cases.

So one of the little nuances too is like we have agents that whose job it is to select what the best model is for the other agents to do.

So it's dynamic.

I believe that people are going to start really realizing that prompts are a form of intellectual property.

And just like you have assets and IP, it's not just going to be trademarks and patents, it's going to be your work flows and your prompts and things like that.

So there's best practices that we're kind of figuring out and defining.

For example, all of our prompts are stored in a database and then that database is ultimately what's referenced within our agents.

And what that does is we can see a history of every change.

We can do machine learning to optimize our prompts.

Like you don't want something I I speak about when I do a keynote is?

You, you want to challenge status quo with your SO PS and processes.

And what I absolutely hate is when you ask someone, why do you do something the way it's done?

And they just, oh, it's the way that we've always done it.

You want to, you know, be always rethinking.

Is it still smart to do this?

Often times it's just no, no one's ever thought to look at something.

And it's the same with prompts.

If you come up with a prompt today, you know, if you don't re evaluate and tweak and optimize, that's that's foolish.

And so you want to have a prompt optimization strategy.

And if you start putting it into databases, you could see history better kind of AB test and see the results.

Also, if you want to protect the IP and not have everyone be able to tweak or touch, you know, it does give you that extra layer of protection, but it it really makes it a lot more of an asset.

Then you, you invest as a company in a really good prompt for, you know, how to categorize an e-mail or how to build this contact card and you're not having other people reinvent the wheel or potentially start diverging and have having different prompts.

You really invest one time and it's stored somewhere.

You know, I think about just even situations like how this stuff can get really real time.

So, you know, you have a conversation, say that a competitor comes up on a on a sales call and immediately you have, you know, a request, go to product marketing or your product marketing agent and like that can kick off, you know, workflow and then update like a battle card with the latest information about.

I mean, this stuff is going to be fast.

It's going to be real time.

All the workflows that make sense are just going to happen with these agents, and they're going to be very specific and be really good at the one outcome I would say that they're trying to produce.

You know, so again, the devil's in the details like you want to be mindful of, you know, what you're feeding into these things.

So sometimes you want to be thinking about kind of like pre layers to things.

So you're not just necessarily uploading the entire transcript of a call, but you have like a pre process that you know, like a a fellow or one of those types of tools will understand it's a sales call.

So this is exactly what we're looking for in a sales call.

And it's not just a general summary, right?

Because if you upload a transcript to one of these tools and say summarize it, you're just going to get a general summary.

But you might have a certain certain style and certain key elements that you want to extract from sales calls.

So you have your own framework for a sales call.

It could even determine was this a sales call, a biz dev call, an internal call.

Maybe there's different things you want to strip out, but you, you need to define that.

And then you, you, you get the summary, but it's not the general summary.

It's your unique custom summary.

And then that's fed back into the CRM or the agent so that you're not kind of just drowning it in gigabytes and gigabytes of crap, but you're helping it to be more focused on just important data to make the best recommendation or produce the best result.

And just like you would with a human right, so you want to give them like focus information.

And even if you had like, this super powerful EA or chief of staff, if they were going to brief you on a meeting, they would tell you what you needed to know, right?

They wouldn't tell you all the things that you already knew and waste your time.

Yeah, there's a lot of stuff happened, but here's what you need to pay attention to.

And you're right.

Like the intellectual property is in these prompts, and it's really important to iterate on them.

I think when the reasoning models first came out, it kind of changed the way that you needed to prompt.

And they're constantly changing.

Nick, I was just going to ask you very quickly about what goes on in the background with these agents.

Like how does one go about building them?

There's a lot of different platforms.

It's a moving target.

Like right now we're doing a lot of stuff in N8 N which is like a Zapier competitor.

It's quite technical.

So the average company, it's probably not going to be a fit to try to go and learn that.

You probably want to hire, hire someone for that.

You know, there's a bunch of different platforms where you could do it.

I think, you know, if you want to long term, like once we start giving this or selling this to clients and integrating this, the strategy is going to be different.

You know, it's going to have to be custom code and like a stand alone web application that we build for for more, you know, prime time enterprise level.

But for prototyping or purely internal, we're using NADN.

You got crew AI, there's a whole bunch of them.

I know we're we're running up against time, so this was a really cool demo.

If people want to get in touch with you, what's the best way to do that?

Well, on LinkedIn, you can look me up and I'm sure in the show notes you can put it, but I'm Nick Sonnenberg or Nicholas Sonnenberg.

On LinkedIn.

I think it's Nick Sonnenberg.

My book is Come Up for Air.

So come up for air.com if you want to check the book out and the training and consulting company is get leverage.com.

So you should be able to find me from any of those.

Oh, and I have a podcast, the optimizepodcast.com.

So if you're interested in inefficiency, take a look at some of those.

This is amazing Nick, Thanks so much for doing this.

Glad we finally had a chance to to make it happen and till next time.

Thanks for having me.

And that's it for today.

Thank you so much for tuning into this episode of This New Way.

If you like the content, be sure to rate, review and subscribe so you can get notified when we post that next episode.

See you next time.