Kyle Richless on How AI Turns Three Days of Work Into Seconds

00:00:00:00 – 00:00:32:29
Unknown
Hello, this is Julian Castelli. I’m the host of the Growth Elevated Leadership podcast, where each week we talk with inspirational entrepreneurs and leaders in the tech industry. Today’s podcast is part of our AI Sherpa track, where we talk with leaders, builders, and experts who are building and innovating with AI. This episode is brought to you by Growth Elevated Growth elevated as a community of tech founders, CEOs, and CXOs who are committed to working together to share best practices and learnings in an effort to help all of us become better leaders.

00:00:33:01 – 00:00:54:05
Unknown
We do this through educational programs like this podcast as well as on our blog, and of course, our annual Ski and Tech Summit in beautiful Park City, Utah. So if you enjoy skiing and talking with our tech leaders, check us out at Growth elevator.com. And please subscribe to this podcast wherever you listen to your podcasts. My guest today is Kyle Richards.

00:00:54:08 – 00:01:21:02
Unknown
Kyle is a lawyer, startup operator, co founder, venture capitalists, emerging fund manager and now a second time funder founder, according to his LinkedIn. I’m going to ask him questions about this. This is path of all those things today. Kyle is joining us to talk about his latest venture, zygote AI. Zygote calls themselves the outsourced AI team for owner led and mid-market companies.

00:01:21:04 – 00:01:42:12
Unknown
They map the workflow, build the systems, and keep improving after launch. With AI. Kyle recently joined us for an in-person workshop at Growth Elevated, where he walked us through some amazing use cases, and I’m excited to bring him on to the podcast to go through those today so we can we can share them with our broader, broader, broader audience.

00:01:42:21 – 00:02:05:15
Unknown
Kyle, welcome to the podcast. Thank you so much. It’s an honor to be here. It’s good to see you again. I love the AI Sherpa framing. I think I’m going to steal that and borrow that from you. And secondly, when you just chronicle everything I’ve done professionally, I’m feeling particularly old now that I’ve just. I’ve just heard every, every stop across my, professional journey.

00:02:05:15 – 00:02:24:22
Unknown
But but it’s good to be here. I love talking, I love talking about. And the work we do with our clients. So excited for the conversation. Yeah, well. Well, you know, you’ve been an entrepreneur, you’ve been a lawyer, you’ve been a professional, you’ve been venture all the things that we talk about around the tech tech ecosystem. So you’re a great person to talk to.

00:02:24:22 – 00:02:53:21
Unknown
I think, you know, our audience is, consisting of tech CEOs and CXOs. And so I think you get to relate with a lot of them. They’re all in various stages of adopting AI. I think there’s a lot of AI FOMO out there. I think there’s a lot of AI confusion. And, as you recall, when we we did our workshop, we’ve put together a growth elevated, a three stage framework to try to create some structured process, or at least a, playbook for implementing AI.

00:02:53:23 – 00:03:22:22
Unknown
Stage one is, enabling your knowledge workers with AI. Process and coaching and using tools. And it’s basically the basics around setting up an air charter, choosing your tools. And it has a bunch of, learning frameworks, that, that are, is set up to have your teammates using AI, sharing it with each other, learning from each other, celebrating it as a company, and making it part of your culture.

00:03:22:24 – 00:03:46:05
Unknown
Stage two, which we talked about when you joined us here in Park City, was choosing specific initiatives to underwrite as a company investment, to try to improve a process better than has been done historically by using AI and upgrading a workflow. And you shared some great use cases in that and looking forward to to going through that today.

00:03:46:07 – 00:04:07:13
Unknown
And then stage three is actually putting AI into your product, particularly if you’re a software company. But there are there are real world companies doing it as well. So that’s the framework today. We’re going to focus on that, that middle section. And that’s what you do it at zygote. Right? You’re the expert that people, call to bring in when they decide they want to automate something with AI.

00:04:07:15 – 00:04:34:12
Unknown
How did you decide to start zygote? Well, very quickly, before I answer that question, I, I would double click on what you said. You said there’s a bunch of FOMO and confusion around AI that, that might be understated. I think those two elements are true. I’ll add a third and fourth dimension to FOMO confusion. I think there is also a lot of insecurity about AI.

00:04:34:15 – 00:04:55:14
Unknown
My competitors using AI is they are going to replace my job. And then there’s a fourth bucket. What I wish I call political antibodies to AI, which is the opposite of FOMO, which is oh my gosh, who is Kyle? What is this engineering team and what might that mean for me? Good, bad or ugly? Yeah, it’s a threat for sure.

00:04:55:16 – 00:05:19:09
Unknown
So, to to to add that here and then as far as the work at Zynga and why it shows this path, you know, as you as you had mentioned, when you outline my, professional journey, I was a venture investor prior. Up until 2022, in the opening AI moment. And two things really dawned on me when open AI became the verb and went mainstream.

00:05:19:12 – 00:05:44:23
Unknown
The first is some of the best entrepreneurs were telling me that they might not need my capital or anyone’s capital to go 0 to 1, because I don’t have to spend on an expensive tech team, because now coding terminals have dramatically reduced the cost to get to an MVP or to get a product to market. And that is a tough pill to swallow if you’re in the business of selling capital, which is what venture capitalists do.

00:05:45:05 – 00:06:13:07
Unknown
But the second thing that dawned on me is the first version of OpenAI came out one. By the time they released their second model, Anthropic and Dario broke away, he started anthropic. And by the time they had released their first model, you already had a sense that the pace of acceleration was moving rapidly at a minimum and might be moving faster than any technology that I have seen in my lifetime, at almost 40 years old.

00:06:13:09 – 00:06:46:13
Unknown
So if that’s the case, then the technology is going to be that fast while you have disruption, which equals opportunity. But I arrived at this conclusion that there’s a bet I wanted to make, and the bet was the pace of this technology supercycle is going to far outpace businesses and individuals ability to adopt the technology. And so I thought a tremendous business opportunity would be at the implementation layer, which is how do I get smart about AI?

00:06:46:13 – 00:07:05:28
Unknown
What models do I use? How do I deploy agents? But like how the how do I even think about this in the context of my company and my customers and my products and my competitive landscape? And who is there to bridge the gap from an implementation perspective, because it didn’t seem like it was going to be McKinsey or Bain or BCG or these legacy players?

00:07:06:15 – 00:07:30:07
Unknown
Did are not native to this new technology. And so absolutely a huge opportunity that was that was kind of the impetus and taking it and run with it. And it is it’s a battle that is paying off every day and is proven to be even more true than I kind of realized at the time. Awesome. I’m going to I’m going to do that so they know to break this.

00:07:30:09 – 00:07:57:06
Unknown
I’m getting a bunch of feedback from your mic. So so we can we can splices together. But are you hearing that when you’re talking. Are you hearing you’re hearing a cracking when I talk. No I don’t hear anything so weird. Keep talking. Can you hear me now? Any better? Keep going. What about now? It’s like it’s a it’s a it’s like a crackling behind the.

00:07:57:09 – 00:08:21:07
Unknown
Yeah. I don’t just read a sentence of your of your transcript just so we can hear, hear a flowing sentence. AI doesn’t create value where people are busy or less. So it’s more about where work is stuck or piles up waiting for approval of information. Replies, reviews, finding things, moving data. Businesses don’t. Okay, that’s better, that’s better.

00:08:21:09 – 00:08:33:14
Unknown
Okay, let’s let’s let’s just try, try again. I’ll just so that it’s in there and the transcript will will work on that says is to 827.

00:08:33:17 – 00:09:01:18
Unknown
Cool. Well thank you Kyle. What is one of the biggest misconceptions that business leaders have about AI now that you’ve been working with them for a while? Yeah, the biggest thing that jumps out is the following. We’ll talk to business owners and they’ll tell us, I know I need to get moving on. I oftentimes the impetus there, the catalyst there is I have a feeling my competitors, other players in my ecosystem are getting moving on AI.

00:09:01:20 – 00:09:25:05
Unknown
So let me quickly buy AI tools or upgrade my legacy software stack with its newfound AI add ons and not all quell my fears and maybe that will make me up to par with others that are savvy, are front footed when it comes to AI. The issue is if you have an adopted AI, you don’t have an AI problem.

00:09:25:07 – 00:09:58:06
Unknown
But if you haven’t adopted AI, what you haven’t done, which is what your competitors might have done, is attacked what I call decision latency. Okay, so so so what is decision latency means? Decision latency is the time between when you want to know a particular piece of information or data about your company, and then when you know it, number one and then number two, once you know it, how quickly can you then act on that information to make an action, make a decision, or do something that’s in service of your business.

00:09:58:06 – 00:10:25:07
Unknown
And so, you know, A.I. isn’t about just finding tools. It’s not just about replacing labor, but AI at its best. As of July 7th, 2026, it’s about eliminating waiting, and it’s about escalating cycle time so more people can get more work done quickly. And work that shouldn’t get done by humans gets offloaded to agents, and humans do only the work that they need to do.

00:10:25:09 – 00:10:52:03
Unknown
When that happens, actions get taken not in three days, across six meetings and 14 emails, but in an automated fashion that happens in seconds. And so, yeah, that’s kind of the biggest misconception. It’s not about we don’t need to adopt AI for the sake of AI. We want to adopt AI to move more quickly, to make people more productive, and to give humans time back to do things that only humans can do.

00:10:52:06 – 00:11:13:27
Unknown
Now that makes a ton of sense. What, what? You know, how many, how many projects have you been doing now? Zig zag over roughly. We have two dozen that are alive right now. Okay. So quite a few. And in these projects, almost every one of them, I think, except one is on a monthly retainer basis.

00:11:13:29 – 00:11:35:08
Unknown
And so we operate in short sprints. We will build a module we call modules, and 4 or 5 weeks. That’s the full way to production, live deploy, etc.. And then move on to the next thing. And the reason why we work on a retainer basis, and frankly, the reason why we keep our clients is because we articulate that line.

00:11:35:08 – 00:11:56:23
Unknown
What is the ROI going to be from this module or this project across this month? And then once it’s deployed, if the ROI is there, then, well, great. That’s awesome. What you just automated or made more productive, more efficient in finance. Now let’s go to operations. Right. So we’ve automated or done something operations. If the ROI is there time back more productivity etc..

00:11:56:25 – 00:12:21:21
Unknown
Let’s go to sales. And so we we will move kind of department by department or team by team or even person by person across companies and 4 to 5 week sprints. And again we do that across about two dozen companies currently. No, it’s well after after working with over dozens of companies. What surprised you most after, you know, completing these AI implementations?

00:12:21:24 – 00:12:47:16
Unknown
Yeah. What’s really interesting is, you know, we’ll get the question of, hey, if this doesn’t work or if this breaks or the ROI doesn’t come to fruition, you know what the hell happened? And it must be right. It must be because of the model soliciting. Right? But when things don’t go as planned, or the ROI doesn’t show up in a short time frame, or have what we call have rapid time to value, it’s almost never about the model.

00:12:47:16 – 00:13:10:11
Unknown
It’s almost never about hallucinated elucidation. And really, it’s about two things. One, you need leadership to initiate the change management exercise because you need people to buy in. You need the utilization. And then once agents have been deployed or genetic frameworks have been deployed, or AI is embedded into your organization, people have to change behaviors, and habits have to be broken.

00:13:10:11 – 00:13:28:08
Unknown
And so I’ll give you a non-tech example and then I’ll give you an eye example. If you have a broken down Ford Bronco that you just can’t stand because you can’t drive to Park City without, you know, having to pull over the side of the road. And then I replace it with a brand new 2026 Tesla model Y.

00:13:28:08 – 00:13:52:13
Unknown
The drive itself is like, oh my gosh, this is what an improvement. But you’re essentially using the the car has to drive the car the way you always have. Right. And so there’s a behavioral change in habits that need to change for things to be successful. And so what’s a practical I example. Oftentimes we build a use case which is about customer service because there’s a media narrative that legal and customer service and software development are some of the low hanging fruit.

00:13:52:13 – 00:14:17:00
Unknown
And there’s some truth to that and it’s direction. Correct. And so what’s what is the customer, service example. You know, let’s say you have Zendesk, you got a customer support ticket. The customer service rep typically will look at the question and then manually go look at the support docs or the help docs or the ethics then has in his head.

00:14:17:00 – 00:14:40:28
Unknown
Or I sent a previous email. There was a similar question last Tuesday. Let me go pull that up and and send it. And maybe that entire exercise takes 45 minutes. But that’s the ingrained habit. What I can do is read the inbound ticket or API access, or MCP access to z index automatically send an agent to find the answer in your help docs, then auto populate a draft.

00:14:41:03 – 00:15:04:22
Unknown
Have that draft sitting in your outbox such that all the rep has to do is go to his apex, do a quick scan, and press one button. So rather than 45 minutes that took 10s, but it took the behavioral change to do it differently. And so the buy in and the leadership that inspires new behaviors is the precursor to actually being successful with AI.

00:15:04:22 – 00:15:20:04
Unknown
And I guess, you know, I’m giving you a long winded answer. But but the last thing I’ll say, Julian, because I think it’s just super important. You asked me two years ago I would have told you this is about a third leadership and change management buy in and two thirds of technology challenge. You asked me about nine months ago.

00:15:20:04 – 00:15:45:00
Unknown
I’d say maybe it’s half and half. Leadership change manager technology challenge. You ask me today, I’ll tell you it’s two thirds leadership buying and change management and one third a technology challenge. And that is real and true. And my reactions to doing this day in day out for multiple years across, you know, many, many clients. Well, that makes a ton of sense because it really is change management.

00:15:45:00 – 00:16:00:26
Unknown
And that that that three stage framework I talked to you about was all about you know, hitting that and making sure if you don’t have CEO buy in, if you don’t have literacy, if you don’t have a culture where you’re going to do things differently and a mandate to do things differently, it’s not going to work. You’re right.

00:16:00:26 – 00:16:27:04
Unknown
Technology itself is not going to solve the problem. It’s just a different way of doing doing work. Kyle, I’m going to suggest that we start start over or just do things because the cracking is really at the point where I don’t think it’s going to be a professional recording. So I’m going to stop and then. Yeah, that’s so important, Kyle, about it being a change management exercise for the organization.

00:16:27:07 – 00:16:42:27
Unknown
You know, how about ROI? Are you seeing ROI? I know that there’s been some articles in the journal and other places that people are saying, hey, you know, people are spending all the money, but they’re not getting the ROI. What’s your take on that? I’m sure you get that question before people sign up with you. Oh, absolutely.

00:16:43:10 – 00:17:04:03
Unknown
It’s the the the the Crucible question with respect to AI, and I’ll, I’ll answer it, in two parts. The first is, are there organizations or even most organizations that are not generating the ROI that they want to or that they underwrote when they started the initiative? Absolutely. That’s the case. Whether the Wall Street Journal is right or not.

00:17:04:06 – 00:17:25:24
Unknown
It is the case that most of these initiatives are not generate the ROI with the payback period. A huge piece of that is organizations trying to bite off more than they can chew. What you don’t want to do, from an ROI perspective, is pay a large sum of money to get a very extensive AI roadmap that is transforming the entire company in a short period of time.

00:17:25:26 – 00:17:47:25
Unknown
That just does not work. That’s too much, too soon, and that does not set a foundation of change management and leadership. Thing number one. Thing number two, with respect to ROI, is, oftentimes people are adopting AI tools without even articulating what is the ROI in the payback period, even going to going to be. And how do I even dictate whether this was successful?

00:17:47:28 – 00:18:12:26
Unknown
And so I have a very strong view, which is if you cannot articulate the ROI of a project in two sentences, that’s not a project. You prioritize. So so have two. It does have to be one. So, so a one sentence ROI is here is better clarity on lead scoring with respect to my best prospects, so that my best sellers are only dealing with my best projects, and then their quota attainment goes from 80% to 120%.

00:18:12:28 – 00:18:42:18
Unknown
They literally are generating more money that okay, that’s a direct derivative one sentence articulation of ROI. So what’s a $0.02 derivative? ROI. We have a, company in the payment space. They have a team that is scanning huge amounts of payment documents. That team we’ve been able to automate and deploy agents across that division. The team has been able to shrink from 7 to 2, but we had a plan to repurpose those five people.

00:18:42:18 – 00:19:03:02
Unknown
Those that are now sales engineers and the sales engineers are increasing the conversion rate for the sellers. That’s a second derivative ROI. So that’s the second sentence. But if it’s either a first derivative or a second derivative, I prioritize the project. If it is this elongated and I think this might help someone, it might help someone or someone, and this could be cool, or this would be novel.

00:19:03:04 – 00:19:34:02
Unknown
That is not ROI that I quantify. And that’s not a project I will prioritize. Got it. Okay, well, listen, I, I know when you oh my goodness, I’m sorry. We’ve we’ve got a problem. Now I’m hearing a double echo. Oh my gosh. You were you were crackling and now I’ve got a double echo. Shoot testing one, two three.

00:19:35:13 – 00:19:50:01
Unknown
It’s not going to work. Maybe I’ll come. Let me jump. So call based on all of these examples, if you had one piece of advice for CEOs today, well, what do you think would be.

00:19:50:03 – 00:20:16:18
Unknown
The biggest thing I see from CEOs is they want to vet the vendor or the partner, and we show them what we’ve got. We show them case studies, we show them your testimonials, and then the mistake is, okay, I’ll go talk to my CTO and then the CTO is okay, this is great. Show me functionality I don’t know, or tools that my team doesn’t use or agents that are deployed differently from how my team might already be operating.

00:20:16:21 – 00:20:40:07
Unknown
But if there’s a single pocket in an organization that’s probably leveraging these tools might already be well versed in cloud code. It is the tech team. And so when you’re giving superpowers to the team that already has a taste of the superpowers, there is less alpha to gain. And or, you know, even worse, you know, the tech team I had mentioned political antibodies.

00:20:40:15 – 00:21:00:12
Unknown
Right. Sometimes I wondered about that. Yeah, I know, sometimes the tech team is like, well, if there’s an AI partner that’s moving three or 4 or 5 times as fast and is a fraction of the cost of, you have to use, maybe we don’t want to talk to this partner. And so rather than thinking this is a technical solutions, let me route it to my technical expert.

00:21:00:15 – 00:21:29:08
Unknown
What a CEO should do to avoid this mistake is let’s have a conversation across my executive leadership team at or at the board level and then identify there where are slow processes that are slowing things down, where is their operational drag? Where are humans getting stuck? And where could things move more quickly and very frequently? That leads you to finance or ops or recruiting or sales somewhere that is pretty distant from the tech team.

00:21:29:11 – 00:21:47:16
Unknown
They don’t have the political antibodies, they haven’t experimented with this stuff. And there’s much more alpha and ROI to gain. And frankly, buying happens faster. And so my my big piece of advice for CEOs is this is not about just kicking it to your CTO. And this is not just about your technical team getting smart about it.

00:21:47:19 – 00:22:08:13
Unknown
Every job is not going away, but every job is going to change. And that’s commercial. That’s operational, that’s financial, that’s legal, etc.. And so you have to think big picture. And again the lens is how can I make things move more quickly, more productive and give time back? Not necessarily. How can I sharpen the pencils of my tech team?

00:22:08:15 – 00:22:24:27
Unknown
Yeah, that makes a lot of sense. All right. Great. Well, listen, Carl, you do such a great job of sharing those use cases. I’m going to pull up the presentation that, you shared with us and ask you to, walk us through these, these use cases. And I’ll probably ask you some questions as we go through it.

00:22:24:29 – 00:22:31:06
Unknown
Awesome. Let’s do it.

00:22:31:08 – 00:23:25:18
Unknown
All right. Can you see the screen? Yes, sir. So recruiting and hiring. So. So people are the lifeblood of every company. You know, this is no surprise. Hiring is hard. There’s this old adage, hire slow, fire fast. No one actually does this. And it’s it’s time intensive. And pre I, for most positions, you have a recruiting funnel similar to how you have a sales funnel and someone in HR or talent or whatever you call your recruiting team or division is making a job description that hasn’t been optimized with I collecting tons and tons of resumes, whether a platform like in deed or through LinkedIn is looking manually at resumes and trying to

00:23:25:18 – 00:23:49:10
Unknown
decipher those which jump out as a fit for the company, a fit against the job description, and then the best resumé fits or the best fits on paper. Or oftentimes we know nepotism exists. The, you know, the the, CEO’s neighbor’s son is part of 15 or 30 or 50 screening interviews just to get to the middle of the funnel of your candidate funnel.

00:23:49:10 – 00:24:21:10
Unknown
So, so a ton of time and energy and inefficiency is wasted just trying to populate your candidate funnel with candidates that you could plausibly hire. So how does I hope to reinvent and transform the process? Well, number one, uploading your JD to the anthropic models, which are very good at this sort of thing, can help sharpen your job description and sell it more when you have a better job description, that job description, let’s say you’re working with with LinkedIn because that’s the predominant hiring platform.

00:24:22:11 – 00:24:53:11
Unknown
Your JD’s on LinkedIn. When resumes come in, an agent can grab the resume, compare them to the job description, and compare them to a rubric. And traffic is very good at this. And stack rank. The candidates could be 1 to 100, 1 to 50 and then score them. But because I has the ability to use logic, you can also give you a 1 or 2 sense description of why the candidate was scored when it what it was for the candidates that surpass whatever your threshold, they’re the top ten out of the 2000.

00:24:53:11 – 00:25:32:01
Unknown
They scored 87 or above. Then an agent can send an automated and personalized message on LinkedIn to the candidate, almost instantaneously saying, you know, hey Julian, thanks so much for applying for this technical PM role at Zeigler Labs. I’m particularly interested in ECS experience or why expertise, which makes me feel there could be a potentially very strong fit, and that makes for an unbelievable candidate experience, because as soon as you applied, rather than just waiting and wondering which is the typical can experience, you immediately got a personalized message that feels as if a human wrote it specifically for you.

00:25:32:03 – 00:25:54:29
Unknown
So now, with a great content experience, you have earned the right to have that candidate invest more time because they just applied and they quickly heard back in a personalized way. So now you’ve earned that ability to get more time from the candidate. And so then the next thing that you can do is rather than jump right to a screening call, an agent then sends a second automated message.

00:25:54:29 – 00:26:27:13
Unknown
This is a DM right here. And LinkedIn can be and indeed can be via email. And the message asks the candidate to record a five minute or less loom. Video looms a free video recording platforms out loud and they submit the loom video, answering 5 or 6 questions that you conveyed to the candidate. When the loom video is finished, a webhook triggers that alerts the agent, grabs the transcript from the loom, brings it back to Claude, and then scores that transcript against a rubric.

00:26:27:16 – 00:27:04:10
Unknown
This is a beautiful thing for many, many reasons. One, you get objective evaluations of candidates. There’s no bias when an AI is scoring a video transcript. The ad doesn’t know black, brown, female, male, European like it doesn’t know your age. So it’s unbelievably objective and it can happen rapidly. And so now you’re at a place where rather than having a probably a junior recruiter making job descriptions, scanning hundreds or thousands of resumes and then doing limit lists, what I call a top a funnel screening calls.

00:27:04:13 – 00:27:26:09
Unknown
Now they’re just watching five minute loom videos of the very best candidates of thousands. And then by the time you actually want to engage, you have a wealth of information about that candidate. And so earlier when I was talking about giving time back as a core catalyst for AI, this is a beautiful use case, and it just tremendously expedites the process.

00:27:26:13 – 00:27:45:05
Unknown
It gives time back, but it also gives you better candidates because again, I can score better than a human can. An AI is not objective. AI is more objective and less subjective than humans can be. And we know humans in interviews typically make snap judgments in the first 15 seconds of conversations. Right. So a beautiful you know, that’s a perfect one.

00:27:45:05 – 00:28:08:00
Unknown
The recruiting use cases is, is huge. That’s such a, you know, digestion of information, lots of data. You know, you have a structured job description. So, so, you know, a lot of the, the, the aiming pieces have been built. It’s just been processing, manually with humans. And AI is perfect for that. So great example there.

00:28:08:00 – 00:28:24:18
Unknown
I love that one. Let’s move on to sales and revenue. Yeah, totally. Sales and revenue is another use case that’s near and dear to my heart. It’s similar in the sense that recruiting is a funnel many apply for. You will be hired. You know sales similarly is is a massive funnel. You want a ton of qualified leads in the top of the funnel.

00:28:24:18 – 00:28:45:03
Unknown
Hopefully your conversion strong and a few will actually amount to paying customers. But any great seller would tell you that the actual meeting or demo or presentation, no matter how good of a listener you are, how high IQ you are, or how charismatic you are, it’s only part of the battle. And sales is a grind and it’s preparation.

00:28:45:09 – 00:29:07:29
Unknown
And yes, it’s the demo meeting, but then it’s follow through and follow up and wrangling cats, and there’s just a lot of effort and volume that goes into sales. And so I applies to sales in the sense that it can better prepare you for a meeting in a fraction of the time, and then it can do all the follow up and a hyper personalized and expedited fashion.

00:29:07:29 – 00:29:27:26
Unknown
And so me as a salesperson, you know, the first 11 or 12 years of my career, I was leading sales teams. And when I would have a sales call or meeting, I would always reserve at least 45 minutes so I could be properly up to speed and abreast of who is Julian? What is your background? What commonalities do we have?

00:29:27:28 – 00:29:49:28
Unknown
What about your background might make you find my value to be particularly interesting or adept for what you’re doing. And I would compile a research one pager, and I did this for every single call, and I did it for every single call so that the first five minutes of that sales call I won the candidates trust, they could feel Intel.

00:29:49:28 – 00:30:10:26
Unknown
I was hyper prepared, and I’m investing the time in them so that they can reciprocate. Now, it can happen in a matter of about 30s. So what happens is the best practice is, you know, let’s say a meeting gets put on my calendar and let’s say I’m Microsoft. 365 shop APIs hooked up to Microsoft Graph, which is the Microsoft API.

00:30:10:28 – 00:30:34:11
Unknown
As soon as Julian schedules on my outlook, that pings an agent in parallel and multiple API calls happen, one goes to LinkedIn. What’s Julian’s background now? You’ve been a CEO, you’ve been a board member. You run a community. What’s that? I’ve been a CEO. I’ve been part of. Okay, that’s really interesting. Growth elevated. Is there information about an in PitchBook or Crunchbase?

00:30:34:11 – 00:31:01:26
Unknown
If you were a public company, agents go to SEC filings and you can have agents rapidly in a matter of like literally seconds, compile tons and tons of information. That’s step one. Step two is if you upload 5 or 6 example one page prep doc as your template, then as the agents are grabbing all this information, then they can synthesize and format it into the exact template that you like because they had previous examples.

00:31:01:28 – 00:31:20:09
Unknown
And so now I’m putting a call that I have just because it’s on my calendar. I have a one pager in the format that I want, with the content I want, with the length that I want, with the opening. I want emailed in slack to me five minutes before the meeting, and now my preparation, rather than taking 45 or 50 minutes, is just a five minute review.

00:31:20:12 – 00:31:37:27
Unknown
Okay, so then what’s the two sentence translation to ROI? Maybe I’m better in the meeting, but also I just got that 45 or 50 minutes back and I can schedule another sales call during that window as opposed to prep. So the beauty of the use case and the ROI, this is why it makes it magical in the context of sales.

00:31:38:00 – 00:32:05:04
Unknown
That’s awesome. Yeah. And then the flip side with respect to to sales is, you know, there’s prep, there’s follow through, and then there is the magic in the middle, which is the actual demo or sales meeting. And if you manage the sales team, you know, there’s an 8020 rule and there is a small group of if you have a large enough sales team, there’s a small group of al performers, and you have outliers that are probably generating a multiple of revenue of everyone else.

00:32:05:06 – 00:32:26:15
Unknown
And what’s interesting is, you know, for the longest time we’ve had this concept of like a ten max engineer in the Bay area. And now, ironically, with AI, there’s 100,000 engineers. But this principle actually also applies to sales. And so sales leaders will call me and say, you know, Kyle, I have a disproportionate amount of my revenue that’s flowing from just a couple people, and I know they do certain things differently.

00:32:26:17 – 00:32:47:23
Unknown
But how can I distill and and unearth what they’re doing so that the entirety of myself can leverage their best practices the way they objection handling the way they talk about pricing. They navigate to next steps when they push back, when they act as a challenger sale. What’s their talk time? It turns out it’s 40%. The best people talk the least.

00:32:47:23 – 00:33:09:16
Unknown
They do a lot of listening. And so how do you have all of these insights? 5 or 7 years ago, they were kind of like the first era or a first generation of call recording calls or is gone. Of course, now everyone has their, you know, their note taker, whether it’s its audio or its order or its granola or you name it, there’s about like 100 of them that have have gotten venture capital funding.

00:33:09:18 – 00:33:29:08
Unknown
And, you know, now you have transcripts and so many you can grab every single transcript, put it in a common place. This could be as simple as a Google Doc or a folder in your G suite called transcripts. And then you can hook up a rack database or a small language model to those transcripts. And then the sales leader can ask any question he or she wants to ask.

00:33:29:11 – 00:33:45:20
Unknown
You know, what did Julian say when he was talking about pricing in the last month? Because he’s closed three times as many deals, or Julian just spoke to the same prospect that we haven’t been able to win the deal from 4 or 5 teams prior to what was said differently. And so then you start to unearth best practices.

00:33:45:27 – 00:34:07:15
Unknown
But the beauty of it is these are not best practices as it pertains to sales writ large. These are best practices in the specific context of your business, your value prop, and your customer. And it compounds and get smarter and smarter and smarter over time because you unearth the best practices they get shared, the team adopts them. There’s still some people on the team that outperform.

00:34:07:21 – 00:34:31:14
Unknown
Then what do they do next? How do they evolve? Now? Let’s capture those best practices now. Let’s disseminate those across the team. And oftentimes this is so. So how does it go. So that you know that that’s fantastic. You can go and you can you can measure what people are said. How is it, you know, if you’re building an application like this, is it turned into a training program or how does it manifest itself in terms of trying to help the bottom reps?

00:34:31:16 – 00:34:50:29
Unknown
Is is this kind of a feedback and training you get, you know, score for your, your your sales call? Or does it kind of give you, hey, here, five things that the best players do that you might want to try next time. How does how does it work? If I’m in a sales organization that’s implemented this? Yeah, in two very specific ways.

00:34:51:01 – 00:35:09:02
Unknown
There’s the there’s at the individual player. I see level individual contributor layer and then at the team level. So at the see layer Julian, if you’re the VP sales you’re my manager. We have our weekly sync. We sit down. You’ve asked 5 or 10 or 15 or 20 questions about all of my transcripts, and then the variance of my transcripts to the the best songs transcripts.

00:35:09:09 – 00:35:33:03
Unknown
And then you come with specific coaching points. And we have a very rich and vibrant discussion about how I can get better based on the data. Okay, a better one on one than you’ve ever had prior with your sales manager, VP sales. That’s how the I see level at the team. Wide level every week. Whether it’s Monday or whether it’s Friday, you can circulate a document of like, these are the best practices, these are the new best practices.

00:35:33:03 – 00:35:57:04
Unknown
This is something that we’ve uncovered. This is something that was just tried, that worked, and it’s disseminated as kind of like a weekly coaching or learning update. And the metaphor that I like to use for this kind of like weekly sales best practices capture is it’s like a playbook. If you’re a football team and you know your football player each week, you’re going game to game and you have you watch the film and you have a new set of learnings, and then the whole team can apply that new set of learnings.

00:35:57:04 – 00:36:16:09
Unknown
And so it tends to be a one pager that is disseminated. It can be talked about in a team like call or an all hands meeting. But oftentimes just an email of the new best practices have been captured that week that you should apply. Go for it. So at the team level and at the player level. Got it.

00:36:16:11 – 00:36:39:25
Unknown
Keep rolling. Yeah, yeah. This next one is a sales operations which is the bane of every salesperson. Right. That, sales operations and and, data hygiene with the CRM. Right, is, the bane of all salespeople. I would imagine everyone that will listen to this podcast and will watch this podcast has been in a meeting with multiple stakeholders.

00:36:39:27 – 00:37:02:07
Unknown
There’s multiple discussion points that lead to multiple decisions that lead to multiple owners, that lead to a variance of deadlines. And you lead the meeting and you feel like you got a lot accomplished. And by the end of the day, the next day. Wait, what exactly was Julien on the hook to do? And I think we made this decision, but someone disagreed or Julien was assigned this project.

00:37:02:07 – 00:37:37:13
Unknown
Or when is a Dubai? How do I track it. So similarly? You know with transcripts. And and the Claude anthropic models are particularly good at finding, you know, the signal in very messy and unstructured environments. And so an agent grabs a transcript of a group call if it’s happening on, on zoom like this. And then it will distill in shock and index that transcript and break down what was said that the decision breakdown, what was said that amounts to ownership breakdown, what was said that amounts to deadlines.

00:37:37:15 – 00:37:56:22
Unknown
And then it does two things. One, it captures the essence of the call, the key takeaways and emails it to everyone who joined the meeting. And then it’s slacks or emails every individual participant of what they’re on the hook to do. And by way, and so if it’s the engineering leader, their message on linear, if it’s the sales leader, maybe it’s their messaged, you know, via slack.

00:37:56:22 – 00:38:14:20
Unknown
And if it’s, you know, someone else in the organization that only lives via email or by text, you know, they get a message, this is what you’re responsible to do. And here’s the time you get to do it. But everything gets captured, so nothing gets lost. And what does this do? This eliminates let me ping the VP of engineering because I’m not sure if he heard what I heard.

00:38:14:20 – 00:38:32:26
Unknown
Let me ping the VP of sales because I’m not sure if she’s on the same deadline. Now the CEO calls. What was the decision? I don’t totally know. I have to ask a few people. And so this is about knowledge capture and then action routing. And it’s another thing that is leveraging the basis of transcripts to capture, again, decisions, owner deadlines.

00:38:32:26 – 00:38:50:28
Unknown
And it’s it’s a beautiful way to organize yourself, to move more quickly and not have drop balls coming out of Google. Yeah, I love that. Can you put this into some sort of a dashboards that so that it kind of monitors the ongoing processes as well? Yeah, totally. And this is, we can we can jump to this or we can just get to this in turn.

00:38:50:28 – 00:39:23:23
Unknown
But, you know, maybe the best use case or the one that’s probably most marketable is connecting everyone systems. So connecting. Yeah. IP or QuickBooks or linear or Salesforce or Slack or wherever your your email wherever things live. And the AI era, systems are able to talk to each other in a way that really wasn’t possible prior. And so what will happen is, is a group like mine or your CTO of your company would build an orchestration layer, which is a custom piece of software that sits on top of all of your systems.

00:39:24:01 – 00:39:44:06
Unknown
And, you know, one of your systems would capture your transcripts. And in this orchestration layer, not only can it, it can deploy agents to your systems or record populate data, collect data, and bring it back to that layer. But you would also display dashboards to show progress against any initiatives. And then even further more, take it to the next step.

00:39:44:09 – 00:40:03:03
Unknown
You can build in a chat interface to ask any questions about your systems as well. So. So that’s a long answer to a short question. But yes, you absolutely would build dashboards. You can see progress. And then you typically have a chat interface. So you can even ask questions about the dashboard. What’s your monitoring progress. Nice I love that.

00:40:03:06 – 00:40:43:29
Unknown
Where’s we jump to next? We can do some customer facing stuff. Yeah. Why don’t we do this one just because, you had mentioned, being the bane of, sellers existence, and I think that shared inboxes are the bane of whoever owns them. Yeah. Their existence. And so, if you’re listening on audio, you can’t see the page, but, you know, we have a at info, so maybe this is at, you know, gross elevated.info or this, you know, this is support, dot growth elevated or it’s or it’s the sales, inbox.

00:40:44:02 – 00:41:18:26
Unknown
And these inboxes are typically owned by junior person. This is not a job that anyone loves. Let’s, let’s let’s be honest. And what happens is in practice, if you have a shared inbox for, let’s call it sales, people tend to work through them in a chronological order. And so I got a inbound prospect. They probably don’t even fit my ICP at 8 a.m., and I’m going to try to respond to that one or ignore that one, or actually that one, or delete that one before I get to the one that was sent at 920 in the morning, I’m gonna try to handle that one before I get to the one of those sent at

00:41:18:26 – 00:41:41:25
Unknown
1030 in the morning and trying to just grind through, which, frankly, a bunch of junk is shared, inboxes mostly collect junk. And I’m trying to get to inbox zero by the end of the day to put out the anxiety of oh my gosh, here comes more emails. And really, there is very, very few emails that hit these shared inboxes that should be prioritized and should be actioned.

00:41:41:28 – 00:42:05:07
Unknown
So what can I do? Well, I can do three different things. First of all, it can classify the email or the email that hits any of these inboxes. So in your sales inbox, you know, is this a current client that has a rudimentary question? Or is this a current client that wants to churn, in which case it should be escalated to a human human?

00:42:05:07 – 00:42:28:09
Unknown
And so, so, so I can class of and classify a subcategory of the of the inbox. Secondly, because I can use logic and it can use rules that you codify, it can prioritize them. And this is what everyone that has this year’s inbox really wants to understand. If I let’s say I’m out of office, I own the the shared sales inbox.

00:42:28:12 – 00:42:46:18
Unknown
I come back to my computer. There are 450 emails I want a priority list of like, what are the top five? And it’s the top five. Two of them of the 400 are going to be customers. I should spend all of my time on those two, or at least right. 99% of them are a disproportionate amount of my time.

00:42:46:20 – 00:43:16:12
Unknown
And then to take it even a matter further for every single email that’s not, you know, those top two, they’re going to be customers. I can auto populate a response, the MCP or API to your outbox. So all all this person has to do is hit send or if it has a super high priority score, it can auto draft an email for you and then send it to your slack or notify you of this is high priority.

00:43:16:12 – 00:43:39:25
Unknown
Here’s a draft, but we want you to edit it first. Now there there are options to this. You have to build in exceptions and what we call carve outs. Anytime there’s a threat I should handle that. That should immediately alert a human. If you have someone with an angry tone, AI is not the best at countering an angry tone.

00:43:39:25 – 00:43:59:11
Unknown
That’s where it’s nice to have you pick up the phone. So and so there are carve outs that you can, codify and embed into rules, for edge cases. But on the by and large, these shared inboxes have huge volumes that are mostly noise. You’re looking for the signal. I find it, prioritize it, and actually tells you what to say.

00:43:59:15 – 00:44:24:09
Unknown
Yeah, that one’s really, you can you can feel that the weight of all those emails. You’re right. That’s that’s very strong. Where should we jump next? How about finance and operations? Yeah, sure. This one is is fun to talk about because we have multiple clients that had this same exact, I’ll call it, challenge. CFOs or controllers.

00:44:24:09 – 00:44:47:11
Unknown
One was a CFO, one was a controller. Both of them were tasked with what I call, the variance memo or the variance commentary. Oh, yeah. It’s part of the FDA process. So I, I’ve taught dozens of people how to do this. Yeah. It’s a yeah. And painstaking process. And you sit on boards and so, you know and I won’t I won’t, I won’t dox the innocent here.

00:44:47:11 – 00:45:10:06
Unknown
But for one of our current clients, she told me it takes her eight hours the Sunday night before the Monday board meeting to do the variance memo. And this is okay. Cogs are running hot, and I. Now I need to dig into why. Or we had a revenue mess. And, you know, I’m not exactly sure why or inventories miss stocked.

00:45:10:06 – 00:45:34:10
Unknown
And I now need to figure out why. And so you know previously you can use math and data. There’s a Python tool called pandas, which is cheap and easy to use, just deterministic math. And it can calculate, if anything is a standard deviation different than what it should be. But once you find that there’s a revenue mess now, you still have to go do the investigation to figure out what the hell happened.

00:45:34:13 – 00:45:53:21
Unknown
Right? So typically that process you’re you’re opening up the ERP and and, you know, diving into all sorts of, you know, sub tables and, you know, trying to find the the outlying transaction or two that, that mess things up is the are you able to go in there and navigate there. Doesn’t know where to look. It’s exactly right.

00:45:53:21 – 00:46:19:12
Unknown
So a couple things are true. Now I can in agents in particular can navigate unstructured environments. So what’s an unstructured environment. That’s a different software system. That’s a different structure of data. And so an agent is deployed and leaves effectively from the orchestration layer jumps into Salesforce probably in this instance or Monday or at ADL or any of these crumbs that probably won’t exist in ten years.

00:46:19:26 – 00:46:42:04
Unknown
But then they can spot and recognize, okay, our biggest new client this quarter with a verbal commitment. And then it’s a union contract signed on January 2nd. But now there’s an email that also has lodged in Salesforce with the agent reads, which says, oh, the point person internally that’s going to be tasked with quarterbacking the vendor. He’s actually going on PTO.

00:46:42:04 – 00:47:05:15
Unknown
So we want to kick it out from January to March. Now we lost two revenue months. An agent in the AI agent not only can find information and capture that information, but use logic to write the actual paragraph describing what hap. And so you get it’s impressive. You get to a place where this is the mess, and you’re looking at actuals versus budget versus the prior month typically.

00:47:05:17 – 00:47:28:19
Unknown
And they’re getting the answer and you’re getting the logic and you’re getting the context. And then fast forward I could talk about this one for two more hours because I love it and it feels magic. You know fast forward and you have the entire variance memo written and, you know, five minutes and then the controller rather than an eight hour exercise on Sunday night, she’s just doing some late editing as she’s watching Sunday Night Football.

00:47:28:22 – 00:47:54:18
Unknown
Yeah, that’s that’s huge. I mean, that’s that’s the DNA process. Is the memo writing after you know, sometimes days of, of of digging and, you know, swooping in and out of various tables and comparison and pulling up historicals. And so that’s huge. I that one’s near and dear to my heart. Yeah. Vendor invoices. That’s a that’s a huge one right there.

00:47:54:25 – 00:48:16:04
Unknown
Let’s talk about that one. Yeah. I’ll, I’ll try to go quickly because I, I get too excited and go into too much detail. And so if you are let’s say your, your accounts payable and you have a variety of vendors, your enterprise company, you have a huge amount of vendors. And so you get bombarded the first of the month, the end of the month with, you know, a ton of different invoices.

00:48:16:24 – 00:48:35:05
Unknown
And lo and behold, not every vendor did exactly what they wanted to do on tape, on spec, on budget. And so what you can’t do as an, you know, upstanding citizen is just go ahead and approve every single invoice and just pay a whole bunch of people without double checking that you got the value that you’re paying for.

00:48:35:07 – 00:49:04:17
Unknown
And so what had happened in the past is you get the PDF, hopefully it’s by excuse me, you get the invoice, hopefully it’s a PDF. You know, we’ve worked with companies where the invoices are literally handwritten. And then you are breaking down the PDF into its subcomponents, and then you’re trying to find the purchase order or the statement of work, and then you’re trying to find the goods receipt, and then you’re going line by line in the invoice with the purchase order and the goods receipt.

00:49:04:17 – 00:49:25:21
Unknown
And like, did we get everything that we were contracted to get? If we did, we approve. But hey, we didn’t get these two things. And now what do I do? And how do I escalate? And it becomes this incredibly time consuming process where you have these are just business processes that you know, have put humans in to reconcile data and recognize pattern recognition.

00:49:25:21 – 00:49:52:25
Unknown
Now, the I can just do that right. And I can just do it and, and the, the frontier AI models, are native with, with the ability to ingest PDFs. They have contracts with PDF plumbers and others, and the PDF plumber chunks are indexed to the PDF, breaks it down into its subcomponents, triangulate with with the Po and the good receipts, and does a three way match instantaneously.

00:49:52:27 – 00:50:13:10
Unknown
And then the 70 or 80 or 90% of the invoices that have the three way because you got the goods lined up with their approved. Yep. You just put it in the queue to hit approve. And then the ones that you don’t approve it specifically could service. This is what is missing. So not only does it get asked to do a human, this little human, they don’t have to review the whole goddamn thing, but they just have to review.

00:50:13:16 – 00:50:41:12
Unknown
This is the missing, the exceptions. That’s the cue. Like, if you could just focus on the exceptions 100%, that’s that’s that’s what you want to get to. I talk about connecting the tools because that’s a huge that’s a huge deal. Right. Well, the scope really quickly. Before we get to connecting to the tools, you know, if you go to if you go to 18, I won’t you have the ability for AI to continuously watch all of your operational data.

00:50:41:12 – 00:51:07:05
Unknown
I’m not going to dig into why, but but the reason why I think this is particularly interesting is, actually for a very, very sad day for this country, but nine over 11, you know, obviously happened. And what at the time it was called the Department defense, which now Department of War, there was, in fact, some signal that a terrorist attack was being planned and it was being planned against the World Trade Center, or the Twin Towers.

00:51:07:07 – 00:51:33:21
Unknown
And so why did we not take action to prevent it? Because it was one catastrophic signal amongst hundreds of thousands of other signals. And so we missed it. And that catalyzed Alex Karp and Peter Teal. And the earliest days of machine learning, conceptualizing. We need to find a way to use machines to synthesize and find the needle in the haystack so that never happens again.

00:51:33:21 – 00:51:52:15
Unknown
And so I’m not going to double click too much on this use case. But you know, yeah, that’s a great example. Just like the thousands of signals going on, it’s just too much for humans. So we missed it. Yeah. And Palantir is the company in the Nasdaq that trades for the richest, sales or revenue multiple of any company listed in the Nasdaq.

00:51:52:15 – 00:52:11:12
Unknown
And that’s partly because Karp is a, you know, crazy scientist and genius, in part because they had a head start. But it just shows the power of when I can continuously watch data and it can identify outliers. That’s just a human’s not going to be able to do because of the volume. So I just wanted to I remind you that counts for, for, you know, all right.

00:52:11:12 – 00:52:39:02
Unknown
Let’s talk about connect connecting your tools. Yeah. How how does that work? Yeah, we did a question when I was presenting at Jackson’s Hideaway, and someone asked me about the plumbing. I remember, and and years ago, there was, a major, revenue generating initiative for consultants around interoperability. Right. Which is like the silliest eight letter, eight syllable word.

00:52:39:17 – 00:53:02:01
Unknown
That’s that is why legacy software systems can’t connect to one another. And you know why you have to pay the big three big dollars to do so. And then you had the advent of the API era and software systems having an API key. That’s allowing people or other software to access their data. And there is an incentive to have an API key, because if you didn’t, then you’re kind of left outside of this new community.

00:53:02:01 – 00:53:20:15
Unknown
That was connected and you’re left outside the network effects. You can build clouds and kind of manually build connections and bridges to other software stacks. And now, you know, anthropic that has been so amazing in their product roadmap with Co-Work and cloud code. And now, you know, MCP is a way to call tools. So what are we supposed to say?

00:53:20:24 – 00:53:46:12
Unknown
We are in the golden era of connectivity and it’s I don’t want to make, you know, absolute statements, but it’s pretty much the case that you can connect almost all of your software to any other piece of software. You can input data to almost any piece of software. You can extract data from almost any piece of software, and you can bring that all together again into this orchestration layer.

00:53:46:25 – 00:54:12:26
Unknown
And then that allows a couple of things to happen. You know, one, agents can move through various instances of software to get work done. And then secondly, you can have dashboarding in the ability to query and ask any question in your organization, no matter where the information live. So for my first example, let’s say you have an inbound lead that got populated.

00:54:12:26 – 00:54:40:02
Unknown
And let’s say you have like an intake form on your website. So someone grabs that lead and agent brings the lead over to Clay or Apollo or Zoom info to populate additional information. The phone number, the title grabs enrich it, and then an agent pulls it to Salesforce. Populates in Salesforce. Now you have no longer. If it doesn’t Salesforce, it doesn’t exist.

00:54:40:05 – 00:54:55:18
Unknown
Now that you have a full comprehensive picture of the lead, you have the ability to lead. Score has a high lead score. An agent can then make a decision because agents are embedded with logic. Do I kick this to an SDR or do I put this on a nurture track? Or oh my gosh, this is dead set of the ICP and this is the decision maker.

00:54:55:18 – 00:55:18:19
Unknown
Let me kick this the whole way to my number one salesperson. And so because you can connect information now agents can flow across systems, but also because you connect information you have the ability to dashboard progress across divisions and across activities. And you have the ability to have a chat style interface as if it was ChatGPT. But just the data in your organization.

00:55:18:19 – 00:55:37:27
Unknown
And so, you know, the very beginning of this conversation, we were talking about decision latency, which is like, what’s the time it takes to get information, and then what’s the time it takes to action that information. This makes it instantaneous. And this probably feels maybe other than recruiting, this probably feels like the biggest magic trick today with AI.

00:55:37:29 – 00:56:01:06
Unknown
Yeah, that’s that’s incredible. Eager to eager to see see some of that that working. It’s it’s, so many tasks, so many human tasks that you just described, so many kind of connections that maybe wouldn’t happen or if nothing else, would take a lot longer to happen. To your earliest point about making decision loops faster is just amazing.

00:56:01:06 – 00:56:20:07
Unknown
When you can you can wire this information together. Yeah. And, let me just double click on one thing. I mean, as the CEO of a company or Julian as a chairman of the board, think about how much time is spent either in group meetings, getting information, or I’m the CEO of a company. I got to ask my VP of engineering something.

00:56:20:07 – 00:56:39:07
Unknown
Okay. Learn something from him now. Is he aligned with my VP of products? Okay, I got to watch this. Yeah. You’re getting an instant transparency of all the data. I think you mentioned, like, the building, the brain. Right? Like the company brain. Wait, what page that jumped you about on that? That that’s kind of a an amazing thing.

00:56:39:09 – 00:57:03:17
Unknown
Yeah. You know, we can stick right here. When you’ve connected all of your systems, you basically you connect them all regardless of the pathway, whether it’s an API key or regardless, then you have a database. And the beauty is you actually then don’t need to pay the frontier models. You don’t need to pay the tokens from OpenAI or the tokens from anthropic.

00:57:03:17 – 00:57:28:29
Unknown
We all we all heard these horror stories of of token maxing and running up huge bills. But you don’t need the latest and greatest private model that could be a Nobel laureate or get the perfect mCAT score and Lsat score at the same time, because you’re just focusing on your your package that’s on your package data. And think about the size of your package data relative to everything that’s ever existed on the internet in the world.

00:57:29:03 – 00:57:54:06
Unknown
Right? So we’re talking like a millionth of 1%. And so what’s the level of horsepower required to do inference or synthesis or memory or context or querying across just your company’s data? Turns out very, very little horsepower relative to the new fable molecule dropping or open an iPhone. Why does that? But that means that your company brain can be a low cost, open source model.

00:57:54:06 – 00:58:13:09
Unknown
And it doesn’t have to be the Chinese. I know people, we were the Chinese model. I know they’re they’re catching on quickly. I know we’re in a in a in a race for the AI, AI supremacy or for artificial, you know, general artificial intelligence. But there are now American open, models. Yeah, there’s going to be and there’s gonna be locally hosted.

00:58:13:09 – 00:58:39:04
Unknown
Like, the key is that if you can, you know, this goes back to the same ETL exercise. If you could basically extract, transform and load the data in one place where it’s accessible, right. The the data for one organization, even a relatively large company, it’s not like indexing the internet. It’s, it’s a it’s a, you know, consumable, measurable amount of data that that could be managed, like you said, it could be managed in the private model, it could be managed for the hosted model.

00:58:39:04 – 00:58:59:21
Unknown
It could be managed, managed with lower cost resources that are going to come out. I’m sure there will be more these coming out every month and, you know, but but how hard is it getting all the data together? Do you have to kind of do a lot of ETL or can it, can you can it just find them all in their own own repositories and, and, index them together so that it can be useful at this point or that still in process?

00:58:59:21 – 00:59:19:20
Unknown
No, it’s not that hard. Set up a database. It can be pinecone, it can be super base. Find a way to connect to all of your systems. Again, most of them today have a have an open AI key that it plays nicely because it’s in their best interest as a software company to play nicely. To actually deploy the model is like a 15 minute exercise.

00:59:19:20 – 00:59:46:24
Unknown
I like the Google Jama models, but to your point, which was dead on every day a new American open source model comes out and that becomes the brain that can surface any information, that lives anywhere in your company. And I, I guess the other point that’s, that’s super important and worth, double clicking on is, you know, an open source model might be a way to spend $25 a month rather than $5,000 in token costs.

00:59:46:26 – 01:00:09:07
Unknown
But an open source model is also the most secure way to have your company’s data exposed to your capabilities, because it’s not even connected to you. You can even host it locally and have it all in your local databases, right? You have it all in your local database. There’s this miraculous comeback of like the, Apple mini hard drives that everyone’s buying today, right?

01:00:09:07 – 01:00:35:21
Unknown
Right. Their prices through the roof. So this, this, this example right here, this, this company brain is their, is their a good example of of it of it working that like you can point to or you’ve seen like a demo of like, like like it sounds like a, like a brilliant idea, but, I, I’d love to like, see one working or even even, you know, even see a science fiction movie about one working.

01:00:35:21 – 01:00:56:21
Unknown
I think. Is this the Jarvis kind of thing, or is it actually the what I was doing five minutes before we just hopped on this call was chatting with the company brain. We set up for my company, which is which invoices are outstanding that have been countersigned and which invoices are past due that haven’t been paid. And I am not opening any software systems.

01:00:56:21 – 01:01:22:24
Unknown
I’m not even opening my email, just talking to a chat interface on orchestration layer that’s connected to all of our software stack and connected to my email. That’s pretty awesome. Literally, what I was doing before, before we just hopped on this call. That’s a great example. Well, Kyle, this is this is a ton of great information, will, we’ll, I think people really, appreciate these use cases, and we’ll we’ll share the deck out with the audience.

01:01:23:07 – 01:01:42:21
Unknown
You know, if they ask. And, where can they find find you and zyga, if they want to want to get your help on doing some of this. I’m all over the internet, Twitter, LinkedIn. But Kyle Kiely at zygote, I goes, I feel free to email me, find me on Twitter, find me on LinkedIn. And and I genuinely mean this.

01:01:42:22 – 01:02:09:15
Unknown
I’m happy to chat. I my philosophy and business building is the more that you can educate the masses on long term. Greedy and good things come from me helping small businesses and typically what happens with my businesses. So I’ll get introduced to a business owner and I’ll take the time to try to educate them on how they might get started and where there is fast ROI or where there is man hours that that that AI agents could tackle to give them time back, and then that person’s brother in law or college roommate calls me and I have a client.

01:02:09:15 – 01:02:27:21
Unknown
And so I think it’s in my best interest. And just something I enjoy to do is try to educate business owners. So feel free to reach out emails best. But Twitter, LinkedIn or anywhere else you can find me is is is also good. Awesome. Thank you for your time today, Kyle. Awesome. This was fun. Thanks, Julie.

Introduction to Growth Elevated & Kyle Richards (00:00:00)
Julian introduces the AI Sherpa series and welcomes Kyle Richards, founder of Zygote AI, to discuss practical AI implementation for businesses.

Why AI Adoption Is an Implementation Problem (00:04:07)
Kyle explains why the biggest AI opportunity isn’t building models—it’s helping businesses adopt AI faster than the technology evolves.

The Biggest AI Misconception: Decision Latency (00:08:33)
Why AI isn’t about buying tools, but about eliminating delays and helping organizations make better decisions faster.

Leadership Matters More Than Technology (00:12:21)
After dozens of AI implementations, Kyle shares why change management and executive buy-in determine AI success more than the models themselves.

How to Measure AI ROI (00:16:27)
Why successful AI projects start with a clear, measurable ROI instead of broad transformation initiatives.

CEO Mistakes When Adopting AI (00:19:50)
Why CEOs shouldn’t delegate AI solely to the CTO and should instead focus on operational bottlenecks across the business.

Practical AI Use Cases Across the Business (00:22:31)
Kyle walks through real-world AI workflows for recruiting, sales, finance, customer support, operations, and meeting management.

Building the Company Brain (00:52:11)
How connecting business systems with AI creates a searchable knowledge layer that dramatically reduces decision latency across the organization.