Removing Bottlenecks, Not People with Stephanie Sylvestre

0:00 Stephanie: What we do is we look at processes,

0:02 we run them manually,

0:04 and then after a while we automate them.

0:24 Scott: Hello, everybody.

0:25 Welcome to another episode of the How Many CTOs Does It Take podcast.

0:29 My name is Scott Porad,

0:30 my trusty co-host.

0:32 Brad is on vacation this week, and I have a special guest here today,

0:37 someone who I'm very interested in hearing about what she is working on,

0:40 Stephanie Silvestri.

0:41 So welcome to the pod,

0:43 Stephanie.

0:43 Stephanie: Thank you for having me.

0:44 I love the title.

0:46 How many CTOs does it take? And looking forward to a great conversation.

0:51 And I'm with Avatar Buddy.

0:53 I started Avatar Buddy 10 years ago.

0:55 So I was in AI before AI was a thing. In fact, people are like, I don't know what you're talking about.

1:01 And I still think people now just go,

1:05 whatever.

1:05 I still don't know what you're talking about.

1:09 And

1:11 I have been super excited with the evolution because 10 years ago I said we can we can use AI to help people be the best version of themselves and people taught that's

1:24 insane that's ridiculous um it's too big and I'm like it's supposed to be that big because that is how big AI can be and so to be able to see how big AI has become um is kind of validating and we have been

1:44 doubling down,

1:45 tripling down, quadrupling down on the fact that AI plus human takes us to a whole different level.

1:55 And I have been taking people to task whenever they talk about AI to eliminate jobs or

2:03 eliminate people.

2:05 And so that is that has been like my latest thing.

2:08 In fact, we just launched our remove bottlenecks, not people campaign.

2:14 That's all around.

2:15 Let me tell you how you can identify bottlenecks in your organization. And then let me show you how you can remove them while keeping the people and increase throughput,

2:28 increase profitability.

2:30 Increase impact.

2:31 Scott: Okay,

2:32 there's a lot. I have a lot of questions. Just in like one minute, I have a lot of questions. So can we go back to the very beginning?

2:39 Stephanie: Yes.

2:40 Scott: So 10 years ago,

2:42 I feel like

2:45 I was just a baby in diapers when AI wasn't even on my radar 10 years ago.

2:50 I mean, it was a thing maybe we heard about we saw in science fiction movies.

2:54 How did it come to be that 10 years ago, you were already thinking about this, that you already knew about it? Like, what was that? How did you get to the starting line a decade ago?

3:04 Stephanie: So I've been in tech for over 30 years and one of the things that I saw when I first started

3:10 In tech was people would learn a language,

3:13 focus on that language,

3:15 the world would evolve around them, they'd get laid off,

3:19 they'd be angry,

3:20 and then blame other people. I was like, you should blame yourself.

3:26 And I said, I don't want to be like that.

3:29 And secondly,

3:30 I always like to learn new things. I like to keep my skills up and come in. I like to think of myself as the oldest Gen Z.

3:40 And soon I'm going to be like,

3:41 I like to think of myself as the oldest alpha.

3:45 And so I was at the point in my career where I was like,

3:49 I need to learn something new.

3:51 What am I going to learn?

3:53 VR.

3:55 AR or AI and VR is virtual reality AR is augmented reality and then AI artificial intelligence so I didn't want to do VR because I I was like this makes me nauseous augmented reality I just could not see why people would use it it just did not have a practical implementation in my mind you know

4:22 I come from the school of if the solution takes more than two clicks,

4:25 it's high friction and people are not going to use it.

4:29 It's just not,

4:31 you're not going to be able to deploy it.

4:32 And then I said AI.

4:35 AI covers everything anyway,

4:38 so I got into AI.

4:39 And I was at dinner one night and the restaurant was crowded,

4:44 it was loud.

4:45 And my friend said I should create a digital mentor.

4:49 And I was like, oh my God, that's so great.

4:51 That's such a beautiful,

4:52 that's a wonderful idea.

4:54 We should definitely do that.

4:55 And then she was like, that's not what I said. I said, well, it doesn't matter what you said.

4:59 Scott: Ha,

4:59 Stephanie: That's

5:00 Scott: ha, ha,

5:00 Stephanie: what we're going with.

5:01 And that's how Avatar Buddy was born by mistake.

5:04 I was just telling some of my employees that you have to fall into success,

5:10 right?

5:10 Or

5:11 Scott: ha.

5:11 Stephanie: fail forward.

5:12 And so I want them to go faster.

5:16 I'm like, you need to move faster.

5:18 They're like, well, we got to take time.

5:19 I'm like, no,

5:20 no, no.

5:20 make all the mistakes like the faster you move the more mistakes you you make and the more faster we find the what's not working so we can fix it like I don't want to take five months to figure out something we could figure out in five weeks no and they're looking at me I'm like yeah fail forward

5:37 Scott: what did the first version of avatar buddy look like i mean 10 years ago you didn't have you know now we've got all these tools the frontier models are available to everybody

5:46 What were

5:47 Stephanie: I mean,

5:48 Scott: you doing back then?

5:49 Stephanie: it took we did lots of research.

5:52 We didn't we and then we got some.

5:57 Teenagers from after school programs around Miami to write out requirements for us. We taught them how to write user stories.

6:05 That's how agile methodology uses to capture requirements.

6:10 And then that was in 2019.

6:13 So I mean, it took a lot of time to research,

6:15 to find, to figure out, to find anybody that even understood what I was talking about.

6:22 Scott: Yeah.

6:22 Stephanie: It's like.

6:23 and and um people were talking about you know crypto and i was like no it's not crypto it's not blockchain it's it's different so we got all our requirements and then covid hit but the good thing about it is my current developer he had been telling me that that like this is just insane you can't that can't be done it just can't be done stephanie and i was like yes it can't be done i can think about it it can be done i'm not that smart

6:49 he's like no and so he started doing some research i said oh this is what stephanie is talking about and

6:58 Scott: Mm-hmm.

6:59 Stephanie: so

7:00 um he built out this really cloozy uh prototype on that we deployed on the martin luther king day on 2021 and

7:11 Scott: Hmm

7:12 Stephanie: then we were just kind of like building this stuff and he's like stephanie i don't think you can build this stuff by yourself and i was like we'll try so we're you know we're there we're trying it's not really doing what i want it to do i was like nah it's not

7:26 We've got some teenagers to test it.

7:28 They're giving us feedback where we're rewriting the software.

7:31 It was just crazy, crazy, crazy, right?

7:34 And then we got into open AI and back then you had to apply to be a developer and write justification and it was like competitive and they would say no or yes.

7:46 So we got into open AI developer community in 2021 and that really helped us.

7:54 But, you know, back then.

7:58 The GPT was just a mess.

8:01 It was like a hot

8:01 Scott: Yeah.

8:02 Stephanie: mess.

8:02 Oh my god.

8:04 In the summer 2022, we got 145 teenagers and we taught them how to do software testing and how to validate our requirements.

8:18 And we had like three pages of defects that we sent off to OpenAI.

8:23 and so I tell them I said we were part you were part of creating chat GPT because OpenAI would do is give the stuff out the developer community would find all the defects and then we would send it back to them they would fix it we test it and and so when they went out in November I was like this is cool I'd already planned to go live in January because I was like

8:47 I'm at a point where we need to put it into production and finish testing in production.

8:53 I know people don't like when you say that,

8:55 but we all know if you're a good CTO that you have to test in production.

9:00 Otherwise,

9:00 Scott: Right.

9:00 Stephanie: I mean, no matter how pristine your test environment is,

9:06 you have to test in production because your testers and your developers.

9:13 do not think the way the end users think and

9:16 Scott: Right.

9:16 Stephanie: and so we had um our first first customer I still I have the check I took the check I put it in a little frame put it on a shelf in in my in my live in my living room and it's just kind of like um where I look at it ever so often remind me of where we've come

9:36 And, and that's that that was it,

9:39 right?

9:39 We just started refining,

9:40 we realized like, you know what?

9:43 This LLM stuff is by itself is just crazy.

9:47 And I remember back my day when I worked at Burger King and remember portals, remember when portals first came out and people

9:55 Scott: Yeah, of course.

9:56 Stephanie: were like,

9:56 what,

9:57 what is this? And it was just so complicated and difficult to implement.

10:01 We said,

10:02 you need a data vault.

10:04 We called it a small language model.

10:06 Now people call it rag.

10:08 I don't like that term.

10:09 um and so now we call it data vault because small language model people like what

10:14 so data vault and so that's like one of our differentiators because we figured out data vault and

10:21 What are people now talking about?

10:23 Data vaults. Like you got to give the AI just a small subset of curated data.

10:28 That's when it's best.

10:30 We figured that out in 23.

10:32 Scott: Can we skip ahead for a second?

10:34 Because I realize I know something that maybe people who are listening don't know. We kind of need to skip to the end of the story for a second.

10:41 What is Avatar Buddy?

10:44 Because I realize we talked before and I looked at your website before the podcast,

10:47 but we haven't actually said for people who are listening.

10:48 People are listening.

10:48 What is AvatarBuddy?

10:50 Stephanie: Avatar Buddy is a managed AI as a service company that builds AI agents on your behalf and manage manages them and then provides AI advisory services to ensure that your AI investment is maximized. So basically, we're your AI team.

11:09 Scott: Mm

11:09 Stephanie: But

11:10 Scott: -hmm.

11:10 Stephanie: most places you'll find are either doing product or they're doing service.

11:14 We're doing both.

11:16 And if you're keeping up with the space, you just saw that OpenAI just made some deals with the large consulting companies.

11:27 Anderson and or is it Mackenzie Mackenzie and BCG group we've been doing that since 2023 as well that we're helping you figure out how to implement AI configure it for you

11:44 And we've maintained tight control over configuring the AI.

11:48 We don't let people configure AI on our platform because,

11:51 in my opinion,

11:53 the way you configure AI is the most dangerous part of

11:57 Scott: Hmm.

11:57 Stephanie: AI.

11:59 And you see with Metos now,

12:02 the way they've configured that.

12:04 And now they're trying to figure out what they're going to do.

12:07 Scott: Talk about a genie you can't put back in the bottle.

12:10 Stephanie: One of the things that people like to ask about when I go talk, they're like, what's your background?

12:19 Like what's your background?

12:20 Do you know anything about AI?

12:21 Scott: Mm

12:22 Stephanie: And

12:22 Scott: -hmm.

12:22 Stephanie: I will say that people that just know about AI,

12:25 they are the most dangerous people walking this planet right now. Because if you don't understand how software works,

12:32 even though AI is not really a software,

12:35 but it's adjacent to software.

12:38 If you understand how software works,

12:40 you're going to build AI products that are just horrible for the society.

12:46 So like I think about Mythos or Claude Code or any of those things and I think to myself,

12:53 these guys don't even understand how software works because if they did, they would not even think that this was a good idea to build. Because all they did was.

13:03 made it easier for everybody to be hacked and it's in such an out of scale and level it's so unsafe right like you we have we literally have no safety anymore it's like everybody's house even though you locked your door the bad guys have the keys the master keys

13:24 Scott: Mm

13:24 Stephanie: Even

13:24 Scott: -hmm.

13:25 Stephanie: though you have an alarm set,

13:26 the bad guys have the master key to override your alarm.

13:30 And just for your people listening out there, just let me break down why I have an issue with this.

13:38 So when you build software,

13:40 you never have a defect-free software.

13:43 What you have is a piece of software that has the least amount of defects. If you're a really good QA person.

13:51 You're going to have, you're going to build software,

13:54 you're going to test it, you're going to test it in a way that 95% of the bugs have been shaken out.

14:01 Mm But

14:02 Scott: -hmm.

14:02 Stephanie: we all know there's a 5% chance.

14:04 And, you know, I remember we used to say this thing that it's possible,

14:08 but not probable.

14:10 So.

14:11 It's possible that somebody will find this defect,

14:14 but it's not probable.

14:17 Scott: Right.

14:17 Stephanie: And you look at probability and then that's how you determine whether or not you should release the software.

14:24 Now, we know that lots of large...

14:30 Fortune 50 companies still run on mainframe. That's the dirty little secret that we rarely ever talk about.

14:39 And over the years, what they've done is they bolted the latest software on top of the mainframe.

14:47 And then they jimmy-rigged. I mean, jimmy-rigged, that is a valid IT term.

14:57 Right?

14:58 We're

14:58 Scott: Yeah.

14:59 Stephanie: just going to generate this.

15:01 There's lots of hard coding happening in the back end.

15:05 And then those developers leave,

15:07 they retire.

15:08 And how can it help you if you're no longer friends with them?

15:13 Because now when you go try to fix that code,

15:15 you can't because you don't know what else you're going to break. Because is there any comments in those codes?

15:24 Probably not.

15:26 Documentation can't find it.

15:30 So that's the state of our software globally.

15:34 Nobody has clean,

15:36 pretty,

15:37 pristine, defect-free software.

15:39 Now you're going to build a solution that can find all the defects in under nine seconds.

15:45 How is that okay?

15:47 Scott: Can it fix them in under nine seconds?

15:49 That's the question. If they can find

15:50 Stephanie: No,

15:50 Scott: them,

15:50 Stephanie: they can't.

15:51 Scott: can they fix them?

15:53 Stephanie: No, they can't.

15:55 Because one,

15:56 remember AI doesn't have state the way we have state, where I can be like,

16:03 Scott,

16:03 remember that piece of code you wrote 10 years ago?

16:06 Can you remember?

16:07 And if I probe you enough and...

16:11 shake you enough you'll be oh yes this is why I did it this way AI can never get to that so how can you fix it does

16:20 Scott: Yeah, it doesn't know well to use the buzzword. It doesn't know the context.

16:24 Stephanie: not know the context and and then um we all know you write a piece of code

16:32 And you fix it in one place and it breaks 15 other things that your

16:36 Scott: Mm hmm.

16:37 Stephanie: logic going like,

16:39 how did it break that?

16:40 That couldn't have worked.

16:42 Like, how are those two things connected?

16:44 Scott: Yeah.

16:45 Stephanie: Because in the developer's mind, it made sense.

16:48 Scott: Mm

16:48 Stephanie: And

16:48 Scott: hmm.

16:49 Stephanie: that developer is not there fixing that code.

16:51 Scott: That's right.

16:52 Okay.

16:53 So if let's say that my company wants to work with Avatar Buddy,

16:57 like on the services side,

16:59 let's start there.

17:00 What's that kind of engagement look like? How's that go?

17:04 Stephanie: So, first of all, we have a scoping call.

17:07 It's like the first date.

17:09 Like, do I like you?

17:10 Do I like your energy?

17:14 And so...

17:16 I'm very quickly moved through the state of any customer is a good customer like no to if the customer does not align with our value system does not bring us joy I don't have any enjoyment speaking with them I don't want them as a customer and and people will be like well you can't do that I was like yes you can because

17:39 This is about making the world a better place.

17:41 Why would I want to have a bunch of customers that doesn't bring me joy?

17:46 That just means that I'm going to be a miserable person and anybody around me will be miserable by extension.

17:50 Scott: Mm-hmm. Mm

17:51 Stephanie: So

17:51 Scott: -hmm.

17:52 Stephanie: that's the first thing we do.

17:54 Then the second thing we do is we do an AI opportunity mapping session,

17:58 which is I just sit down and I say,

17:59 Scott,

18:00 what problem are you trying to solve with AI?

18:02 Scott: Mm-hmm.

18:02 Stephanie: If you can't articulate what problem you're trying to solve,

18:05 I can't help you.

18:07 Scott: Yeah.

18:08 Stephanie: And some people get tested with that and people are like.

18:13 Why do you do that on the front end and part of your sales cycle and you do it for free?

18:18 You could charge two,

18:19 three thousand, five thousand dollars for that.

18:21 And I go, if I understand why they want to solve what problem they want to solve for and we get an alignment and agreement,

18:32 I will more than make up for that five thousand dollars on the back end.

18:37 Scott: Mm-hmm.

18:38 Stephanie: But if I don't, I will spend it will cost me more than the $5,000 in the back end.

18:44 Scott: Okay, so can you offer like a little bit of advice?

18:49 Stephanie: Yep.

18:49 Scott: what is what's an example or two of clients you've had who have come like where good problems like a good like you've had a you've had a client that's come prepared and they've they've said x this is it and i'm sure you've had clients that have come unprepared and they've said why like what does prepared look like and what does unprepared look like

19:10 Stephanie: So prepared looks like.

19:12 um you understand your business process you brought people that do the work to the meet-in so

19:18 Scott: Mm-hmm.

19:19 Stephanie: like i just one person goes i don't need to talk to the team i was like yes you do they're like no we don't need to talk to the team i was like no this is not a good customer because

19:29 Scott: Okay.

19:29 Stephanie: if the team doesn't have buy-in and the team's not helping solve the problem then

19:36 You know, I mean, we've both been leaders in organization and somebody comes and says this is the problem.

19:44 And then you might have been busy and you say,

19:46 OK,

19:47 let's solve that problem.

19:48 And then halfway through solving the problem,

19:50 you go and you you finally go talk to the people that are dealing with the problem. They're like, no,

19:56 like that's not the problem.

19:57 Scott: Yeah.

19:58 Stephanie: I'll give you an example.

20:00 This company said, we have an email problem.

20:02 Can you help us put in an AI solution to help with our emails?

20:07 And I said, well,

20:08 no,

20:09 I'm not the company. I don't do that.

20:10 Don't automate emails.

20:13 Scott: Mm

20:13 Stephanie: I

20:13 Scott: -hmm.

20:13 Stephanie: said, you have an email problem?

20:17 I said, tell me why you have an email problem.

20:20 Well,

20:20 it turns out they did not have an email problem. They have an email.

20:24 problem email the email is a symptom of a bigger problem so

20:28 Scott: Mm-hmm.

20:29 Stephanie: there's two things one is they didn't have a systematic way to collect all the requirements so what people would do is they'd send these emails back and forth back and forth to CYA because

20:45 I don't want to like Scott didn't give me all the information. I want to black and white. It's Scott's fault if this

20:50 Brad: Mm-hmm.

20:51 Stephanie: deal goes through because he's so I said, look,

20:56 we'll sit with your person that prices out the orders and we'll ask them all the different parameters that are needed based on whatever type of orders you get.

21:10 Always give it to the AI and we'll tell the AI that if they don't give you all the information you asked it for all the information.

21:18 When I send the email to Scott.

21:21 for the order fulfillment he has all the thing one email as opposed to seven did you have an email problem so

21:29 No.

21:29 Well,

21:30 you did. There was an email problem of too much emails, but that was really the symptom of the deeper problem.

21:36 that's what we do in an AI opportunity mapping session just go deep and understand and find the bottlenecks

21:43 and look at the processes and say,

21:45 we're going to just automate this piece of the process, we're going to capture this piece of the process,

21:50 we're going to document this piece of the process, and the rest of it you just carry on, and that automizes and saves.

21:55 Okay. So part one is, Hey, the right people have to be in the room.

22:00 And part two is you have to have an awareness of what.

22:07 root causes are the problems you can't just say if if you call Stephanie up you call avatar buddy up you can't just say make my people more productive I mean you could and I suppose you could spend a bunch of time with the team trying to figure it out but um that's not there's a little bit of awareness that has to happen by the client about what's going on

22:28 Yeah, everybody knows where the problems are.

22:31 So I say to the leadership,

22:33 like, we have to create a safe environment where

22:35 Mm-hmm,

22:36 people can tell you the truth.

22:38 So

22:38 mm-hmm.

22:38 I was actually very refreshed.

22:41 One of the sessions I was at,

22:44 these people were like,

22:45 the CEO is a bottleneck.

22:47 Everything has to go through him.

22:48 He's never here.

22:49 And then he changes stuff.

22:52 And so, you know, I get my work done on Monday and then on Thursday night.

22:58 he's sending me all of these revisions and now friday i'm scrambling to get it done and

23:05 Mm-hmm.

23:05 you know it that somehow it's my fault that it doesn't get done so

23:11 we're like okay well let's let's capture how how he works run run the the your finished product through the ai

23:22 That is his digital twin and then get all the edits done so when you send it to him, if he reviews it on Thursday,

23:31 he has no edits or maybe one little edit so you're not scrambling on Friday.

23:35 Interesting.

23:36 So what have you found in the people that you've worked with, the clients you work with, what are the types of things or even an example that's been just a tremendous success?

23:46 Like what should...

23:48 What should we all be looking out for in our organizations as the kinds of things that seem to work really well?

23:55 We know in the engineering space,

23:57 we're seeing it starting to work really well, the writing software.

24:00 There's still pitfalls and we need guardrails and things like that.

24:03 What other parts of these businesses,

24:05 not these businesses,

24:06 of businesses are you seeing where this stuff works really well?

24:10 So I think that's um

24:12 workers spend a lot of time staring at a blank screen trying to like get their thoughts out if

24:22 Oh, yeah.

24:23 the AI knows how to do your job the way you do your job and has the data it can help you get those thoughts out faster like

24:33 Yeah,

24:34 that makes a lot of sense.

24:36 like I'll give you an example

24:38 I wanted to do the remove bottlenecks, not people campaign.

24:43 And I don't know how to run a marketing campaign.

24:47 I don't know how to write up a marketing campaign. I don't know how to do any of that stuff. Right.

24:50 But I did know that everything needed to be aligned.

24:54 And I just said to the AI,

24:55 like, this is the campaign I want to run.

24:57 This is the emotional hook that we want to ground it on.

25:00 And it needs to go across all of these. These are all these channels that I'm using. And it wrote.

25:06 a pretty slick marketing campaign for

25:09 Mm

25:09 me and

25:10 -hmm.

25:10 then I went through and I just tweaked a little things like it used some terminology I didn't like so I had to go back and I had to fix it then I had to go through the document make sure it's all and then then then I just played it through right I talked to each person on the team

25:27 Like, hey,

25:28 how does this work for your thing?

25:30 This is what we want to do.

25:31 And so like that's an example, right?

25:34 That took me about six hours to do.

25:39 If I had hired a marketing consultant and I brought somebody in,

25:45 like that would have been like a six week effort or maybe even longer.

25:49 Yes.

25:51 And.

25:52 And that's the beauty.

25:55 But why was I able to do that? One,

25:58 I had an AI agent that's configured to be a marketing guru.

26:02 Mm And-hmm.

26:03 we configure that based on a marketing guru,

26:06 like the AI is based on an actual human person.

26:11 And two,

26:12 we gave the AI all the context and data for our company.

26:18 So it took our existing campaign,

26:22 everyone needs a buddy,

26:24 safe,

26:25 simple,

26:26 fast,

26:26 human digital handshake, and it pulled all of that in along with remove bottlenecks, not people.

26:37 Mm

26:37 And

26:37 -hmm.

26:37 that was the level of connectiveness that I've been trying to get and I was not able to get.

26:43 That's an example of like,

26:46 you know, what a human have been able to do that? Probably not.

26:52 And if they had, it would have taken so long that I would have lost interest and wandered away.

26:57 Mm-hmm. Mm-hmm.

26:59 And that's that's where.

27:02 You see AI really having a true impact.

27:05 So now I say remove bottlenecks, not people,

27:08 or eliminate bottlenecks, not people,

27:10 remove,

27:11 eliminate their interchangeable, you know, you tell the AI every, every, we got a social media buddy,

27:21 every social media post has to have.

27:24 Save simple fast,

27:25 manage AI as a service,

27:27 eliminate bottlenecks, not people as a hashtag.

27:31 I was thinking, as you were talking,

27:33 I was thinking about how are you going to scale your business?

27:41 And because meeting with each client one-on-one is time-consuming. I was thinking.

27:51 Is it the case that there could be an AI agent? You just talked about the introductory call or maybe a few calls you might have with a potential client to try and understand the real root causes of what they have going on.

28:02 And I'm wondering,

28:03 could that itself be an agent?

28:06 No.

28:07 Now you think that needs to be a human?

28:09 I mean, you could obviously be a human, but based on our value system,

28:15 it's not going to be a human.

28:16 What we do is we look at processes,

28:20 we run them manually,

28:21 and then after a while we automate them.

28:24 Yeah.

28:25 So every time something becomes too much.

28:29 we automate them like i just had a conversation with my cfo yesterday i said let's fix the ai agent so you can close in one day end of month close should be one day she's she's like man like the agent is great uh something that would have taken me a few days i took just a couple i took um four or five hours and and she's like i don't like how it writes

28:53 And I was like,

28:54 well,

28:54 okay,

28:55 we'll fix that.

28:56 teach it that's very interesting the close in one day because where you can you can say it's taking us n days to close the books and you can look at all the different parts that cause that time to add up and it's a very objective measure of is it getting better and you can look at that piece by piece i'm reminded of a time that just reminded me of a time

29:19 Hi, this is a long time ago. When was this? This is about 2010.

29:24 And I was at this company and it was taking forever to deploy our software.

29:30 And we set a goal for ourselves that we wanted to take 10 minutes to deploy.

29:35 And the first thing we had to do was measure how long does it take to deploy now.

29:41 And it took I remember this, it took 174 minutes to deploy.

29:45 And so then it was a matter of shipping away.

29:48 And so it's interesting that the kinds of processes that have a very measurable, like closing the books for a month, the kind of processes that have a very objective measurement of like,

29:59 is it getting better?

30:03 Are well suited for this kind of work.

30:06 Exactly.

30:07 So now when I meet with her, we're

30:11 Mm-hmm.

30:11 having real conversations about the business.

30:16 She's saying,

30:18 oh, by the way, I think once we get down to we can close in a day.

30:25 I said,

30:26 once you, we're in this woman's group together, I said, once you come in, once you get to get it to close in, close in one day,

30:34 we're going to host a meeting, we're going to host a pop up with this with the women from,

30:40 and you're going to talk about how you're using AI to close the books in one day.

30:44 And that will, that's how we're going to sell it.

30:47 And people are going to be like, whoa.

30:50 And so if I take the time, right,

30:53 like on the front end, if I take four hours on the front end to get to know a customer,

31:01 learn what they're trying to do,

31:03 onboarding them becomes easier, managing them becomes easier,

31:09 building out their solution becomes easier. We have some AI.

31:14 agents that we give other AI agents to create right not all our AI agents are created by hand it's

31:21 Turtles all the way down.

31:23 it's exactly um and and so it's like kind of like the same thing right um we have a marketing company and I was just like I am tired of reviewing the same thing every month and making the same freaking edits every month

31:41 So I was like, if you don't use Marketing Buddy to help you write it, I'm just going to send it back.

31:48 Mm-hmm.

31:49 And they started using Marketing Buddy and they're like, yeah,

31:52 like it took five, it used to take five hours to create the content calendar. It takes an hour and a half.

31:59 I was like, it's in your best interest because I pay you a fat rate.

32:03 So if it takes you 20 hours,

32:04 it takes you an hour and a half,

32:06 Yeah.

32:06 you're still getting paid the same amount of money. So you want to reduce it.

32:11 And that's where.

32:13 the service and the product comes in right it cannot divorce the two and when people try to divorce the two you don't have the impact it's why Anthropik is winning right now Anthropik Enterprise is winning because they figured out that they have to give consult with the product it's why OpenAI just went and bought a little boutique consulting firm out of London and is making all these deals with these companies because they they

32:40 They realize that like I like to tongue in cheek say Sam Altman has like a wire in my bug in my house and whatever I think about like three to six months later,

32:52 he comes up with the same thing. So he's following, he's following me.

32:56 The whole point is, is that

32:59 Because I understand software delivery,

33:02 because I understand how business works,

33:04 because I did IT quality,

33:07 because I also ran,

33:09 helped run a small government, a local government entity. It was only $184 million.

33:19 Pretty substantial.

33:20 I understand both sides of the table.

33:23 I have a master's in economic development,

33:27 so I have the social skills.

33:29 So when you come to Avatar,

33:32 buddy,

33:32 what you're buying is my IP.

33:36 And as AI becomes everywhere and so it's nowhere,

33:43 what will differentiate people is do you have IP that other people want?

33:51 that's not in the AI or if it's in the AI it's in an AI that's behind a firewall so you so you can't get to it it's the IP it's the absolute

34:00 Yeah.

34:00 IP so so that's that's that's how that's how we're working that's what makes us different and once I realized that um

34:10 It made us unlock so many things, right?

34:13 And I was talking to a company, we just entered into a partnership with a company out of South Africa.

34:21 And, you know, I just got back from South Africa beginning of this week and I came back with a signed contract with an exclusive partnership.

34:30 um to do work for them right and i said to her like we were presented at one client this morning and i said what we have to do is the client wants our ip and you can't you can't be at all the clients unless we create this solution this solution captures your ip

34:55 And then we're just going to sell it to the client and a client will push a button and two, three hours later,

35:01 we'll have the reports they need.

35:02 We don't need to we don't need to do that work and we just interact with the client to make sure that, you know, their AI investment is where it needs to be.

35:14 You've touched on something that I've thought about recently,

35:16 which is.

35:19 Right now, folks like you understand how these tools work,

35:25 how to create agents that can do things for people.

35:29 You've got, as you said,

35:31 IP or insight that others don't have.

35:34 I was talking to a friend of mine.

35:37 I guess it was two weeks ago we were having dinner.

35:40 He owns a coffee shop, actually has a few coffee shops, but he has some coffee shops.

35:45 And he has been using, I don't know if he uses Clot or ChatGPT or whatever,

35:50 to sort of summarize his financial reports and help him understand what's happening with his business.

35:57 he has sort of a CFO type person who would give him all this stack of reports and he could, it was always, they were always having to have these long conversations. And so he just fed it into, into the chat and said,

36:08 you know, help me understand this.

36:10 And I thought to myself,

36:11 well,

36:12 someone could create an agent or a skill to do that.

36:15 And then being who I am, I thought, and that could be a software product that you could sell to people, right?

36:23 Yep.

36:24 Which is what we're

36:25 So

36:25 doing.

36:25 we can sew that same thing over and over again.

36:27 But then I thought to myself,

36:29 what kind of, you know, moat or protection is there around that?

36:36 I guess what I'm asking is, it seems like maybe in five or 10 years,

36:41 probably closer to five, like a lot of these things are a lot of this knowledge is going to be out there, so to speak, kind of like open source software.

36:50 And I guess we keep,

36:53 now I'm going to answer, this is a terrible.

36:55 Question, Stephanie,

36:56 because now I want to propose an answer to it.

36:58 Please answer it because I want to hear your answer because I have an answer too, but I want to hear your answer.

37:03 I guess what's going to happen is the humans are always going to be learning a little bit ahead of the robots.

37:09 We're going to teach the robots a bunch of stuff that we know,

37:12 and then we're going to go out and learn more stuff.

37:14 Original learning is going to happen by the humans,

37:17 and we're going to feed it to the robots. So every company,

37:20 look.

37:21 QuickBooks is going to build their report summary agent into QuickBooks, right?

37:25 And then that's all going to be not that valuable of intellectual property anymore. But we're going to learn more as humans and we're going to then,

37:32 that's always,

37:32 the humans are always going to be the cutting edge. That's my answer.

37:35 You're in your team with your friend that's using chat GPT to summarize,

37:40 right?

37:41 He looks at his financials his way.

37:46 Yeah.

37:46 You look at your financials your way.

37:48 I look at my financials my way. You can't have the same agent.

37:55 It's what you

37:56 That's

37:57 have is you have the same base of an agent,

38:01 but you have three different agents behaving slightly different.

38:09 The AI that I keep telling people is because it does not take that much effort.

38:16 Once you create the base agent,

38:17 it's just an incremental amount of effort to create the agent to work the way you want it, the way your friend wants it, and the way I want it.

38:27 And we're now all happy.

38:29 Yeah, that's interesting.

38:30 Not everybody works the same way.

38:33 That makes sense too.

38:35 I mean, and that's that's been the biggest friction in corporate America from beginning of time.

38:40 You

38:40 Mm-hmm.

38:40 must fit into this box. If you don't fit into this box,

38:44 then you are a bad employee.

38:45 And if you're a bad employee,

38:47 depending on our level of tolerance will either.

38:51 keep you and marginalize you and then just make you a miserable person then do you just then in turn makes everybody a miserable person or we just fire you and we come up with and if we if we have no spine we'll just reorg your your um your position out the door if we have some spine we'll come back and we'll give you constructive feedback

39:12 Yeah, that's a very vicious way of thinking about it. I even think about I've worked at.

39:16 work with three different companies that all kind of do the same thing but they do it you know they do it different ways um and and you can sort of say well that's not that's not how we did it at this last company but i guess i can see why you guys do it that way so i guess we'll do it that way even though it's the same same business you know people choose to do it their own way

39:36 Exactly. So now all of a sudden, I mean, like, that's like,

39:39 that's where.

39:41 It come back to helping people be the best version of themselves because I no longer have to work the way Scott works because Scott's my boss.

39:52 Mm

39:52 I

39:52 -hmm.

39:52 can work the way Stephanie works because I have an AI agent that's helping me work the way Stephanie works.

39:59 That makes me be the best version of myself.

40:02 It builds self-agency. And because I am feeling good about myself,

40:06 I am loving Scott and everybody else on the floor and I'm getting my mom to bring cookies in for Scott and Scott's like,

40:13 oh my God, this is great.

40:15 Scott: Well,

40:15 that sounds wonderful.

40:17 And I feel like that's the note we should end on with one other small addendum. If people want to find you or want to find AI Buddy,

40:25 where should they go?

40:26 Stephanie: AvatarBuddy.ai,

40:28 or you can look us up on LinkedIn.

40:31 If you look for me,

40:33 Stephanie Sylvestre, T-R-E,

40:35 AvatarBuddy.

40:38 And if you mentioned that you saw me on this podcast,

40:42 I will happily give you a half an hour to chat about anything.

40:49 You want to learn about Africa burn?

40:51 Sure,

40:52 we'll talk about Africa burn.

40:55 Scott: Well, thank you so much for coming on today.

40:57 This has been really fascinating and frankly a little bit inspiring to me, which is good.

41:01 So I'm glad it's always inspiring to talk to you, but I'm a little bit motivated now to go try and figure out where there are some bottlenecks in our world so we can transfer tech on them.

41:10 Thank

41:10 Stephanie: Yeah.

41:10 Scott: you, Stephanie.

41:11 Appreciate it.

41:11 Stephanie: Thank you.

41:12 Scott: This episode of the How Many CTOs Does It Take podcast was produced and edited by Safran Heptakow.

41:18 Our theme song was AI generated by Brad.

41:22 Please follow us on all the social media platforms at HowManyCTOsPod.

41:27 You can find us on YouTube,

41:29 LinkedIn,

41:29 Instagram,

41:30 Facebook,

41:31 Threads, X,

41:31 TikTok,

41:32 and Patreon,

41:33 or at our website at HowManyCTOsPod.com.

41:36 Dot com and remember to rate review and subscribe on the podcast platform of your choice.

41:41 Thanks for listening