From Skynet to Cashnet: AI Profit Survival Guide for your Business - EP 017

00:00:00: Svetlana: Hello and welcome to the AI boardroom.

00:00:03: Today we'll be talking about how to calculate ROI for AI projects.

00:00:09: So we'll be sharing two methods with you for how you can figure out whether

00:00:14: an AI project is worth investing in.

00:00:17: And today, as usual, we have Edgar joining us.

00:00:21: There he is.

00:00:24: Edgar: Yeah thank you very much for introducing us.

00:00:26: Hello and welcome also from my side.

00:00:30: Whoo!

00:00:31: Been some time.

00:00:32: The last episode.

00:00:32: Svetlana: Yes, and please do

00:00:33: not dislike the intro.

00:00:34: We are really trying.

00:00:35: We're trying our best to spice things up.

00:00:38: Please let us know if you like this type of intro.

00:00:40: If you like us to spice things up a little bit.

00:00:43: But, yeah, it was a try.

00:00:44: Edgar: It

00:00:44: definitely helps to ask for If

00:00:50: Svetlana: you appreciate that, the spice up, please give us a thumbs

00:00:53: up because truly appreciate it.

00:00:55: We do it for you.

00:00:56: Edgar: Okay.

00:00:56: ROI was funny.

00:00:58: You came up with this and I think one or two people.

00:01:01: Two days later, I went through my LinkedIn feed and just had some conversation with

00:01:06: people like ROI working with AI, one was even saying I still don't get over it.

00:01:12: Like he was saying, AI is like raising all boats equally.

00:01:16: And I was like, What?

00:01:20: So if you take this argument, we don't need to do software and companies at all.

00:01:24: Everyone has access to software.

00:01:25: So where's the benefit?

00:01:27: Yeah, I tried to explain it to him.

00:01:29: And his only answer was, yeah, I hope wish you good luck with that.

00:01:33: Like I did like half, half a page of a message, explained everything, gave him

00:01:39: examples and he was like okay, screw you.

00:01:44: Svetlana: Best of luck with you and your AI.

00:01:46: Edgar: And I was like, yeah, thank you very

00:01:47: much.

00:01:49: The thing is, this, he was illegal.

00:01:53: And if I were illegal, I would be really worried.

00:01:57: Svetlana: Yeah.

00:01:58: Edgar: That

00:01:59: means that's for another day.

00:02:00: Svetlana: Yeah.

00:02:01: And I think we, just two perspectives because I think we

00:02:03: just kind chatted before starting.

00:02:05: So really I think that there's two situations for why, from what I'm

00:02:09: seeing is why people want to figure out what ROI is for certain projects.

00:02:16: Because one, they don't know how to calculate it.

00:02:18: So small firms and they have out of the box solutions.

00:02:21: They're looking for inspiration and they're looking in the industry and as

00:02:25: you mentioned before those kind of the use cases are quieting down because Probably

00:02:29: a lot of people are already adopting things So there's just no it's old news

00:02:33: at this point the out of the shelf and then the people who are implementing it.

00:02:37: They may sometimes you know lack talent in house.

00:02:41: So they pause projects, they don't take them and scale them across the enterprise.

00:02:46: So they never realized ROI and they've been burned by some of those solutions.

00:02:50: And yeah, there's just like the realization of those solutions.

00:02:54: But you did, you made a really interesting point.

00:02:56: And this is something where I operate on the daily.

00:02:59: You do have companies who are actually implementing AI as part of their

00:03:06: like core part of their operation.

00:03:08: So it's part of their intellectual property.

00:03:10: So it's in their best interest of not actually like disclosing

00:03:14: what efficiencies or sometimes how they're implementing things.

00:03:17: Cause that's going to invite questions and that's going to invite, other

00:03:21: things for people to inquire into.

00:03:22: So there's a lot of.

00:03:24: I could say, especially big companies that are doing with AI and they're

00:03:28: getting a lot of efficiencies.

00:03:30: So if you look at the Fortune 500 kind of reports on AI and that adoption, like

00:03:35: amazing numbers, like in the 60s, 70s of adoption, but you're not seeing that.

00:03:40: Edgar: You have to,

00:03:40: you have to be fair.

00:03:41: I've looked that up too.

00:03:43: I was curious, like we are so deep into this, like there has to be some

00:03:48: like I think let's say normal company or not like huge mega corporation

00:03:52: company who actually benefits from it.

00:03:55: Most of these are questionnaires to the people.

00:03:59: And the people saying like my productivity gain is like 70 percent of people say

00:04:03: that they gained a lot of productivity.

00:04:05: That's what the numbers say.

00:04:06: And I was like, okay, I get why people don't care.

00:04:10: Like you can't connect this to ROI because this is completely subjective.

00:04:13: So what's the objective number?

00:04:15: And one, one way I always try to at least picture it for someone is if you have

00:04:22: someone in sales and he makes 20 calls a day and he has to prepare these calls.

00:04:28: Now.

00:04:29: You get AI in helping him with the preparation.

00:04:34: Now he gets not 20, but 25 calls a day.

00:04:37: You have a 20 percent increase, like closing rates stays the same.

00:04:40: Everything stays the same.

00:04:41: He just gets more things, stuff done without doing anything else.

00:04:46: If you then just add maybe a better quality and also repeatable and

00:04:51: improvable system on vetting your leads.

00:04:54: Then you might increase your your overall closing rate too, and then it

00:05:00: compounds and you have these like strings of small gains, which compound to huge

00:05:06: gains where you have, sometimes you can double your sales or double your

00:05:10: Profit per invested dollar or something.

00:05:14: That's really where the magic is.

00:05:17: And I think it's, that's why I'm so reluctant to accept the copilot way.

00:05:24: Microsoft promotes is all the solutions you can build with the copilot are either

00:05:31: needing like huge expertise outside of copilot and then integrating it into it.

00:05:37: Or you're just stuck with insufficient tooling to actually get something useful

00:05:42: from it, except of reading documents.

00:05:44: Like it's hugely beneficial reading documents.

00:05:46: I think that's already a huge improvement for a lot of companies.

00:05:48: Don't get me wrong, but is it worth like 35 bucks a month per user?

00:05:53: I don't know.

00:05:55: Svetlana: So I think with solutions that are out of the box, I think Intuitively,

00:06:00: it's hard to use out of the box.

00:06:03: There is a learning curve that you have to adapt to those new tools.

00:06:06: You have to maybe sometimes customize or maybe retrain yourself to a new process.

00:06:11: And so you're losing efficiencies because of that learning curve.

00:06:14: So you're not able to maybe again demonstrate some of the promised are

00:06:17: why initially, because you do have to adjust your own press processing.

00:06:21: You have to adapt to those systems.

00:06:23: I think the unique aspect of building something that is completely customized

00:06:28: is that you can't keep your existing process if it works well, and you can

00:06:34: basically expedite certain aspects of your workflow with a custom solution.

00:06:40: And that's why I think for this reason, I think integrating solutions, especially

00:06:43: AI solutions into the workflow is so important because you're not losing

00:06:48: those kind of learning efficiencies.

00:06:49: You're not if the process works, there's no reason to break it.

00:06:53: But if it helps to bring velocity to that process, to bring more capacity,

00:06:59: Which AI can do that completely at scale.

00:07:02: Why not?

00:07:03: And I think this is where both of us kind of shine is by building those

00:07:07: AI, custom solutions to understand the workflow, understanding what works,

00:07:11: what is the unique aspect of the business that could be streamlined more.

00:07:15: It's not sometimes putting more people to actually increase capacity.

00:07:19: Edgar: Sometimes you can do that with technology.

00:07:21: Yeah, so I think also

00:07:22: highly.

00:07:23: Highly personalized.

00:07:24: I like no personalized, but like optimized for the company.

00:07:29: And that's why I'm always struggling with the copilot approach, which

00:07:33: says yeah, we like off the shelf solutions, because I think they

00:07:38: won't be able to deliver properly.

00:07:44: You have to add a lot of customizability, which I think is not the actual plan.

00:07:50: So yeah, do it from the get go properly.

00:07:52: And like you say, integrate in the workflow because what I was.

00:07:55: Finding myself coming back to over and over again is close to all off the shelf

00:08:02: solutions lose a lot of value because they're not integrated in everything else.

00:08:07: And because I cannot integrate them the way I need to, even the

00:08:11: workflow toolings I find myself avoiding again and again, I really,

00:09:02: All the off the shelf tools don't work for me because I need to jump

00:09:06: through extra stuff to get this work.

00:09:10: And it's not integrated into my workflow.

00:09:12: And that's why I'm over and over again, end up just doing my own

00:09:16: stuff and deploying that instead.

00:09:20: It's not that I don't want to, I'd love to have something where I can just

00:09:23: like quickly come up, sketch something.

00:09:26: I just, I would love to have my apple pencil you to use like

00:09:30: just to stretch down something.

00:09:32: And then I builds the workflow for me, for example.

00:09:35: And I'm pretty sure there are stuff like as API actually has this, you can type

00:09:39: into Zapier that's what I want to do.

00:09:40: And then I trust to build you up a workflow.

00:09:44: Yeah.

00:09:45: So if you don't want to do like the whole way, like the proper AI, then at least do

00:09:50: some workflows you can also, if you're on the Microsoft side, just power automate,

00:09:54: like the stuff is there's not a lot of magic in the in the overall like single

00:10:00: parts, but they have to work together.

00:10:03: And that's the other thing you won't get an ROI.

00:10:07: When you're not implementing stuff properly and break down problems properly.

00:10:13: To give you an example, if you like, I also talked to someone in finance

00:10:18: and she was like, yeah, in finance, like AI is delivers great results

00:10:22: and we are going to the future, but it's not working for finance.

00:10:24: And I was like, what exactly does not work for you?

00:10:27: I was like, Oh, what would be your wish?

00:10:30: And she was like, yeah, I would love to just, hand over a pile of data

00:10:34: and then ask do something with it.

00:10:37: And I was like, yeah, okay.

00:10:39: I get this.

00:10:40: And that's actually what gen AI is good for, because what gen AI can

00:10:45: do, and that's part of the solution design is break the stuff up.

00:10:50: If you get the question, I had this on stream.

00:10:52: I had like, how was the product launch?

00:10:55: That's the question.

00:10:58: It's in and of itself, it's ambiguous and can be anything.

00:11:01: If I asked someone on the street, he was like, he will be like,

00:11:05: what are you talking about?

00:11:06: But if you go ahead and break it down, you have different information in there.

00:11:11: You're like one information is okay.

00:11:13: We seems we have a new product means I can look up the marketing campaigns or our

00:11:17: product list, which is the newest product.

00:11:19: And then at least find out which product he means.

00:11:23: Then how.

00:11:24: Did it go means, okay, how were the sales?

00:11:28: How was the ad spend that got us to the sales?

00:11:31: Did we actually make a profit?

00:11:33: So that's all.

00:11:34: The questions you can get from it.

00:11:36: And that's something your AI has to do.

00:11:38: If you have an AI workflow, we have to, and you want ROI from it.

00:11:41: You have to do these steps.

00:11:43: I said it like 10 times on the stream.

00:11:45: You have to put in the work that's actually important because.

00:11:50: If you don't do it, then you come up with a lackluster system, which only

00:11:54: gets 60 percent of the work done, and you don't have any benefits.

00:11:57: If you want ROI, you need to get this number up and get to in the

00:12:01: 70s, 75s, 80s and from the 80s it's really valuable already.

00:12:05: But that's where you need to be.

00:12:07: to get.

00:12:08: And that's the only way how you can also reduce the consequences

00:12:14: from failing processes.

00:12:16: And there will be processes like it won't be 100 percent for the foreseeable future.

00:12:23: Svetlana: Yeah, I think.

00:12:24: And I think just to Maybe there's a common saying that says yeah, you can

00:12:28: add AI to a broken process and it's just going to get worse, basically.

00:12:33: So I think, and I think what you're saying, and I'll maybe caveat.

00:12:38: things in a different way.

00:12:39: So you can actually throw data at generative AI and just like very randomly

00:12:45: just ask for it to derive insights.

00:12:47: I think there's a use case for it, how useful it actually is

00:12:50: and how practical really depends.

00:12:53: But ultimately whatever you end up doing with it, you're putting

00:12:56: it through some sort of filter.

00:12:57: Whether you do that before you prompt the system or you actually do it after you see

00:13:02: all of the insights, you're still curating some part of that list, those insights

00:13:07: that are useful for a particular purpose.

00:13:09: So there's, and that kind of comes from education, how to use these systems.

00:13:13: Sometimes it's just better and more efficient to direct the system to

00:13:17: give you a specific outcome you're looking for instead of playing this

00:13:20: guessing game of what you what the A.

00:13:22: I.

00:13:23: System thinks you're looking for.

00:13:25: And very similar in.

00:13:26: You can't just throw all kinds of data and ask.

00:13:30: I think in your example me going to you and selling like, Hey, can you

00:13:34: give me, tell me how to sell a product?

00:13:37: Some of the best practices is going to be, you're going to be like,

00:13:39: is this a physical product or is this a digital product, right?

00:13:42: There's this new different nuances you have to specify and very similar with AI

00:13:46: systems, there's almost like some guide.

00:13:49: to how to extract the most value out of those systems.

00:13:53: And they are well aligned to the type of techniques, again, machine learning

00:13:57: knowledge graphs, they come with their own kind of specifications and best

00:14:01: practices for you to extract the value.

00:14:02: So of course, if you're not using them per their instructions, then

00:14:07: I'm going to give you the best results that you're looking for.

00:14:09: And you're going to blame the system for not working well.

00:14:11: Whereas, as you mentioned, you have to put in the work.

00:14:15: To actually read the manual and to actually use it appropriately and put

00:14:20: it in the right context, put it in the right workflow, contextualize it, and

00:14:24: then personalize the results if needed.

00:14:26: But garbage and garbage out, but you gotta have to put in the work.

00:14:30: Edgar: You need to understand this whole.

00:14:33: way that these things work is built for being generic.

00:14:38: Like it's generative, but it's also generic by design because you have

00:14:42: a huge, like I compared it to a cloud that's raining over the ocean,

00:14:46: that's AI and it's most simple state.

00:14:48: And if you only.

00:14:49: Ask like a simple question, which is completely ambiguous and don't have

00:14:52: any details and no context at all.

00:14:55: You only go out with your ship and you get wet, but you

00:14:58: don't even know where you are.

00:15:01: So you have to get this on land and you have to narrow down the cloud so that it

00:15:04: rains exactly where you want it to rain.

00:15:07: So that's basically how you have to imagine it.

00:15:09: And that's only achieved if you provide the right context.

00:15:13: And because we said it like, of course you can ask any question, but if

00:15:19: it's ambiguous, you have to design your AI, you have to design your AI

00:15:23: system in a way where it actually can ask back questions and is not

00:15:30: inclined to just generate anything.

00:15:32: And because that's going to fail most of the time.

00:15:36: So it has to be an interaction.

00:15:38: It has to be a conversation.

00:15:40: You wouldn't dare to employ someone new.

00:15:44: He comes in the first day and he has to know everything.

00:15:46: Like you also would give him some context so that he can work with.

00:15:50: And that's how you have to view it a bit more with AI sometimes.

00:15:55: Svetlana: And I think just to maybe summarize some of these

00:15:57: points, like why is this important?

00:15:59: I think that there is inherently a learning cycle.

00:16:02: They have to go through AI and part of it is setting realistic expectations.

00:16:07: You hear that a lot, but what does that mean?

00:16:09: You can't, it's not a plug and play solution that you plug in these A.

00:16:14: I.

00:16:14: Systems and then make them available, co pilot through everyone's computer and then

00:16:17: the month to see the immediate results.

00:16:20: They do take time to actually realize and we'll speak about kind of the

00:16:24: frameworks for what we came up with it.

00:16:27: You have to set realistic time frames.

00:16:29: So you're not going to see immediate returns in weeks or maybe even months

00:16:32: sometimes, depending on how complex the system, how long the cycle is for

00:16:37: you to realize some of those returns.

00:16:39: But you also have to take into account the learning curve of your employees actually

00:16:43: getting accustomed to that system.

00:16:45: So they do tend to, if you ever see a adoption curves, they to tend to.

00:16:49: be quite slow at the beginning, and then you get into this flywheel, which, the

00:16:56: more users, the more data, the more kind of better algorithm, better quality data.

00:17:02: And then you basically reach that exponential scale where you, your

00:17:05: system, your organization truly benefits.

00:17:08: But you have to go through that initial cycle.

00:17:10: I would say it's.

00:17:12: Common to all systems or all AI systems that there is a learning curve, whether

00:17:17: it's gen AI again, and some of the other solutions that I've mentioned,

00:17:21: Edgar: you already started.

00:17:22: So what we brought with us today are basically two methods

00:17:27: on actually calculating stuff, which we want to briefly go over.

00:17:32: If you want to have some more in depth information, how this works, we have some

00:17:38: reference to it in the end of the video.

00:17:41: Yeah.

00:17:41: But yeah,

00:17:42: As you are mostly the author of this

00:17:45: the stage is yours.

00:17:47: Svetlana: All right.

00:17:48: So there's really two methods as we talked about.

00:17:51: So one is called the comprehensive ROI.

00:17:55: Again you'll get the details of the step by step with examples.

00:17:58: And our comprehensive guide, we'll link it up link it down in the comments

00:18:02: or have a link I think at the end of the video that you could click.

00:18:06: So the comprehensive ROI, think about it this way.

00:18:09: So if you have, Some metric to go against, so maybe it's sales figures

00:18:15: that or maybe ROI assessments that other companies have shared commercially in

00:18:20: the past, and you've seen those costs cost savings, or maybe revenue increases

00:18:27: what was the third one in the, or the time savings, use that as an anchor

00:18:31: to your Basically, ROI calculation.

00:18:34: You have some baseline to compare your ROI results to.

00:18:38: With the second approach, which we're calling incremental ROI, you don't

00:18:42: necessarily have any sense of how much improvement you can gain with that system,

00:18:48: but what you do know is the value of ROI.

00:18:50: A percentage increase in a particular workflow again in revenues, the value

00:18:55: of adding an additional customer to your basically bottom line.

00:19:00: What does that cost you?

00:19:01: And you can work backwards, but estimating the cost of your I solution.

00:19:05: How much percentage increase in the current Revenues do you need to create

00:19:11: in order to for this ROI to make sense?

00:19:13: And so it's a quite different approach.

00:19:15: Once it's more influenced by established solutions and what you've heard, what

00:19:19: the potential of these AI solutions is, and then the other approach

00:19:23: is a little bit more bottom up.

00:19:24: So the key steps in the comprehensive ROI solution, that's the one

00:19:28: that you have a reference point.

00:19:29: You'll select the metric that matters.

00:19:32: So revenue is typically linked to some success metrics, whether

00:19:36: it's higher conversion rates.

00:19:38: More customers through the door, whatever have you.

00:19:41: So you link it to the to some specific metric.

00:19:45: You quantify that metric.

00:19:47: So you know how much money that's bringing in.

00:19:49: You do a cost estimation and you would figure out basically there's

00:19:53: an easy formula to a state of the formula to calculate the ROI.

00:19:59: So basically ultimately in sales, Lead scoring what you want to know.

00:20:06: So for example, if you want you to increase the number of leads

00:20:11: that are coming through your sales pipelines, you want to make sure

00:20:15: that you're paying attention to the customers that are responding to.

00:20:20: Your lead forms that have the highest propensity to buy.

00:20:23: So how do you know when you have a hundred leads, who do you call first?

00:20:26: And how do you prioritize?

00:20:28: If I was to implement an AI system and it told you ranked, gave you an ordered list

00:20:35: and prioritize list of people to call, and the matter of introducing that system

00:20:39: will increase your sales by let's say 30%.

00:20:43: Will that be impactful?

00:20:44: And would that be.

00:20:45: Worth and then I'll put a figure onto it that's going to cost you 20, 000.

00:20:50: Will that be of benefit to you?

00:20:52: And so it becomes an easy comparable to with some numbers that are well

00:20:56: established in the markets that are reachable by other customers and then

00:21:00: you compare it to the cost and you determine whether there's basically

00:21:04: a net positive value back to the organization and you consider that.

00:21:08: Basically a done deal or decide to go with a different solution

00:21:12: if it doesn't make sense.

00:21:14: But anything that I've missed there?

00:21:17: Edgar: No, pretty much sums it up.

00:21:20: I really liked the incremental ROI where you just say, okay let's just

00:21:24: assume 1 percent improvement where would we get already with that?

00:21:29: And, but also You can have different processes, like different start

00:21:33: starting points for your personal AI journey with your company.

00:21:38: you can say, okay, 1 percent here, or here, like which one will be

00:21:43: the most beneficial from the start.

00:21:45: So we start with that one.

00:21:47: Because that's what we usually when we work with companies and their

00:21:51: AI approaches, we try to pinpoint the one, two, three processes

00:21:56: that are actually impactful.

00:21:58: That's the strategy for the next 12 months and go for it.

00:22:01: Because yeah.

00:22:04: That's what we need for people not to go out and tell everyone there is no ROI.

00:22:11: Svetlana: And then it becomes like

00:22:12: a more of, I really that analogy because it becomes like an

00:22:14: apples to apples comparison.

00:22:16: So if you really, understand the value of a 1 percent improvement and

00:22:22: again, like a time saving, what is a minute in a specific or a 1 percent

00:22:27: improvement in a specific process worth?

00:22:30: What is a 1 percent improvement in cost savings worth in a particular process?

00:22:36: And then you can also prioritize projects.

00:22:38: I think that's a really great point that I think harder to do with a comprehensive

00:22:41: ROI when you're just again benchmarking against what else Other people are

00:22:45: doing in the industry and what results they're achieving, where you can truly

00:22:49: incur the incremental ROI and do a.

00:22:52: Almost like a comparison for what's more valuable for us to implement and again

00:22:58: There's some calculation that goes into it from a cost perspective because AI

00:23:02: systems do cost and ultimately You want to implement systems that give you a net ROI,

00:23:08: but it gives you an additional benefit that you can't compare Against and like

00:23:13: really do an apples to apples comparison.

00:23:16: One other benefit with the gradual improvements is that you can do

00:23:21: this for very complex projects where you do have a longer term horizon

00:23:26: for implementing these systems.

00:23:27: So let's say you can't, you think it's possible to 30 percent improvement,

00:23:34: but you can achieve that in year three.

00:23:36: So you're gonna, you probably are going to front load a lot of the benefits initially

00:23:41: and you're going to achieve incremental improvement over time, but then at the

00:23:45: same time, it's actually the effects are bigger because you reach that exponential

00:23:49: scale and flywheel effect, but your system is going to continue improving,

00:23:53: not as a result off further investment.

00:23:55: It's actually going to improve as a result of Further feedback.

00:23:58: And of course, there's gonna be cost maintenance.

00:24:00: But ultimately, you're projecting that net result over a longer duration of period.

00:24:05: And then you're promising some incremental improvement.

00:24:09: And again, even if it improves 1 percent percentage over the existing process, or

00:24:15: it's a net 1 percent positive over the amount invested, you're still winning.

00:24:20: So you're still, and you're actually going to reach higher scale and

00:24:23: more benefits over the longterm.

00:24:25: If you can commit to that path, you just have to pay, have the patience

00:24:28: with it of realizing that that timeline of reaching exponential scale.

00:24:33: So I would love to

00:24:35: Edgar: add one, one last thing.

00:24:36: And then I think we have it covered pretty well.

00:24:42: I've also read on LinkedIn, someone, some marketing agency owner.

00:24:47: He was like, I forbid my people using Chachapiti and because

00:24:51: results are not good and they are only good if I'm tinkering around

00:24:55: a lot and why should I do it?

00:24:56: And I was like, that doesn't make any sense because yeah, of course

00:25:00: you have to tinker up, tinker around.

00:25:02: That's what I say, put the work in, but then you have a repeatable

00:25:05: system which you can improve.

00:25:07: And I also always work with samples and rules that you hand over, people

00:25:15: using the system don't even see them, but they are in the background

00:25:18: and they can automatically improve.

00:25:24: By AI, the, I can improve itself without needing any training or

00:25:29: whatsoever just by updating its own rules based on the user gives a thumbs

00:25:34: up or thumbs down for the message.

00:25:37: So yeah that's something to keep in mind.

00:25:39: Like I said, the system will improve if you design it properly.

00:25:44: And that's why I always tell get yourself an expert.

00:25:48: The value you get from investing a bit more into your strategy up

00:25:52: front, will hugely improve any ROI you could have achieved without.

00:25:58: Because of course it's this is, you could argue this is a standpoint which is

00:26:03: beneficial for someone consulting on AI, but it's not making it less true because.

00:26:10: You need someone who really has experience, who really thinks in

00:26:15: the right direction, in the right path, to set you up properly.

00:26:20: Because when we are done, usually you can go and hand it to everyone who can

00:26:24: build even at least some AI application.

00:26:27: And he will give you, he will deliver you a proper system, but you have

00:26:32: to put in the work beforehand.

00:26:35: Svetlana: I would say I, it totally resonates with me and I

00:26:38: think it's maybe also historically has come Maybe the pitfall of A.

00:26:43: I.

00:26:43: M.

00:26:43: Y.

00:26:43: Again, there's so many failed projects is because people think that they can

00:26:47: figure things out without an expert.

00:26:49: Because they're seeing this as yet another I.

00:26:52: T.

00:26:53: Project.

00:26:53: So we'll figure it out.

00:26:55: Yeah, it's just, how difficult could it be?

00:26:58: Except I think, and you've built, you've been in the space for such a long time.

00:27:02: It's not an problem.

00:27:03: Easy thing to transition from just, let's say mobile or traditional

00:27:07: software development and then just jump into a I and be an expert.

00:27:11: It does take a really long time to actually learn this stuff.

00:27:15: And that's another reason why you can't promise any power systems.

00:27:19: Edgar: It's not only learning.

00:27:21: It's this whole mindset shift developers, your development team,

00:27:25: how it works today without AI is by design set up the wrong way by design.

00:27:34: You people think the wrong way because we programmers, we think sequentially, we

00:27:39: think in loops, we think in if statements, stuff like that, AI doesn't work like

00:27:45: that and you have to completely switch it.

00:27:48: You have to start from like, how would I build it if I only would use AI and then

00:27:53: you add the classical stuff to it and that's how you have to approach it and

00:27:57: this is like I said, that's why an expert is so important and so helpful if it's

00:28:01: a good expert, of course because he can bring in this mindset, teach your people,

00:28:06: bring them up to speed and enable you to do stuff you wouldn't be able without.

00:28:13: Svetlana: And I would say that's exactly the you're basically shortcutting

00:28:17: the time, the learning path.

00:28:19: So instead of figuring it out over a longterm, you get expert guidance

00:28:22: ultimately of walking you through that process and really distinguishing

00:28:26: the differences, but then also the upscaling and the training is so crucial.

00:28:31: But ultimately I do say that even though you're bringing in the expert,

00:28:34: hopefully you are upscaling your teams, internal teams over time.

00:28:39: That's like the whole benefit is not to be continuing to be handheld.

00:28:43: It's actually bringing that expert to actually like upscale and instill

00:28:47: that mindset into your folks.

00:28:49: If that's going to be a part of your culture going forward, because

00:28:52: you want your internal teams also to be upscaled and you want

00:28:55: them to adapt to that mindset.

00:28:58: From a trained expert.

00:28:59: So just want you to like, do not undervalue someone who is very experienced

00:29:05: in this space and bringing them in to help expedite some of the deployment

00:29:10: and realizing the ROI of your solutions.

00:29:14: Edgar: Yes.

00:29:16: Okay.

00:29:17: I think we are done for today.

00:29:19: It was interesting.

00:29:21: Yeah.

00:29:22: I think we will have to do some more content on that.

00:29:25: Interesting topic.

00:29:26: Hugely beneficial if you get the right mindset and you get the right

00:29:31: decisions and the right strategy in line, then you can reach benefits that

00:29:36: no other software project has reached for you in the last two decades.

00:29:40: And, but that's a big a big.

00:29:46: thing to just have in mind, the gap between prototype and scalable

00:29:50: product is as huge as it ever was.

00:29:54: Because we working with different systems with a different mindset, A

00:29:59: is not B anymore, and that's okay.

00:30:02: You can deal with it.

00:30:03: You just have to know how, and then you can get your returns.

00:30:10: Svetlana: Awesome.

00:30:10: Thank you so much for tuning in.

00:30:12: I hope you enjoyed this podcast.

00:30:15: Again, as we mentioned we prepared a guide for you this time.

00:30:18: We will include a link in the description.

00:30:21: For you to download it.

00:30:23: Again, we invested a lot of time actually putting together and, together with

00:30:28: Edgar to make sure that It's visually appealing, but then it's also, it gives

00:30:32: you the step by step guidance with examples for how to actually think

00:30:37: through each one of those methods.

00:30:39: So I honestly think it's gold.

00:30:41: It's probably one of the first guides in the most, most comprehensive guides we've

00:30:45: provided so cannot recommend it more.

00:30:49: So and if there's anything else, say again.

00:30:53: Yeah.

00:30:53: A lot of confidence.

00:30:54: But yeah, it is.

00:30:54: Of course.

00:30:55: Yeah.

00:30:55: It is good.

00:30:56: I

00:30:56: think it's, I think it's totally.

00:30:59: really proud of it.

00:31:00: Edgar: Give this to all your

00:31:00: Project leads and they we'll be able to, yeah, give you better

00:31:05: numbers to for your AI project.

00:31:09: Svetlana: We're just cross promoting it.

00:31:10: Yeah.

00:31:10: Yeah, we'd love your feedback if there's anything that we could improve or do

00:31:15: less of the intro that I did today.

00:31:18: Totally happy with it.

00:31:18: We'd love to hear your feedback.

00:31:20: So let us know.

00:31:21: Don't forget to subscribe and I guess we'll see you on the next one.

00:31:24: Edgar: Yeah.

00:31:25: Thank you.

00:31:25: Bye.

00:31:27: Bye.