Transcript for Episode 2:

Michael: Welcome to the Disambiguation Podcast, where each week we try to remove some of the confusion around AI and business automation by talking to experts across a broad spectrum of business use cases and supporting technology. I'm your host, Michael Fauscette. If you're new to the show, we release a new episode every Friday and as a podcast on all the major podcast channels, and then also as a video on YouTube.

And then we post a transcript on the Arion Research blog and of course on the on the player on the stream. In your show today, we'll take a look at what AI Solutions innovations are available from the Salesforce ecosystem, from the app exchange to help with sales and marketing. I'm excited to, to talk to, and I'm joined today by Jakub Stefaniak.

Jakub is the VP of Technology Strategy and Innovation at Aquiva Labs, a top tier Salesforce services partner. He's a Salesforce certified technical architect and has a professional certification as a CTO from MIT as well as several professional certifications from Harvard and Stanford. Jakub hosts the podcast AI and Innovation meets app AppExchange.

So let's bring Jacob in.

Jakub: Hello. Hey. Hello. Welcome. Thank you for the invitation.

Michael: Thank you for joining. I really appreciate it, and I appreciate that you can share some things with the audience today that I think they'll find really exciting. There's a lot of interesting things happening around Salesforce and the ecosystem and ai.

So why don't we start though by just tell me a little bit about Aquiva Labs and what you do there.

Jakub: Sure. Aquiva Labs is quite unique company because we are part of program called Product Development Outsourcers, and it's a super small niche in Salesforce ecosystem when we are providing consulting services not only to Salesforce customers, but as well to independent software vendors. So company who publish the application on app exchange Marketplace. And it put us in a very nice spot because in many cases, ISVs. Are not only developing the applications, but as well thinking how to utilize new technologies like generative AI and different kind of the ai. So on the one hand, our customers are very tech savvy and like much more willing to innovate in a typical sales customer.

But on the other hand, we as well work with. Salesforce direct customers. So we are helping them to make more let's say mature decisions about what is the kind state of technology and the fact that we are doing this all research and development for core of our PDO business give us as well very nice connection with the rest of the ecosystem.

And talking about my role in the company. A few months ago, I left my lovely engineering department in European Union, which I was building for the last two years. And together with our CEO Nikolai, we started something, what we called AI Task Force. So we really, we have a very small group when we are thanked to answer a few questions.

One is getting all this AI knowledge, which then we can help to transfer and adapt to our customers. And the second is leading the digital transformation inside of our company. Because with the new wave of artificial intelligence, we see lots of potential to improve our own development services. So we are helping our engineering department to adapt to the new processes, new technologies, and re-imagining how the outsourcing company can work in 21st century.

Michael: Nice. I should say that I met Jacob through an association with Aquavit Labs. They're doing some work for one of the startups that I work with LocatorX. Definitely excited and we certainly had a really good relationship with you guys. And I think you've certainly done a lot to help them get their product on the app exchange.

Let's jump into the AI piece of it. So there's so much buzz out there particularly over the last year or so, AI, GenAI, business automation, all of those things seem to have finally come together in in a way that have gotten people really excited. And like with any new technology or.

I guess not really that new, but newly available or at least people are aware of it, there's a lot of excitement. So I'm curious cause you talk with a lot of customers around using AI, implementing AI, what they should be doing. So what are you hearing, what are customers doing with AI today? And then what are some of their concerns and fears that might be, maybe holding them back a bit?

Jakub: That's amazing question and I think we should take a look on the history of computer science because it's not the first time when people get excited about AI, probably in seventies and then in eighties you have this first big waves, and then we have a term winter of AI because of this super high expectations business start putting lots of money into doing stuff just to find that technology is not good enough. Yet in nineties when IBM's deeply won Chess match with Kasparov people as well like thought, okay, tomorrow is going to change the world. The theory is that we had a pretty good AI for playing one particular game, but it was not enough to generate the business value in this particular moment.

I think with the current level of hype, it's good to ask exactly the same question. Where do we exactly stand and how could we benefit from this current situation in the market? When people are thinking about ai, I think the very confusing part is that, okay, we have some technology which help to discover new drugs and fight with a cancer.

We have some technologies for super good quality image recognition, and we have ChatGPT, which can help with, for example, getting sentiment and classifying text. So for many customers it's very difficult to even think, okay. Where should we start? And then one of these options can cost me hundreds of dollars or thousands of dollars per month, but another is going to cost me millions of dollars.

But as a not technical person, I can have literally lots of problems to even think, okay, maybe solving cancer is a good plan before my business. But then I will find that this technology is not so easy as just using generative ai.

Michael: So it sounds there's levels of jumping in on this technology, right?

So the, there are things you could just simply automate using very simple applications of generative AI. But then the more you go up in the chain, either from an automation standpoint or from a, from an assistant, data analysis and decision support tool. It can get considerably more expensive and more complex.

Is that's accurate?

Jakub: Yeah, that's correct.

Michael: So as you're talking with these customers and obviously a lot of Salesforce customers there and Salesforce has gone all in on AI with the AI cloud and, Einstein GPT, and, a bunch of different offerings that are coming out or have come out.

Is there, what's the discussion like? I usually when you talk about implementing the technology, there's like a build versus buy versus should we partner? How does that play out in the decision making today for Salesforce customers?

Jakub: That's a great question. Even if we are thinking only about option to buy, Lots of people struggle to make decision what to buy.

What are the real options available to me and the big concerns, for example, the data quality. If I know that I have 10 years of technical debt with my database, is it like good enough to get any benefits from AI or not? Then tricky question is how much is it going to cost me with even AI? APIs we are paying per every API call, but then my CFO would like to know more or less how much we are going to send them at end of the month.

And doing this calculation is not revealed because we have 10 concept of tokens, which are like words, but not exactly have words. So most of our direct customers are coming with exactly the situation that. They feel that something is going on. I really like the metaphor. There's exactly, really a moment when First iPhones premieres occurs.

Many things which are now obvious, like delivery routine. The Uber are obvious to us now, but like before this specific moment in time, there was not possible even think about business models. So we see customers exploring these options, but not sure where to start. On the ISV part of the business. When our customers are a software development company or product companies, it's a little bit different.

So we have lots of customers who have specific use cases in mind, and then I'm thinking what is the best option to make it in the secure and business efficient way, because now we have to keep in mind that if you put AI as part of your product, you're taking responsibility for it. So you probably want to really understand what kind of the level of risk bias and so on you're going to struggle with.

Michael: Yeah. That, that's certainly been a big concern. I think. And I just published recently a study on AI adoption and you could see there were a lot of there's a lot of fear around certain parts of it. One of, one of the biggest things that came out was concern around what they call data accuracy.

I also slipped the hallucination term in there, and I think people didn't necessarily know what I meant by it. A lot of people don't know what I meant by hallucination anyway but the data accuracy, hallucination, whatever you want to call it, is definitely a concern. One of the, one of the limiting factors is data and data cleanliness.

And like I said recently, that survey we did, we saw about 40% of the respondents listed data quality as their number one concern. Related to data prep and. As we know, CRM has been is a bit infamous about data quality or poor data quality, I should say, especially around sales and sales force automation.

Have you seen anything and I should say, in your podcast, obviously you talk to a lot of Salesforce ISVs. And so I'm just curious, what have you heard, what solutions have you seen that can really help accelerate your data cleanliness, data quality?

Jakub: So one pretty nice and simple to adapt use case, which came to my mind is application called Alfred, created by a company native video.

What is going to help with if you have situation that your salespeople are working in the field? They usually struggle to find time to fill all details from the meeting on the opportunity object and so on. So Alfred allows you to record your voice and based on the transcription of your voice, use generative AI to fill the summary of what you set on the level of the record.

And then like in this way, interesting is taking matter of seconds to have a meaningful summary in your salesforce org. So it's pretty good first step and it's, I think something that is going to start happening more and quite often because on the upper exchange ecosystem. There are more than 10 different companies providing some kind of the either calling solution, something related to voice, let's say.

And at this moment, none of them have this generative AI capabilities, but implementing them sounds to me like a natural next step. And now it's a place when Aquiva Labs can help with, because the tricky part is that then if you have, let's say, one hour meeting and you would like to make transcription of this meeting and then make a recording.

You have quite lots of tokens slash words to proceed. So if you're going to pay out $1 per making summary of this meeting, it can be quite expensive app. But then external knowhow allow you to strongly reduce the number of tokens. So price of your API cost, and then as a product company's much easier to feed your cost into your final financing because you can deliver lots of value to your customers, but in the same time, it's not so expensive for you to develop these capabilities.

Michael: So on your podcast you talked to a lot of AppExchange ISVs and I know you've seen a lot of interesting approaches to using AI and, kind of innovation around AI. What are some of the things that you've looked at on the show? What are some interesting solutions and applications of AI in the ecosystem?

Jakub: We have a full spectrum. So one of our customers cover is very big company, they have more than 700 people doing proper research and development in field of AI. So for them, we are mostly helping them to with their app exchange product, which is integrating a very complex AI event product out of their system.

And I would say it's a good example that. If you would really want to do something very innovative, very novel, and be a company building and delivering AI, this scale of 700 technical people doing the research is a good example. What kind of the investment do you should, do you expect? On other side of the spectrum, we have fair example, capital at other company.

Very nice small product with which is leaving eight bit out of the Salesforce core platform because Salesforce, per se, is usually sales and service. Cap Steel is OEM partner, which means that they're not connected to the standard sales process. They have their own platform dedicated to travel industry, and what they're doing as travel industry is not, technology.

They don't want to build this AI capabilities, but they see that, for example, for onboarding of new users today, system, if there will be some AI driven capabilities, they can strongly reduce time of the onboarding. So reduce costs, and then it's going to open them, open new business. So for companies like that, for example, we are going hand with hand and helping them to participate in a pilot program from Salesforce, which is allowing to start using Salesforce as a provider of AI technology because then on the level of trust and marketing for on customers, fact that you are not using external system, but everything is coming back different.

Salesforce as well can be a competitive advantage.

Michael: That sort of makes me think of another piece of this that I think is interesting. And actually, I just wrote something about platforms and services the other day, and I'm curious how is this, Salesforce went into this with a strategy of being open and also of making all the services available through the platform.

And I'm curious, has that accelerated the efforts for the ISVs, are you seeing a lot of adoption around that in use cases around Einstein and AI cloud in the ISV ecosystem?

Jakub: The tricky part with Salesforce is they have many different AI products. Some of them I have available from quite a few years, and we are speaking about.

Einstein discovery predictions, this and so on. These things are generally available for quite a long time, so I don't think there is lots of a new excitement, but there are like some companies try using this feature, right? On the other hand, we see more and more announcement about. AI Cloud, I believe it's already outdated term.

I, at least for the info speakers, we get notification that we should not use AI cloud anymore. We are going back to the Einstein terminology can be tweaky, but as part of it Salesforce is building their trust layer. The layer of integration with AI. So for ISVs s it's going to be a decision, okay, do we want to build this in integration on our own or do we want to pay for part Salesforce middleware layer?

But then it's going to help us to reduce some complexity, and then it's not real decision, it's just a business decision. I do have some costs. You have some benefits, and you have to do your math.

Michael: Yeah, that makes sense. So I know one of the, as I talk to Salesforce customers, one of the things that I've heard over and over, and frankly a couple companies that I've worked for that had Salesforce is that it can be a bit of a challenge to administer the to administer your organization properly.

It takes some training and, certainly. There's the trailblazer trailhead all the, paths that they have for learning and all. But nonetheless, it seems like AI could perhaps provide some assistance in that making the Salesforce admin job a bit more scalable and efficient.

Have you seen anything in the ecosystem that you think could help Salesforce admins with this?

Jakub: Sure. So that, Two products, which came to my mind. One is Elements Cloud. They have pretty nice tool, which currently thanks to generative AI is literally allowing you to start chatting with your metadata and for example, then if you would like to find all pieces of the system related to some business case.

It can help making assumption that your descriptions in your metadata are good. So if you did your homework, you filled the description fields, you have a meaningful API names, then the results are going to be very good. If you are using NICE API names like Field One, field Two, field three, or for example, you are using the same field for different use cases, then good luck because AI won't be able to help with this and probably your new employees are stacking exactly the same results.

And quite similar project, which we are building internally in Aquiva. It's early stage, so we are calling it Aquiva, GPT, but for sure before it will be market ready. We are going to change the name. It's a tool helping with onboarding of new developers. So it's focusing on the. Chatting with your classes and as the good use case, we have a situation when, for one of customers who is like pretending with us to build this, they have like hundreds of classes, hundreds of thousands of lines of code.

And for a new developer, if you would like to write a simple function, for example, to verify email address. There is other class in the system which is helping with this. But to go through all existing code base for a new person is not realistic. So they have lots of duplicated code and with our new capabilities, we are providing the product, which are the developer to firstly ask existing code base.

How in this sales earth we are solving this specific problem. And then thanks to this, they know, is it like a new problem or something, what they can use from existing set of tools and coding standards.

Michael: So it's almost like a map to what you already have so that you then can either leverage what you have or augment that or implement something new.

If you don't have anything that, that helps that's really interesting. I know there are a lot of companies that do build custom apps and extensions and that sort of thing on the Salesforce platform. So I imagine that would be very interesting, particularly in the idea of getting new developers up to speed really quickly.

And then also preventing a bit of the collision that can happen if you write something that's not that doesn't take into account. The previous iterations of what's been developed and done there. So yeah, that makes a lot of sense. I, one of the areas that I've done a lot of work in lately is around this idea of customer journeys.

And I know, journey maps have been a thing for many years, but I. It's always bothered me that it's a very static kind of activity. You build a map and then customers are supposed to somehow know they should do that. And as we know, they don't do that. They do whatever they want to do.

So it seems to me that there's a lot of opportunity. Leveraging AI to do what I like to call intelligent personalization. So you know, some way to make that customer journey more interactive and automated. So I'm curious, have you seen anything that could help provide a better customer experience, make it more, personalized or interactive?

Jakub: So luckily the answer is yes. One of our customers Coveo, their product can help with something like that. And is this example with 700 smart people thinking how to do it and building this product from many years. So I would say, I totally understand your pain, but it's not trivial to solve it, and it's like a very good example when if you struggle with this, the probably buying existing solution is going to be much more cost effective than trying to build something in-house because it's not trivial, unfortunately.

Michael: I know part of the problem is just simply understanding whatever clues or data you get from the customer that helps you understand where they are and what the context is. How can you make the interaction relevant? So it sounds Coveo’s a solution that could at least help.

Companies improve that capability. And it sounds like it's something that and I've looked at this a little bit before too, and I do agree. The technology to do this and the idea of putting behavior. Individual behavior in a context that helps a system understand what you're doing and then serve up something that's that, an experience that's relevant to whatever it is that you're doing.

That's not a trivial solution at all. That, that makes sense. Anything else around that though that you've seen? I know that's a, it's a big issue for a lot of, a lot of customers

Jakub: So maybe I can just give some story how it happened that we started to work with this specific company.

Because for us as product development of source partner, the very nice place that we really, in the middle of this new ideas coming out of the Salesforce ecosystem going into AppExchange and then through to Salesforce customers. And in this particular case as PDO we participate sometimes in something called co-innovation.

So in this case, even not the ISV itself is our customer, but Salesforce account manager, see that they tackle with, for example, some technical problems or maybe with adoption in the new market or like something would Salesforce see as a potential to solve and partner with PDO, so we can focus on the specific problem, which is like very well defined.

To help the ISV adapt to the app exchange reality and in this case, The very nice output is that at the end we not only solve the customer problem, but we are getting this very nice understanding of what kind of problem they're solving. Plus we can help other ISVs to go in the same direction.

What is quite important to notice that we discussion about current wave of AI, of generative AI? My recommendation would be to do not start with such big plans, and the reason why is that if as a company would like to start building something around AI, it's very similar to all discussions about digital transformations, which happened like 10 years ago.

You should think about some easy wins, some low hanging foods to build momentum. If you will start with building something too complex, you can just find that after two years you're going to be out funds, your business colleagues are not going still see any benefits, and then your program can be terminated.

So probably for now, the native AI sounds like the place and in many different types of business, you can find these low hanging foods, which can immediately give you some value. Yeah,

Michael: that interesting. I, in a former iteration of my tech career, I ran large ERP implementation projects back in the late nineties and early two thousands.

And that, that was always a real challenge. 'cause I felt like customers wanted to get, such an acceleration to the value, to the end value to the vision they had of what this would do, that a lot of times they didn't understand. And one of the things we tried to help them in is to break everything down into steps so that you could realize value early.

And continue to generate value with each of those phases, versus I'm investing everything for three years and then eventually I'm gonna see some return out of this and it's gonna be wonderful. It, it seems like that is from a business standpoint not a good strategy.

Jakub: It. It is never a good strategy to have like super high expectation at the beginning when you don't really understand what are these technical possibilities.

So my recommendation would be to start from a little bit different angle and they're like, really at this moment, two options. One. That you start with educating yourself, what are the capabilities? The tricky part with doing that on your own is fact that AI landscape is evolving extremely fast, like Italy, few hours per week.

For me, it's just following newsletters and checking the current state of technology. Yeah. The reason why it's happening is fact that even if you make some research three months ago can be already outdated. Yeah. For even generative AI, we have few different like language models. Of course everybody knows about ChatGPT from OpenAI, but Google has Bard, Meta published their LAMA and so on.

If you want to start using AI, some of them are more expensive, some of them are less. But for your specific cases, maybe some of them are not as good. Like we made, for example, for one of our projects, a comparison, how this generative AI works for specific questions, which we ask and we see that there.

Quality of the responses vary, but depressing as well is different. And after a few months, what we are now doing, we have the automatic job, which is like sending the same prompt when we know what should we expect as the answers on the regular basis. And we see that even if we didn't touch anything in our app, The performance is changing and the quality is changing.

So it really, either, as a customer, you have to establish something as our Aquiva AI task force and have somebody at least doing it halftime to try to stay off top of curve. Or the second option is to partner with somebody who is doing exactly this and can provide this value. And for us, to have some texture on this.

We started with just getting some theoretical knowledge. So our CEO Nico, I spent some time on business school, university of Navarra in Spain on a program artificial intelligence for executives. And because he's more business guy, as I'm a little bit more technical, he sent me a to MIT to learn about machine learning.

So we really, when we have this. Complex requirements to build complex system. We as well have the right level of technical knowledge to start this journey for Salesforce customer. I would expect that 10 years from now it's going to be part of a chief information officer, Chief Digital Officer. So generic this capabilities, which now sounds I get black magic are going to be as common as having like good mobile strategy or social media strategy. But it's a future for now. Probably just by thing with somebody who is working on this can be like the best first step.

Michael: Yeah. It's funny, one of my previous guests one, one of her conditions for doing the show was that I had to publish that episode really close to the interview because she felt like things changed so often that she had been embarrassed a few times in other shows she had done that were, released two months after or whatever.

And things had completely shifted in the in the AI landscape, which as we know you really do have to keep up with this weekly, right? It's a challenge. So we're getting close to the end of time, but I wanna make sure, one, one of the things that I really wanna make sure listeners can take away from this is, what advice do you have? You're doing this with a lot of different companies, ISVs, end user, customers, what advice do you have for companies that want to jump into this AI enabled solution space particularly for Salesforce customers but in general too. What would you want them to look at?

Jakub: So I think the most important part is that Bloomberg has essentially make some prediction that for big companies, like in this year, the typical budget for AI research was between one to 2%, and in the next year is going to be between 10 to 20%. My feeling is that with this kind wave, we should always keep in mind that.

AI is related to data. So if you start using AI early, you get more and more data, and this data can give you competitive advantage. So in terms of AI driven digital transformation, it's still more than effort that it's always better to lead than to catch up. Because if you have 10 different companies doing quite the same and one of them is going to be this early adopter who is start getting lots of value from the new technology, then of course they can deliver it is very to their customers and to catch up is going to be much more difficult.

My. The biggest recommendation, biggest advice, is to start now, and I'm not saying start building now, but start thinking, start exploring, start asking great questions and think how to prepare for all these new opportunities that are going to occur.

Michael: Yeah. My guest last week when I asked her for advice for marketing teams around AI, she used the old Nike slogan “just do it”.

It's like you gotta get your feet wet, you gotta get in, you gotta get your hands in it to learn. And if you aren't, then you're just falling behind. I fully agree. Yeah. So this has been really fun. This is all the time we have for today. Jacob I really wanna thank you for joining me and before I let you go I, one question I always like to ask a guest is, could, can you recommend somebody, a thought leader, an author, a mentor that has influenced your career, particularly around ai?

Jakub: So I'm just checking the title of the book. There is the one professor from MIT who wrote a book called Life 3.0 Max Tegmark, and he is like not, he's not speaking about generative AI, rather on AI in general. But it's opening a very nice perspective, and I will strongly recommend it to all people who is in this podcast.

Michael: Great. Thank you. Yeah that's great. I will try to put that in the show notes as well. And I will put a link to your podcast in the show notes so that people can check it out. It's, really interesting. I listened to a few episodes over the last few days just to get a feel for what you were looking at. And it goes through and talks to leaders from many of the ISVs in the in this, in the AppExchange channel that help you really understand what innovations they're bringing to market. So I wanna thank you all for joining us this week. Remember, hit that subscribe button. And for more on AI, you can check out an Arion Research report that we just published on AI adoptions based on a survey that we did in August. So it's very recent. It's on the site as a free download. Can't beat free research, so you should go check that out and definitely join us next week as we look at the use of AI and content creation for marketing. I'm Michael Fauscette, and this is the Disambiguation Podcast.