Real-World AI Integration in SaaS

OK, we have our first viewers. So I think we live and we are on the requisite platforms. So hello and welcome to Momentum Officials. My name is Yash, and I'm joined by my co-founders Jay and Koushik to discuss topic of the week, which is the real world AI use cases in SAS. Our goal with these conversations is to provide you with actionable insights and practical strategies that you can apply to your own business.

Throughout the session, we encourage you to engage with us by asking questions and sharing your thoughts. This is a fantastic opportunity to learn from each other and gain new insights that can help drive your digital initiatives forward. Let's get started. Jay, Koushik, how are we doing today?

Doing great.

Thank you

Nice, and this is a part of a client meeting.

Yeah, this got out of a client meeting.

Yeah, so initial discovery phase kickstart meetings, we're trying to lay down base rules with respect to expectation and execution.

So this is honeymoon phase right for any project to start when you're just starting out. This is discovery which is where we get to be the expert and ask questions as to why have you not done this, why did you think of this, why have you not looked at these possible opportunities for increasing efficiency and so on and so forth. That ends when we actually have to do it right.

So we have the consultants for now, and then we'll become engineers a bit later. But awesome. So today we are going to talk about real-world AI integration into your SaaS. before we begin and have a conversation around this, I want to start off with a disclaimer. So no one is an expert in AI, because it and of itself is sort of growing and becoming better.

things that I came to today from six months ago are already outdated, and the AI landscape is filled with people making ridiculous claims of ROI as well, especially AI influencers. Actually, just as a matter of fact, yesterday FCC in the US, which is their agency that controls the marketing activities of companies.

And you would see them largely crack down on food companies and health care products and stuff like that who are making, you know, this really big claims like lose weight, no diabetes, whatever, whatever, have now cracked down on AI on companies making tall claims without any, you know, anything to back back up. And they are now saying that, hey, you know, you cannot make these world changing, universe denting, whatever claims with with AI. So

So you know, hold your horses of sorts. But yeah, so we are here to talk about real-world AI integration. And let's begin.

So yes, so yes, I started the basic one. So I'll sort of frame my question from where you just mentioned, right? Everyone wants to have it, but no one knows how to go about having it in the first place. You could be a business, you could be a product or anything. like, I'm a, so how do I kick start this process within my,

framework of my product. Like if I could be a product manager in a company or I could be the founder. So whoever is in the decision making, how do they go about kickstarting an AI initiative or an integration within their product?

So a way to think about it, and so there are different answers to this. So firstly, I'd sort of try and categorize as to what are the possibilities. So the first case is where you are thinking about starting a business, you identified a problem, and so you have the opportunity to build AI-native solutions. And we'll come to that in a bit. This conversation is largely for people who already have a SaaS product out there.

And they are trying to identify what, from the AI landscape that is unfolding right in front of us, what are some of the things that they can bring within their product. And so a good way to go about it is, essentially, if you have a SaaS product, are most likely used by a department within an organization. Maybe the finance team uses you, or the sales team, or the marketing team, or the C-suite.

whatever department or that group of people are using your product for, a good way to start is to talk to them and identify what other work are they getting outsourced, what other work are they getting extremely expensive consulting for, and what other work is causing some amount of repetitive tasks within the organization. So these are the three sort of

buckets, like what work is outsourced, what work are they paying extremely expensive consulting fees for, and what are the things that they are doing repetitively within that particular department or within that work. Once you make a list of these things, so talk to like 100 customers, 200 customers, and then once, and don't just talk to the decision makers, talk to the people who are actually doing it, because one of the things that we've realized in Koushik,

more than I, is that decision makers generally tend to have skewed way of looking at things. And so it's generally also better to talk to a person who's actually doing those things hands on. So talk to them, understand their journey and workflow, and then categorize them across these three. Most likely, there is a great opportunity within these three. And then identify what are the sort of ancillary

add-ons or opportunities that you can build for within your product. The other thing that I would also say is that within that conversation that you're having, you should also try to figure out whether they will pay for having those problems solved as well. A lot of times, like if you ask a customer, hey, know, do you want this, do you want A product or B product? The answer that the customer will always say is yes.

They always want more. So just make sure that while you are scoping things out and while you are working with your customers to understand their problem statements, you're also making sure that you're asking questions around whether they'll pay additional for having these problems solved for them. But that's a good way to kickstart.

So that's interesting. So let's say we are doing user interviews and getting a set of features from different set of people who are actually users. Now in this case, is there a way to differentiate what particular feature would be that cool AI feature and what would be the one which is high impact one from product owner standpoint? Because at times there are

certain features which may look very cool but then may not be that impactful. However, there could be some which are like actually impactful but right now the users will definitely see a value and will ask you for that but then maybe it is not that cool enough for all the set of users. So is there a way to gauge which one to go for first?

It largely depends on the strategy of the company. So you know for a fact that a cool feature is, and so the way that we define cool feature is something that your users or customers have never seen happen before. It could mean a lot of value. It could just mean a moment of excitement and not enough value over a period of time. But more often than not, a cool feature generally generates a lot of excitement.

and it may or may not deliver value later. And so if one of the focus areas for your organization is to acquire customers at this point of time, then a cool feature utilizing AI is a great way for the product team to contribute to your growth and marketing efforts. However, if the priority is not

like within these two, if the priority is not acquiring new customers but it is to retain and expand the accounts of existing customers then that also becomes a great way for your product team to build a feature that may not look cool that works sort of under the hood but just make sure that your existing customers are retained over a period of time and I'll give you an example for this. So an example for this is Canva

has this really valuable feature called background remover. And it's had background remover and it's become better over a period of time. But you can upload a photo and remove the background. You can upload a video and then remove background from the video as well. And that's the definition of a cool feature that will give value over an elongated period of time to your customers. The first time that I saw background remover, no one else was doing it. AI wasn't around. It was like four or five years ago, maybe even more.

And I was just extremely found it very valuable, useful, and it was a moment of excitement as well. Then we look at, let's say, another feature or a module, which is extremely valuable, but it doesn't seem cool. It doesn't seem exciting. So that I would say is, as an example, the meta ads, targeting controls, and how it makes decisions on who this ad is for.

So it doesn't look cool. If you look at the meta ads, the ad account dashboard, it's extremely boring. And you want to get out of it and you want to start living your life. Generally that's how the piece of software looks. However, if you're a business and if you've used meta ads and if you've done the targeting as correctly as possible, then it gives you value over a sustained period of time. So you know that meta, when it was building this platform, the ad account center,

Meta's priority was not to acquire customers or to generate excitement. Its priority was that they will come for sure, and then we want to offer them value over a period of time. So there are all of these. There's other examples also. another example of cool feature, which will not give sustained value, is the ability to create your digital avatar.

So there'll be platforms like Canva itself also has this possibility where I can create my own digital avatar from my profile pictures. So there's a balance that you want to maintain. If it's acquisition, you want to build cool features that can generate excitement, may or may not be of value. If your goal is retention, account expansions, then you can build features that will offer sustained period of value to your customers and they'll keep coming back for more.

And then there are times when you can catch lightning in a bottle which is where you are able to come up with something that is cool, generates excitement and then also offers value over lot of time. Hope that answers your question.

Yeah, that does. And also, guess if the acquisition is the prime key, then maybe also prioritizing features which a close competitor is also providing altogether. Yeah.

for sure. Absolutely. So a way that we would go about it is instead of a blanket looking at the comparator and looking at all the features that they have, you are not really... So a SaaS company, especially in its early stages, has certain amount of resource constraints. It doesn't matter how much capital they have and so on and so forth, but they have some amount of resource constraints.

So ideally, what we used to do was we would look at competitors, make a matrix of the features and modules that they have, and do it for five or six competitors. And then visit their G2 page or visit their software advice page. And then look at the reviews. Look at the top reviews. So look at five-star and four-star reviews to identify which features are offering most amount of value.

to our shared ideal customer profile. And then look at their negative reviews also and remove all the negativity from it. Just very objectively try and analyze as to which features are offering the least amount of value. It can start to feel very good if you read like a two-star review of a competitor. just remove that hat from your head.

But yeah, so that's generally a good way to figure out and prioritize which component features do you want to bring within your product.

So coming back to one point, so there is GenAI and then there is machine learning aspects that one could go about implementing within other AI integration. So now, first of all, I don't think people still have complete clarity upon the differentiation between Gen AI and machine learning. Could you go about explaining these concepts a little bit about what are these and how could one go about having it within?

as an integration within their SaaS product or in their upcoming SaaS product.

So I would actually argue that it's not important for them to learn and understand the difference between Gen.AI and machine learning. Because their customer doesn't care as to whether this is generative AI or machine learning or how many lines of code has been written or what data, what amount of the scale of data that you have trained on and stuff like that. So the customer essentially cares for efficiency gains. The customer essentially cares for access to new markets or growth in revenue or

So like, it's either increase your top line or improve your bottom line. So that's sort of a conversation. So I rather argue as to look at the use cases that you and your customers have and then leave the differentiation of Gen AI and machine learning to folks like us. So we understand the difference. We know the difference. We'll be able to use the right set of tech stack and right set of people.

and right set of data to be able to build and bring along solutions that are valuable. As a founder of a SaaS company, primarily your job role is to understand very, very deeply what your customers' problems are and what their life looks like. And then ideate along with your product folks or along with your other stakeholders, could be investors, could be team members. Ideate along with you as to what could possible solutions look

Execution of those solutions and implementation of those solutions requires you to develop expertise that is not valuable to growing and building an impactful and sort of a good SaaS product. So that's sort of where I am. However, I'd still sort of, given the question, I'd still sort of bring up that.

Another way, the thing that you should know are what are the capabilities of AI so that you are sort of able to bridge that gap when, whether it is through Gen. AI or machine learning or through natural language processing or through whatever, right? So whatever category of things that I'm seeing, you still want to know what are the capabilities that AI can sort of bring in so that you're able to bridge that gap between the problem statement that the customer is seeing and the solution that you want to imagine.

So a good way to think about it is that AI can see things. So as an example, one of the capabilities that AI has is that it can see things. can convert images or videos to text. And so there could be use cases around this. So a use case can be, as an example, if you build an APS, like an application tracking system, then the application tracking system can scan a resume.

to do requirement matching. So AI can see things. It can convert images or videos to text. AI can uncover insights. So it knows your market. It knows your competition. It knows your customers. It knows all of these things. And so given a large amount of data and some amount of context, it can uncover insights from. So as an example, insurance companies can use it to evaluate risk. Banks can calculate credit scores based on history.

and so on and so forth. The other thing is also AI can create and generate, which is generative AI. So it can create visual art. can create videos. It can create voice. And it can also create music. And so this is something that we sort of experience on a day-to-day basis, which is you have all of these SaaS companies helping you create short videos. All of these SaaS companies helping you create product demonstration videos.

You have presentations.ai or Tune, which helps you create presentations from scratch. GitHub Copilot can write code and so on and so forth. Then another one is that AI can hear things. It can understand speech and it can recognize voice. And so companies are using it to teach multiple languages. It can understand you and then it can tell you whether you are right or wrong and so on and so forth.

And then another thing is, of course, how AI started, which is it can undertake basic tasks. So think of your basic chatbots. Think of those basic journeys as to looking and understanding at your help center or help articles, and then figuring out as to what are the right answers to give to the customer. So that's sort of the gambit and the things that AI can do. Just to recap, AI can see things.

It can uncover insights, it can create and generate, it can hear things, and it can undertake basic tasks. So these are all the five capabilities which you can keep in your mind when you're talking and listening to the problem statement that your customers create.

Any more understanding of technology is essentially taking your company behind because it's very interesting and it's sort of like a snowball. So you just want to know more and more and more. It keeps you away from working on distribution and working on acquiring customers and retaining customers.

So this adds to one real in terms of based on where a SaaS founder is at or where the company is at. I had an interesting question. When it comes to an early stage SaaS companies or some SaaS companies which have already acquired certain amount of growth or a certain amount of user base and the one which are at enterprise level.

in these particular three categories, when it comes to bringing AI initiatives or AI integrations, be it any feature or any other angle, should they be going out for any consultant who is going to help it and get it implemented, or should they have tried to build an in-house team? Could you give some idea on what that direction should incline towards based on the stages?

So I'd say that if you're an extremely early stage company with little to no resources, then you want to invest your time and energy. Because if you don't have money, then you have time, right? And you have lots of it, right? So then you want to select no resources, wanting to solve a problem for very small part of the market.

then you want to spend and invest your time and energy into doing AI yourself. So trying to figure out how to go about it and doing it just yourself. I'm also writing a post on vibe coding, is the new thing, which is where anyone like a person like me, who's also non-tech person can start to code and how can you go about it. So that's the first answer. So no resources.

solving a problem for a small set of the market, you want to go about it yourself. Then for the second category, which is where you have some amount of customers who are already there. In those cases, as a founder, your energy is well spent on, again, deeply understanding the problem and ideating the solution. Let it up to a certain extent and you can decide what that certain extent is depending on the scale of your business and the amount that your customers are willing to pay for solution to that problem.

But up to a certain extent, you want to outsource it because you don't want to waste your time into experimenting and figuring out whether the technology works because the technology works. So there are other implementations that other people have done. And so you want to leverage the expertise of someone, get it to a certain point, which is where you know that, now it becomes repeatable and predictable and consistent, and then hand it over to an internal team who can maintain it and make it better over a period of time, slowly and gradually.

But you're zero to one. If there are lot of experiments that need to happen at zero to one, then you want to control as many variables as possible. So you want to get it done by an expert who's done this before. And at an enterprise scale, which is where you have enterprise customers, can have a bad end to end. I would argue that you want to have a third party do it.

Because if it's an internal initiative, one of the things that we've seen is the amount of ownership is always a question in larger organizations. So amount of ownership that an individual has is always a question in larger organizations. So we've seen that people play games on emails. We've seen that questions get answered by other questions. And the progress of the company becomes difficult.

So I'm not talking about exceptional top 1 % companies like Netflix or Google. Google has these problems as well, but so let's say Netflix. So I'm not talking about these top 1 % companies. If you're the top 1 % companies, by all means, just go ahead and do it yourself. But my guess is like 99 % of companies are not top 1 % companies. That's how you define top 1%. So if you're one of those companies, and you already know it, as a founder, you already know.

that a lot of things happen significantly slower than they should. And if you have that feeling, you want to completely outsource it, and not even just to one vendor. You want to have multi-vendor system where you're not dependent on one vendor. And again, you want to control as many variables as possible, and then take it forward. those are sort of how I would go about it if I were in those three situations.

I have a fun one. Could you give a...

We can have fun. I'm a very fun guy.

Anyway, have you seen any recent AI product? I'm pretty sure like a lot of which could have been a SaaS product previously and now they have started now they start to have some sort of AI capabilities in their product. Recent interesting product and you like the way how they have done the whole picture by bringing in AI also for the end user it feels like a quite a good value addition into the

Did you find any?

So because not everyone are able to crack that we have seen. So could you share about some.

Yeah, for sure. Loads and loads, right? So I'll share three examples. So one is I love how ClickUp has integrated AI into their product. And so when I say ClickUp, like think of Notion as well. And very, very strongly not Asana. Asana has done a horrible, horrible job of trying to bring AI. I tried it once and I got frustrated, right? So not Asana.

But ClickUp and Notion have done a phenomenal job of integrating and bringing AI. I'm able to get understanding. So if your team is using ClickUp actively and effectively, then you as a decision maker have answers to all the questions that you could ask. How many people are working on a particular project? What is the amount of allocation for people on different projects?

in terms of summarizing tasks, summarizing projects, how many tasks have not met timelines, whatever. So all of these things are extremely well done in ClickUp and even in Notion. So in Notion, what I've seen is most people use it for its generative capabilities, summarize this ABC or something like that. But it also has all of these capabilities where it can summarize a database and then give you insights on top.

So that's one. The second one that I would say is Canvas done a phenomenal job as well, which is where you are able to create your digital avatars and fill the scene or generate graphics and you can generate certain slides of presentations and write copies using AI and all of those things. And within that,

My use case for Canva is mostly for presentations. Within that category you have Presentations.ai and Gama as well who also done amazing job of off-bidding AI native presentation software. As an example, you use Gama or Tom or Presentations.ai, you will think that these companies didn't exist before AI came around.

And so they are just AI native, right? If you had AI, then what could be the interface for creating people's index? And the third example that I love is Riverside. So we use Riverside for our podcasts at Momentum. We had run a podcast called Building Momentum with an amazing host, Duncan Riley, who's our managing partner for Australia and New Zealand. so that podcast is sort of hosted, managed on Riverside.

What it does is from that conversation, it is able to extract subtitles, is able to captions, it is able to give us 10 options of possible interesting moments of one minute, one and a half minutes across the whole one hour or 45 minutes of recording. It also allows you to have some amount of editing capabilities and stuff like that. And then you can just directly post from it.

So that sort of also helps us accelerate our content pipeline, as I would say. But yeah, so these three use cases, right? All of them in some way or the other are use cases of machine learning and generative AI and put together, which is why I would say that, you know, don't go after the technology, right? So go after the use case and the mix of technology can always be brought in.

I think the reason why I asked the question was that if you're a competitor or any of the ones that he has just mentioned, you can go and check them out and see how exactly have they identified the right use case. That is one way of one easy way for you to land it how to go about finding the right use case.

Yeah, yeah, absolutely. Absolutely. Awesome. This sort of brings us to the... Sorry Jay, you were saying something?

No, no, I just agreed on the point. It makes sense. Yeah.

Yeah, okay. Got it. Yeah. Awesome. So this brings us to the end of the conversation today. There's loads to talk about for sure. We can keep on talking about this for hours and hours on end. If you are interested in talking about this for hours and hours, you'd be able to find the meeting booking link somewhere around my profile, or you can also visit our website and just submit a form or take out some time. can

understand what your product does and then sort of work with you around in terms of what are the places where maybe AI could come in and then go ahead and offer even more value to the customers. The only mistake to avoid is not to build a technology that is looking to solve a problem. So that it's like putting the cart ahead of the horse. So please don't build a technology that then

goes out there to find the problems to solve. Find the problems first, and we'll solve it for you. So those can always be done. Thank you, everyone, for joining in for this conversation. Hope this was valuable. And we will see you again in the next streams. And I forgot to mention that if you like this content, one of the things that we feel good about is that while we do

some amount of AI implementations and we build digital transformation and we do all of those things. We sort of live in our own paradise and think that we are creating a dent on the universe and we are making world-changing products and we have these initiatives. One of the things, and so there's already a lot of ego over there, but one of the things that still sort of, if you want to make our ego even bigger, please consider subscribing.

and commenting everywhere. that sort of boosts our self-confidence and gives us the incentive to keep on doing this week on week. Thank you again, everyone, for joining in. And until next time, bye.

Bye. See you.