Retention in the age of agents: becoming AI-native instead of AI-replaced

Hey folks. Uh this is David Abramson here, the CTO at Qrvey. And I'm going to be running through a presentation that I previously gave at the CPO Summit in New York City back in March. Uh where we're going to be talking about uh SaaS and the uh AI uh and more specifically how to become AI native instead of AI replaced. And a couple of things I just wanted to kind of first kick it off with talking a little bit about um Qrvey and and what we do. So, we are a fully multi-tenant embedded analytics solution designed specifically to embed into SaaS products and SaaS applications. Uh every single one of our implementations is part of a a SaaS product toolkit providing rich uh analytic outputs such as reports, dashboards, data visualizations, etc.

Um downstream to multi-tenant end user customers. And a couple of the key reasons why companies choose to work with us. Number one, our strong focus on self-service. So, really empowering customers to build their components instead of just offering, you know, pre-canned sets of reports and dashboards. Uh empowering your users and and and end customers to actually design their own uh outputs whether it's, you know, creating dashboards from scratch or even just modifying and editing the templates and components that you're offering to them. Uh everything is also fully embeddable, fully customizable, uh fits exactly with your product look and feel, uh and all of the branding. So, it doesn't feel like you're going into another product. It actually embeds and integrates seamlessly as part of your product toolkit.

Uh scaling is also really important with our embedded built-in data engine allows you to scale up to support you know not just uh the large volumes of data but also complex data structures and data models as well as bringing data together from multiple different types of data sources databases as well as semi and unstructured data so you can transform enrich and make the data as analytic ready as possible. Also our multi-tenant architecture so fully designed to support you know your use cases for your your tenant workspaces security entitlements etc. And then Qrvey is also fully deployed solution so even though we work exclusively with SaaS companies we ourselves do not operate as a SaaS we deploy and embed directly into your cloud environments so you get to not only own and manage all of the security and data as well as the infrastructure and upgrades and everything that goes

along with making it as compatible with your SaaS product life cycle as possible. So let's jump into our topic where we're talking about kind of the AI readiness of SaaS products and platforms. One of the things that's kind of really important to think about today is how users are increasingly starting to use AI tools as the interface. These are exploding right everybody's going to Gemini or Claude or open AI and that's the prompt that's the that's the front door to not only searching for information but asking good questions analyzing data doing things day-to-day tasks automating things those are becoming the interfaces and part of the challenge there is the the risk rather for SaaS companies is AI then just becomes the front door right um, your UI that gets bypassed and increasingly the end user loyalty shifts to the AI layer that um, those gateways,

those prompts, those interfaces for um, not only just day-to-day operational tasks, but potentially um, tasks that your software would normally be um, accomplishing for your customers. And so, I guess you know, fundamentally the question is if the AI agents can provide enough of those capabilities to your to your customers, do they really need to use your product or your interface to to accomplish what their their goals are? Um, and existentially this this poses a risk because we've built our products as SaaS companies um, to make them very sort of AI-friendly, right? Um, APIs and data and structured data models and um, tools and automated rules and things like that that be very easily orchestrated. So, um, you don't want to risk your product becoming bypassed um, in favor of all of these these AI tools. And uh, again, the data's centralized, the

APIs are exposed, your security models and architectures are probably already fairly AI-friendly. Um, so it makes you kind of an easy easy target um, for these types of of shifts in user adoption and user loyalty. And again, as I just mentioned, um, SaaS platforms are typically already designed to support this use case, right? We've got structured data. We've got multi-tenant security rules and logic. We've got APIs that are great interfaces to do all, you know, a variety of different tasks within our products uh, and potentially also integrated with workflow and automation uh, engines. And so, it's a recipe for um, UI shifts um, and basically we we provide all the ingredients for what these AI agents need and users can leverage them to do the things that they want to do and potentially we lose that stickiness and the fact that they're not going to be in our platform as much.

However, this also does pose some very interesting opportunities that are brought forth to leverage that AI. So, instead of thinking about the AI as competition, right? So, instead of thinking that you're losing your customer to these interfaces, think about how you can incorporate that into how you want to offer your product capabilities and feature sets. Um and a couple of different ways that this can be accomplished, not just from a data perspective through things like embedded analytic capabilities, but also through MCP integrations, agentic type features, um natural language or conversational user interfaces, um and also exposing more steps kind of in your multi-tenant security process so that you you still own and manage the governance of data and how you want to work with the information through your um SaaS offering and through the tools that you're offering with your SaaS

product. And probably one of the best ways to accomplish this is through the MCP server type interface. Um not only does this enable you to create structured, secure, and you know, define access to all of the tools that you want to expose to the AI agents. Um it also gives you the ability to um create the rules and fine-grained permissions and the multi-tenant logic and all of the different things that you need to do to manage exactly how you want AI agents communicating with your technology stack. Um and so there's a lot of benefits that can be had for your SaaS platform when using these types of AI-ready capabilities, whether it's MCP or even just more compelling and powerful APIs. Um you're really making it AI-accessible by design. Um you still maintain all the control um and also it can open up some

interesting opportunities for um new offerings, whether that's new products, new feature sets, or even new monetization opportunities that you can expose to your customers through these these types of interfaces. So, instead of being queried randomly by those AI agents, um you own the relationship, you provide the official interface, the governance, and also the monetization opportunities to work with these AI tools in the right way. Um and so when we think about how we go about building this into our products, obviously it's it's good to think about it from sort of a phased approach. Um this can either start with just basic AI augmentation. Um maybe it's providing some embedded analytic capabilities with AI tools. Maybe that includes some of the full self-service features, some smart insights where users can ask questions inside of conversational interfaces, but ultimately um exposing

tools like the MCP server to clients um gives you a way to define exactly those AI boundaries. It also gives you some opportunities to now start thinking about how you would monetize end-user consumption of your features, of your capabilities through the external AI services as well. Um and then once you start doing that, you can maybe think about how you would incorporate additional UI. Maybe it's more agentic interfaces and workflows, some of the automation tools, um, some integrations into maybe even other third-party services and their MCP interfaces, um, and then as well as kind of more detailed specialization, um, connecting other external agents, connecting more AI features, um, and adding them into your application. All of these gives you opportunities to expand and grow your AI footprint, um, and and really own the AI relationship between your product and your customer.

Now, if all this sounds like a very daunting task, I think it's important to understand that there are tools out there that can give you a jump start. So, finding the right partners, whether it's folks like us at Qrvey AI or others, um, who can help you get to market faster with some of these capabilities, and then ultimately start to realize that there are, um, compelling monetization opportunities to be had. Um, and now we're moving away from maybe just basic, you know, user licensing models or user-based pricing to potentially some usage-based models, AI usage-based, query-based, agent-based, um, maybe some premium AI feature-based tiers, or maybe even some uh, MCP access fees. So, a lot of flexibility, um, to to start thinking about ways to offer not only new layers, new tiers, but also, um, new models for how you could choose to monetize some of these very rich and powerful AI capabilities, um, beyond your sort of traditional, um, subscription models that you're already

working with. Uh, and so again, thinking about how you can transform building these new capabilities into very successful monetization opportunities and profit centers, um, across, you know, all of your users and tenants within your application models. So, um, thinking about it now, you know, the real competition might not even be other SaaS products. Um, you know, could be thinking about how those AI platforms can be taking attention away from your tools, but, um, as you've seen, there are ways that you can incorporate that and really leverage that. So, instead of competing with them, think about how you can complement or integrate with them, um, through those different types of tools like MCP interfaces or, uh, additional agentic capabilities that you would present and create as part of your SaaS product offerings.

Um, so, it's all about figuring out ways to strengthen not just your product, but the relationship you have with your users, um, so that you don't worry about being replaced by, um, kind of the AI agents and AI, um, user interfaces, and don't become bypassed, um, by the users who are increasingly moving to those for, um, you know, the way that they want to work, uh, on a day-to-day basis. So, um, let's take a look. Uh, wanted to kind of walk through some ideas and examples of how this can work, um, certainly in the context of something like embedded analytics, um, you know, through, um, Qrvey's, uh, MCP style integration. So, we're going to jump into a quick demo, um, and and take a look at how, um, this type of logic can work. Um, so, the first, um, part of the demonstration, um, that we're going to work off of is, um, starting from, um, the external MCP client type interface. So, you know, think about the

gateway into working with data, um, instead of your application can become something like just your Claude interface, where users can ask questions. Um so for example, if I'm connected and I'm using my application, um you know, maybe I just want to see um what data I have access to. And again, you can um communicate directly um with the service and you can see it's it's going through and checking what I have access to based on my tenant security rules and logic uh and searching what types of tools I have access to work with um you know, from the different uh environments uh you know, that we have.

And so all of this can be done 100% through um the external um interfaces um you know, that that we have. All right, let's take a look at the data that I found within my KubeAI environment. So, found that I have some data sets to work with um fully loaded

and ready for analysis and it's asking me what I want to do with this. Um so perhaps, you know, I I I just want to communicate with it Um uh Yeah, maybe I I want to see some um some key metrics um out of some of my data within the current environment. So, it can go through and and retrieve the data. It can query it. Um it can calculate some metrics. Um it can go through and and build some aggregations. And so, it's going to go through and and identify, you know, finance metrics, demographics, activities, um you know, different level of details um that we have. And um you know, you know, maybe I want to see you know, what it looks like you know, comparing

um activity trends year over year. Um and so, it's going to query the information that's there and give me, you know, um level of detail. And and understand kind of all of the the the detail that we have within the data. And as you can see, it's going through and um again, sort of analyzing the information, um understanding things like filters and um you know, different level of details and metrics. Um so, we can see uh you know, different pieces of information um that it's generating. And all of this, again, can be done 100% um within the external interfaces um that we have. So, summaries of of year-over-year activity, key observations, um do you want it to drill deeper, you know, maybe you want to do some comparisons. And again, um leveraging this interface to interact with the prepared data that you have

within your application offers users a very compelling way to consume and work with the data and the details. Now, um we can also do this from um our other interfaces as well. Um again, jumping into um just kind of the the application interface directly, tools like our AI assistant, we can also navigate and work with data directly from here. And maybe I maybe I just want to um summarize some data quickly. I can use the AI assistant to do that, um analyze the results. Um it can go through and not only provide some high-level summaries, but aggregation queries.

Um and again, now what we're doing is we're we're really leveraging the MCP tools and custom interfaces um to leverage anything we need about the data. So, highlights, um you know, key metrics, um demographic data, and again, being able to leverage things like key takeaways as part of um this logic that we're working with. Um and so, you can see all of the different rules and and and capabilities. You know, what are some of the um Maybe I just want to ask, you know, what should I be paying attention to?

All right, so we can go through and get the breakdown. >> >> And it can summarize all of that detail. And through these summaries, I can also then um again, look at some red flags, look at some prioritizations, and then I can have it even do things like build a dashboard. Yes. You know, with the key key metrics. And so, not only can you communicate with it, get answers quickly, get recommendations, get key takeaways, you can have it go through and build a comprehensive set of actionable reports and dashboards to analyze and you know, kind of create the interface for for what you're working with. So, a ton of flexibility um to, you know, create the indicators, create the charts,

um create the dashboards, and build that sort of compelling outputs that you're working with. So, again, going through and and and leveraging all those steps in sort of the um fully prompt-based way gives users a ton of flexibility, whether it's through external tools or um the embedded tools of the analytics solution with Qrvey AI you know, that we're working on. And so, you can see all the components, the charts that are being generated, the KPIs, the different tools that we have, all of that is, you know, part of the logic that we're working with as um you know, inside of the the different interfaces that we have there as well.

Um so, a ton of flexibility to to do the types of things that you need, you know, through these models, and then ultimately generate the the appropriate outputs that you want to work with. Um so, that's kind of a a high-level view of how you incorporate MCP interfaces and AI into your SaaS applications and products. Um and again, um thinking about how you go about launching this into your product, um thinking about what it looks like to partner with solutions like Qrvey AI, ready to integrate tools immediately to get to market um faster and start offering these technologies and capabilities out to your customers so you can focus on the rest of your differentiators as well.

So bottom line, you know, AI is not necessarily going to replace SaaS, but you becoming AI native SaaS um will definitely be the winner when it comes to offering these types of capabilities and um winning the market share for your customers and keeping them engaged with your product offerings um as well. All right, thank you for your time and and taking a look at our presentation. Um check us out curveai.com um to learn more and to um communicate with us and um you know, get a custom demo or just learn more about how Qrvey AI can help um launch these AI native capabilities um directly into your SaaS products. Take care everybody. Thank you.