Applied AI: Turning Data Into Actionable Insights

Let’s face it, data is pouring into your organization.

In the time it took me to say that line, your company’s modern data stack probably took in two terabytes of data.

It’s kind of scary.

You know what’s not pouring into your organization though?

It’s insights that tell you what to do with all that data.

But don’t worry, that’s what we’re here for today.

Hi, I’m Courtney Baker and this is AI Knowhow from Knownwell, helping you re-imagine your business in the AI era.

As always, I’m joined by Knownwell CEO David DeWolf, Chief Product and Technology Officer Mohan Rao, and NordLite CEO Pete Buer.

We also have a discussion with Arman Eshraghi of Qrvey about the embedded analytics platform his team is building and where AI fits in.

But first, spring is right around the corner, thank goodness.

So it’s never too early for another installment of AI in the Wild.

Pete Buer joins us, as always, to break down some of the latest AI news and how it applies to your business.

Hey Pete, how are you?

I’m good Courtney, how are you doing?

I’m doing good.

This week, I’m curious to hear your take on OpenAI’s deep research tool, which Casey Newton of Platformer says just might be the first good agent.

Pete, what should all the business leaders who are listening now know?

Thanks Courtney.

This is going to take a minute, but I think it’s worth taking a minute because the explanation of this tool gives us a real feel for the differential usefulness of agentic AI.

So if you’ll bear with, starting with the basics, it’s available at least for starters to subscribers at a rate of $200 a month.

In the ChatGPT pro tier offering, users are limited to 100 deep research queries per month, kind of reflecting the high computational costs that are involved.

To use it, you type in your query like you would normally do in ChatGPT, but then you hit the deep research button and a very different process flows.

First, deep research asks you follow up questions to clarify the deep research angles that it will then pursue.

Second, it shares its chain of thoughts in answering your question.

You can see the websites it’s visiting, what conclusions it’s drawing from those visits, and how their reasoning is coming together for the story that you’ll receive in the end.

Then third, the output, unlike a one paragraph, one page, fast response from typical ChatGPT is a several thousand word in-depth report.

Very cool.

And you can start imagining how $200 a month might be worth it when you contemplate offsetting the costs of an analyst on a consulting team doing that same kind of work for you.

What’s the catch?

Well, there’s always a catch.

While it’s deep and thorough, it’s not without error.

The author gives a kind of an ironic example where the ChatGPT tool declares that it’s offering preceded Google’s launch of a similar product, of the same name, by the way, Deep Research, when in fact it was Google that came first.

So perhaps counterintuitively, the author concludes, it’s kind of better to go deep on topics where you actually have some foundational knowledge, so that you can sort of fact check the deep research that’s going on to the extent that anyone cares about fact checking anymore.

Consensus though, in this article and others, is that this is one of the first publicly available AI agents that’s really kind of quite good.

And I would just say, remember, this is just the beginning, right?

This is the first tip of a huge iceberg that we can’t see much of under the water.

Wow, that’s really cool.

And I’m already, my wills are spinning on how this gets applied.

I’m curious, Pete, are there some obvious ones to you?

You know, is this kind of, hey, if you’re a marketing person, if you’re in charge of educational content for a business, you know, where are the like immediate use cases?

I mean, you just nailed it.

Like, if you scan across the business, all the different places where deep research is used, by the way, it’s a lot of the business.

And by the way, it’s also a lot of employees.

Those start to be the places where you can get leverage.

And I think the challenge for businesses is how do you use it to make your people more awesome as opposed to, you know, draw a line through your people.

Pete, thank you as always.

Thank you, Courtney.

If data is the new oil, as this thing has been going on for the last decade or so, what kind of engine should it power for your business?

I chatted with David and Mohan about just that topic.

David, Mohan.

Hey Courtney.

Serious business here, okay?

Serious.

Executives today, too much data.

Too much.

A wash with data.

It’s everywhere.

Wash with data.

Data we don’t need to operate, right?

So like random facts, like did everybody know that Courtney was a cheerleader?

Hey, no, I don’t tell people that.

I know.

That’s why I just told them.

Just examples, you know?

Well, I can’t say what I was just about to say.

Okay.

Actionable insights, intelligence on the data?

Not so much.

So how can AI help bridge the gap between data overload and companies?

Too much data, data everywhere.

I don’t have time to go get in the weeds of all the data to find out what I need to find.

And actual insights to help me actually take action.

Again, too much data for me to like weed through it all to figure out the nuggets that I need to go actually take action on.

So can you help us break that down?

And maybe, yeah, giving us some like definitions on the two sides of that bridge.

I think where I’d start that conversation is understanding when you say the word data, when you say action, like I think a lot of these words are tossed around a lot of time without meaning.

And so, you know, one of the things we’ve done and really grappled with at Knownwell is understanding what is the anatomy of intelligence, right?

Of knowledge, right?

These big words that we throw around.

And at the very, very bottom of the pyramid is data.

And when we talk about data, we’re talking about raw, unprocessed facts, figures, symbols, observations.

They don’t have any context.

They don’t have meaning assigned to them.

They are just fundamental building blocks of information is the way I like to think about it.

And that is data.

When you say data overload, that’s what I think of.

People have way too much…

How many BI tools are out there that are taking these data warehouses and just throwing data at people?

Like random facts and they have no meaning.

They have no context.

That’s why when you get a good executive in a room and they’re looking at charts, like they care about is it a line chart or a bar chart?

And do you have a trend line?

And do you have what good looks like and badly?

Those visualizations are so important because they’re visual tools that help the individual turn data into information that actually tells them something.

Can I add a little caveat onto this?

Because I think you’re exactly right.

But I think there are other data that I’m including.

Like, for example, we make this joke all the time on this podcast, we’ve all got note takers now for all of our meetings, and we all get them in our inboxes, and that is data.

That’s right.

But we’re not using it.

To be totally honest, I couldn’t tell you how many times I opened those meeting notes after the meeting.

I totally agree with that, and it’s how we’re expanding this data set, where we used to only mean managed data, we used to mean like what’s in a database somewhere.

Exactly.

I totally agree with you.

That is part of the raw data that sits out there without any meaning assigned to it.

I basically want to say that if you don’t know what you’re looking for, nothing’s going to help you.

So Courtney, if that’s the starting place of data, and I totally agree with you, it’s not just like a piece of singular mathematical data in a database somewhere.

I totally agree with you.

It now includes all this raw content out there.

It’s unprocessed facts.

The second layer of the hierarchy here is information.

This is when data has been organized, it’s been structured, it’s been interpreted to provide context, right?

And so we can start to apply meaning to it, all right?

And here’s where, again, your point of emails, transcripts, all this other information we have is still raw data, right?

With everything in a database, we’ve started to organize it and make sense of it, and so it’s become information that we throw at people.

But in this world of AI, when we’re looking at all of the enterprise data, 85, 90% isn’t contextualized, and it has no meaning.

Okay, so that’s information.

When we start to put meaning to it, it has context.

Then the next step is knowledge, which is the accumulation of this understanding, right?

It’s the assimilation of this information through experience, through education, through rational deduction and reasoning.

We start to develop the next level, which is understanding, right?

So knowledge, and then we start to internalize that knowledge, right?

We assimilate it, then we internalize it, we start to understand.

This is where it starts to become human, right?

We can actually have an understanding of something, not just process it, right?

Then if you continue along the human path, you’ve got wisdom after that, right?

Now, this is becoming…

I’m putting judgment on top of that, and I’m applying knowledge and understanding of that knowledge, combining it with insights, ethical considerations, all of those other human things, right?

But when we’re talking about this, I would point us back to data, information and knowledge.

And then what does the AI do?

It’s the intelligence that’s able to find the data, the information and the knowledge at the right time when we need it, to be able to process it in a way where we can have understanding and it can propose things to us but we’re the ones that ultimately have the wisdom to make judicious decisions.

And so it can use all of this knowledge to rationalize it down to here’s what quote I recommend, but we still need to have the human in the loop to say, does that make sense?

Can we intersect that with all these other contextual domains to apply that knowledge to a situation and intersect it with the ethical, the moral to make sound good decisions?

David, I think the framework you mentioned, I think makes a lot of sense.

But ultimately, I go back to if the leader does not have a vision of what it is that they want, nothing can help you, right?

So you got to know how to run your business.

Sometimes I’ll go to these small grocery stores, many of them are ethnic grocery stores, and they’re wonderful business people, right?

And I always wonder, they didn’t go to any fancy schools, and how are they such great people business-wise?

And that’s because they know whatever it is the most important thing for them, cash flow or margin on unit goods, or whatever it is that they’ve figured out in their business that works is what they’re optimized on, and they don’t even know these fancy words.

They just know how to operate their business, right?

So they are looking for these signals.

So you need to know that as the leader of your business of what it is that you’re looking for.

And then you can use AI to go through the data, the information stack, and get the right things for you in a timely way, and you can prompt the LLMs or build ML models to predict.

You can do all of that, but it starts with having a vision of what it is that you really need to run your business.

Mohan, I think I agree with that, but I need clarification on something, because I would love to understand this more.

I think that while you need to know what you’re looking for, one of the advantages of AI is it can also tell you what you don’t even know to ask about.

How do you reconcile those two truths together?

You just had to look at it as two sides of the book.

I mean, there is an analysis of the data that gives you and illuminates options.

And you can use the AI in that fashion.

So that is one way to do it, where there’s a recommender or an evaluator or whatever it is.

That allows you to, you’re just sort of using it as a co-pilot to just brainstorm and think through.

That’s the scenario that you painted.

And then there is the production, which is running this, getting these flash reports every day or weekly reports or monthly reports.

That is much more of an automation of what it is that you’re looking for.

So yeah, you’re right, you can use it in multiple ways.

One of the hardest things with analytics in general is to figure out the anomalies, all right?

So if you know to look for differentials, that’s great.

You can kind of put them in two columns and subtract.

But if you don’t know what are the, what is like different this month from previous month, and you didn’t structure it that way, you could ask AI to say, how is this different from the previous month?

And be able to find this exception reports that you never thought of in the first place.

And that could be a really wonderful use case for AI.

Well, I think for everybody out there listening, if you’re twitching a bit because you feel this pain, this too much data, the research really backs it up.

Forrester suggests that between 60 and 73%, very specific there, of all enterprise data is never analyzed, just sitting out there.

Layer on top of that, much of the natural communications that we use, the example of our transcripts for our meetings, isn’t even considered part of the modern data stack.

Another recent Accenture study found that just 27% of executives get actionable insights from their data and analytic projects.

Kind of painful there.

So I think the takeaway here for everybody listening is, there’s a lot of hope here for AI and how we may be able to really have something thinking for us, how we should be thinking about all of this data, which is hopefully very encouraging if you’re feeling triggered by this episode.

Well, David, Mohan, thank you as always.

Appreciate it.

Losing a valuable client who you thought was in good standing is a total gut punch, but it happens all the time in the services world.

But here’s a little newsflash.

It actually doesn’t have to be that way.

The warning signals are there.

You just need something that can help you spot them amid everything else you have on your plate already.

That’s why we’ve created Knownwell to deliver real-time insights into things like service quality perception and commercial alignment.

Say goodbye to guesswork and managing based off subjective information.

Visit knownwell.com to learn more and let us show you your data in the Knownwell platform.

We’ll show you how we can take all of the things happening in your business and turn it into actionable insight to drive your commercial intelligence.

Visit knownwell.com.

Arman Eshraghi is the founding CEO of Qrvey, a company that provides an embedded analytics solution for SaaS companies.

He sat down with Pete Buer recently to talk about how companies can tap into technology like AI to make sense of all the data at their disposal.

Arman, thank you so much for being with us here today.

We’re excited to get to know you and learn from you.

Thank you very much for having me.

My pleasure.

May I ask the favor to get us started?

Can you give us a little bit of background on Qrvey?

I hope I’m saying that right, as a company and where AI fits in?

Sure.

Qrvey is embedded analytics for SaaS.

If I had only three words or maybe three words and a half to describe it.

And what it means, it means that we provide the analytic capabilities that can be infused and embedded into SaaS products.

So in other words, we provide this technology to other software companies so they can add it to their product and reach their product from an analytic perspective, and then they can go to market faster.

And the connection to AI?

So we are on the embedded side and on the analytics side on both sides, we are very connected with AI in two different ways.

We are embedded, so we help also software companies to embed AI capabilities.

And also, as you know, the main value you get out of AI is when you have quality data.

And also, you need to add some AI capabilities on top of that data platform that you build.

The main problem that we solve for these companies is solving their data problem, right?

So they essentially have this data that they are building and then feeding the data into analytics.

And I see AI as capabilities, extended capabilities for analytics.

It’s true that for some businesses, AI might be the whole foundation.

But for us, AI is extended analytics, right?

So that’s the way we see AI serving the data.

We have listeners across a spectrum of industries and business models and familiarity with data practices and AI.

So if I may, I’d like to just start with the basics.

When you say embedded analytics, what do you mean and why is it important?

There are two different categories, I would say.

One category when it comes to data analytics is you use it for your internal use cases.

Meaning that you have some people in your organization and the organization has some internal needs with regard to reporting, dashboarding, automating, alert management, subscribing to data.

There are hundreds of these teeny tiny functionalities I can list.

All of those is categorized in my thinking as analytics.

So I simplify that into one word.

And then you address those needs for your internal usage.

Or it might be an external use of these technologies.

Meaning that you are not really developing or creating or offering this for your own people.

You are offering it for your customers.

And that is either you can name it external use cases or you can name it multi-tenancy use cases just because you are serving multiple tenants outside or external, whatever you name it.

That is where you need to really build a product application serving a lot of customers and their users.

And in that purpose, you need to embed something and you need to really make it part of your product.

We are expert and specialized in those multi-tenancy use cases.

And that’s how we define embedded.

Increasingly, we have data everywhere in the business, whether it be around internal or external multi-tenant applications.

Assuming like everything else in life, finding your best use case and making hay with it starts with some strategic thinking.

How should companies, how should leaders go about considering all the multitude of data at their fingertips to leverage for optimizing the business?

Yeah, so it needs to start from business needs, honestly.

Sometimes I have seen so many companies get trapped into really the technology side without understanding the business side of it and what is the value you get.

For example, if there’s an organization that does not yet understand the value of the data being sourced from so many different variety of data sources coming to them, or as a product, as a SaaS application, I don’t see the value of really providing that multiple sources of data being available to my users.

First, they need to figure out the value of that side before they go and just think about what technologies I should use to do that.

But if that value is clear to them and they know that this is highly valuable for their users to get access to the data and to these data sources, the next question normally comes in as, again, in a multi-tenancy world, it’s more about, I’m serving so many different clients and their users on my side, that at any given time, I’m adding another customer, another set of users, and I wanted to give them something that makes them successful with their data journey.

It’s impossible for me to know all the needs of everybody in the future.

The best approach they can take is to offer maximum flexibility and customizability and self-service, so the customers coming are not handcuffed by a certain functionality that they have and that’s it.

That’s the best approach a SaaS application, a SaaS company can take with regard to the data to say, I’m going to really offer more extensibility for my customer, more extension, more customizability, self-service, everything that makes them happy and empowered without necessarily I be the bottleneck and I have 500 customers, I can not just go one on one with each customer and ask about the requirements and implement it.

That’s the big difference between the multi-tenancy and embedded world versus the single tenancy and all of the internal applications on the analytics side that you see like Tableau and Looker and those kinds of applications.

So those are more tailored toward internal organization that you have a certain number of people with very well known use cases, and you sit down with them and you develop whatever dashboard, whatever works for them and then you give them what they need.

Versus in the embedded world, that everything can change on the fly, but everything needs to be driven by APIs and everything needs to be built on the fly based on just the data that you receive and everything needs to be self-service manner, and your customer has to really make some decision, not just you as a vendor.

It’s fascinating.

And you’re getting into the space of my next question.

Let’s pretend I’m a CEO of a SaaS company.

I’ve got a trove of data I want to be able to leverage for multi-tenant customer facing applications.

And I listened to this podcast and I’m trying to imagine what the work is going to look like of doing this project or implementation, whatever noun you would associate with it.

Can you give me the high level overview of what the steps are and what it looks like?

Sure.

So we normally take a very pragmatic course of actions that essentially helps these companies not to wait for month and month and month, even sometimes a year.

We encourage them to go through these phases.

So phase one can be as easy as, I’m connecting to my data and providing a simple view of my data that still is very interactive, and customers and users can really go there and just play with this data in a variety of ways.

That can be accomplished in a matter of, sometimes we have seen weeks, but maybe a couple of months, depending on the organization and the data structure they have.

All of that, by the way, is more really all about the data.

It’s not about the building, the visualization normally is the easy part, if your data is solid and you have the right foundation and structure.

But you decide on the way you want your data pipeline to be, and the way you want to serve the data, and then you go with phase one and make it as simple as possible, because that will give you the entry point to really bring that kind of capabilities to in front of your users.

From there, then they can take face-by-face approach to go to phase two, phase three, phase four.

The ultimate journey, the ultimate goal for them is to really, as I said, provide the maximum flexibility to their customers and be able to define maybe which customer needs to have the self-service data management versus which customers should not have the self-service data management part, but they need to have the self-service data visualization part.

Or maybe self-service automation should be available to premium customers.

It also goes back to, again, on the business side, how do you want to monetize it?

How do you want to create a kind of revenue stream, added revenue stream that comes from the analytics side or more advanced analytics?

So, it’s not just a decision of engineers make on the technical side and go and just say, this is the way we do it.

It has a lot to do also with the product strategy, with the business strategy on the revenue side, how do you want to architect it?

So, those discussions need to happen.

And then a vision will be formed.

And then the vision trickles down, at the very, very bottom, it goes to some engineering steps that are the easiest sometimes to really just take them.

But all of them will go toward the bigger goal that this is really on the top.

This is what we wanted to accomplish in order to, in the next year, we wanted to add 20% to the revenue of the company and provide these capabilities to our customers and added another analytic product.

And in one way or the other, even win over the competition and all of those strategic goals that needs to be clear first before we go into the widths and details.

Critics of SaaS would point out the shortcomings around custom data cuts and custom analyses, the things that have to be done either by paying the SaaS provider’s services team an extra fee to get to the real answers that they’re looking for or assign resources on their own team to do the extra work.

Is the embedded analytics powered by AI solution that I’m starting to understand in your context, is that sort of the answer to fulfilling the promise of SaaS at the end of the day?

That’s true.

The way I personally see SaaS, it’s really the major characteristics of SaaS.

If you wanted to really, really simplify it, oversimplify it even, would be in the subscription kind of revenue stream model or in any way in a usage-based model.

So a different revenue model that SaaS brings to the table compared to the old-fashioned perpetual license model that software does react.

So that’s one major change that SaaS made.

The other one was really faster innovation because you have the software hosted, so you can upgrade it and everyone would use the new version.

And that is huge because then you can really accelerate the innovation.

You can really think and you can collect a lot of data on your site to better understand how the users are utilizing it.

So all of that will help you to faster pace of integration of these technologies and new technologies and build it.

So that, of course, there are some other impacts like the way you look at the customer success because it’s more like now, you are providing this as a service, not software in a way or software as a service.

So you are vested more in to this kind of success of your customers.

So you stick with them throughout the journey rather than the old fashioned way that you give them the software and then that’s their job to make the best out of it.

So all of those impacts that you see now, when you go through the analytics side, in order for analytics side to be part of that success, part of that promise, part of that, okay, you can do what you know.

We need to get to that ultimate self-service capabilities.

That is a tall order.

I agree.

Having said that, good progress have been made.

These progresses have been made by great technologies in the market in all directions.

And AI is one of those directions that automate the automation or make your 10-click, one-click.

One last question for you.

I understand our friend Mohan is going to be appearing on your podcast here soon.

What should listeners know about SaaS Scaled, the name of your podcast?

Yeah, thanks for asking.

So I started SaaS Scaled about two years ago.

And honestly, it just started with my passion for getting in front of SaaS leaders.

Every time I get in front of a SaaS, you know, leader company on the product strategy side, on the business side, on any aspect of SaaS business, it’s just for me personally, I enjoy it.

And I thought that that can also benefit other SaaS people.

There’s no script there, nothing.

It’s just we start talking and then branch out organically, naturally, and just go to any questions and chat about it.

Mohan Rao is one of those SaaS leaders and software leaders, which I highly respect.

And these kind of people just share their insight in a way that they may not even know that what just they said in a few minutes was so valuable.

But it is because it didn’t happen overnight.

Good things in life happens slowly, unlike bad things.

And these people are the people that really, really going through the 20, 30 years experience, collecting details little by little.

And then I would love for them to come to these podcasts and other podcasts and just share their knowledge, because everybody will enjoy it and will learn it in a matter of minutes.

I’m going to steal your words and use them back with you.

Probably you don’t know how much value you’ve added in the last half hour.

And so thank you so much for spending time together.

It’s been a pleasure to get to know you, and we appreciate all the insight.

Thank you so much.

Thanks as always for listening and watching.

Don’t forget to give us a review on your podcast Player of Choice.

Seriously, it really helps our show grow and more people find it.

At the end of every episode, we like to ask one of our AI friends to weigh in on the topic at hand.

Hey DeepSeek, what’s happening?

Quite a wave maker, aren’t you?

Welcome to the show.

This episode, we’re talking about how to turn data into actionable insights.

So what do you recommend?

To turn data into actionable insights, start by cleaning and organizing your data.

Then use visualization tools to spot trends and make decisions based on what you see.

Keep it simple and focus on the key metrics that drive your goals.

And now you’re in the know.

Thanks as always for listening.

We’ll be back next week with more AI applications, discussions and experts.