Please note, this is a special Promoted Episode of the podcast. Please visit their links and help support Experiencing Data.

On today's Promoted Episode of Experiencing Data, I’m talking with Lucas Thelosen, CEO of Gravity and creator of Orion, an AI analyst transforming how data teams work. Lucas was head of PS for Looker, and eventually became Head of Product for Google’s Data and AI Cloud prior to starting his own data product company. We dig into how his team built Orion, the challenge of keeping AI accurate and trustworthy when doing analytical work, and how they’re thinking about the balance of human control with automation when their product acts as a force multiplier for human analysts.

In addition to talking about the product, we also talk about how Gravity arrived at specific enough use cases for this technology that a market would be willing to pay for, and how they’re thinking about pricing in today’s more “outcomes-based” environment.

Incidentally, one thing I didn’t know when I first agreed to consider having Gravity and Lucas on my show was that Lucas has been a long-time proponent of data product management and operating with a product mindset. In this episode, he shares the “ah-hah” moment where things clicked for him around building data products in this manner. Lucas shares how pivotal this moment was for him, and how it helped accelerate his career from Looker to Google and now Gravity.

If you’re leading a data team, you’re a forward-thinking CDO, or you’re interested in commercializing your own analytics/AI product, my chat with Lucas should inspire you!

Highlights/Skip to:

• Lucas’s breakthrough came when he embraced a data product management mindset (02:43)
• How Lucas thinks about Gravity as being the instrumentalists in an orchestra, conducted by the user (4:31)
• Finding product-market fit by solving for a common analytics pain point (8:11)
• Analytics product and dashboard adoption challenges: why dashboards die and thinking of analytics as changing the business gradually (22:25)
• What outcome-based pricing means for AI and analytics (32:08)
• The challenge of defining guardrails and ethics for AI-based analytics products [just in case somebody wants to “fudge the numbers”] (46:03)
• Lucas’ closing thoughts about what AI is unlocking for analysts and how to position your career for the future  (48:35)

Special Bonus for DPLC Community Members

Are you a member of the Data Product Leadership Community? After our chat, I invited Lucas to come give a talk about his journey of moving from “data” to “product” and adopting a producty mindset for analytics and AI work. He was more than happy to oblige. Watch for this in late 2025/early 2026 on our monthly webinar and group discussion calendar.

Quotes from Today’s Episode

“The whole point of data and analytics is to help the business evolve. When your reports make people ask new questions, that’s a win. If the conversations today sound different than they did three months ago, it means you’ve done your job, you’ve helped move the business forward.”
— Lucas

“Accuracy is everything. The moment you lose trust, the business, the use case, it's all over. Earning that trust back takes a long time, so we made accuracy our number one design pillar from day one.”
— Lucas

“Language models have changed the game in terms of scale. Suddenly, we’re facing all these new kinds of problems, not just in AI, but in the old-school software sense too. Things like privacy, scalability, and figuring out who’s responsible.”
— Brian

“Most people building analytics products have never been analysts, and that’s a huge disadvantage. If data doesn’t drive action, you’ve missed the mark. That’s why so many dashboards die quickly.”
— Lucas

“Re: collecting feedback so you know if your UX is good: I generally agree that qualitative feedback is the best place to start, not analytics [on your analytics!] Especially in UX, analytics measure usage aspects of the product, not the subject human experience. Experience is a collection of feelings and perceptions about how something went.”
— Brian

Links

• Gravity: https://www.bygravity.com
• LinkedIn: https://www.linkedin.com/in/thelosen/
• Email Lucas and team: hello@bygravity.com

Transcript

Brian: Welcome back to Experiencing Data. This is Brian T. O’Neill. Today, I have Lucas Thelosen on their line, and you are the CEO—currently—CEO at Gravity, which has a new product called Orion, that’s out. You were also at Google Cloud, as well as Head of Product Data and AI. And then you were at Looker, which was acquired by Google, and you’ve kind of had this whole data product management journey here, and now you’re actually running your own company. So, very cool. Welcome to the show, Lucas.

Lucas: Yeah, thank you so much for having me here. Yeah, I really did. I started off as an analyst, right? I managed analytics teams, and then I was director of analytics, and all the way through to leading product for Google. So, the whole journey.

Brian: Nice, nice. I think there’s a lot people can learn here, given you did come out of this analyst background, but product management has been part of your journey, ran some product, and now you’re literally running a product company that’s doing commercial data product work. So, we talk a lot on this show, there’s kind of two audiences that I serve in my work, which is both companies like yourself with product design, but also internal data teams that are trying to operate in this product-y way, as I like to refer it, treating their internal work as products with a product lifecycle, and thinking about value and outcomes, and not just shipping answers to things or shipping data assets or data containers or connectors and dashboards and things like this. So, I think there’s a lot they can learn. If you’re a forward-looking CDO today, if you’re a commercial data product leader, or if you’re an internal data product leader, and you’re trying to learn how the game is played, particularly when there’s a P&L on the line and someone has to swipe a credit card and literally, like, your value is determined by the income that’s coming in from your product, I think there’s a lot you’re going to be able to learn from Lucas today.

So, I’m happy to have you on but I’m really excited too to talk about this tech that you’re building. So, your company is called Gravity. You have an—AI the way I’m telling myself what this tool is, is it’s an AI analyst. It is an additional analyst or multiple analysts that I can add to my team that’s doing the work of what many human analysts are doing right now. Is that correct? How do you position this product in simple language to somebody?

Lucas: Yeah, exactly. I mean, one of the challenges I had when I was leading an analytics team, was that we never really had enough bandwidth, right? And so, what if I could with, now, these new technologies actually help organizations get the bandwidth they need to elevate their own position to be more of a manager, right? A manager of AI, a manager of analysts. I often talk about, you are the conductor, and let us be your orchestra, right?

If we can play the instrument, or the AI can play the instruments for you, right, then you can actually conduct and focus on the outcomes, right? And so, being able to elevate an analyst to be a data hero in an organization by giving them this AI tool, this AI analyst that does a lot of the groundwork for you, I think, that’s a really exciting opportunity.

Brian: So, is this tool more positioned as, this is for analysts and it’s going to do a lot of the dirty work that you don’t really like to do, so you can focus on doing the work you do like to do and the higher value stuff? Or is this meant to be, no, this is a full A-to-Z analyst replacement or addition that I can add to my team for someone who is maybe managing a team and needs more resour—maybe they’re running product or they’re running marketing or finance and they can’t get headcount. Is it more for that person or the analyst that needs—

Lucas: It’s a good question.

Brian: —more, bet—yeah [laugh].

Lucas: I think it is not just a talk-to-your-data tool, right? There’s a lot of tools out there where you ask a question and it goes and tries to write a SQL query. That’s not—like, Orion does a lot more than that. It actually proactively thinks about it. It does what I call deep analysis. It does, like, different cohorts, root cause analysis, predictions, right, all these things.

So, the way—but it doesn’t really get rid completely of an analyst because you still need someone in organization who takes on that leadership role of what does data mean to our organization and what is the experience that a business leader, let’s say, you know the head of marketing, what experience does the head of marketing need to get the most out of our data? And so, Orion gives you the ability to curate that experience for them, right? We have multiple layers at which you can add context to it, where you can control which data Orion should access for certain use cases, right? But then, once you have it set up, right, then the chief marketing officer, they can go and ask a question, and Orion thinks proactively about it, and can push insights to the CMO, and so it can be a full analyst, right? It is there—I mean, you still need people though, that define the data foundation, that define the vision of what analytics means, and works with Orion on the journey and the user experience that a business user has.

Brian: Got it, got it. Let’s walk back, just to give some context here, like, I’m always like, well, where did this start? Like someone had an idea for something. And I’m always curious, especially with founders, was this a solution in search of a problem and now you found a problem for it, and you’re like, “Wow, we have large language model, we have structured data, we have unstructured data, there’s a lot we can do here. Maybe I’ll see if I can put something together and find a market for it.”

Or did this come out of, like, repeat problem exposure? Like, in your last role at Google, where you’re seeing, like, the same thing over and over again. And why are we throwing all this traditional resources at it, when we could be using this technology to do it? Where did this start? Was it more in a repeat problem thing, or more of a I can do—there’s amazing technology here, and we think we can find a market for it.

Lucas: It was a repeat problem thing. So, the first problem, when we talked already about the bandwidth challenge, right? The bandwidth challenge is very real. Every analytics team I got to work with through my tenure at Looker, but then also at Google, was understaffed. Like, it was never appropriately staffed, and one of the top requests was like, “Hey, when you meet with our leadership, you know, can you tell them what a real analytics team should look like and how many people they should have?” Right? That was a super common request.

But then the second thing that we were asked a lot, and I tried to solve was that companies sometimes, like, they [sigh] you know, if you’re an analyst, you worked maybe at three different companies, and you want analysts there. And there was a strong interest in, like, well, what does good cohorting look like? What does a good root cause analysis look like? How do you do predictions for, you know, revenue predictions, or whatever it might be? How do you do marketing analytics, supply chain analytics, right, all these different scenarios?

And so, what I tried to do is create templates that you can use. Like, best-in-class, this is what I’ve seen working across, you know, let’s say the top 80 companies that I got to work with, this is what top-of-the-line marketing analytics looks like. You know, it could be customer support analytics, whatever the scenario is. So, I created these templates. We created these templates at Looker and then at Google.

But the problem is, like, every organization is quite unique, right? Like, even if you create a template, it still needs so much customization. And then all these, like, all these consulting firms come in and they want to charge a ton of money because it’s such a complicated codebase, you know, to modify it for your business. We kind of put it on ice, until generative AI came around, and all of a sudden I was like, you know what? What if I take these templates of best-in-class analytics, and I put it as the base understanding for Orion, and then I add to it the ability to help with the bandwidth, right? Then I actually have a really interesting product here. I can now bring best-in-class analytics for different scenarios, and I can help with all these grunt work, right? All these like, hey, I’m getting tickets and requests for analysis for this and this, and I just, I can’t do it all. Like, actually, can combine these two and help organizations on both fronts.

Brian: So, I’m assuming then your primary sales channel here is someone that’s managing a team of analysts that needs to reduce workload, or they want to effectively try to replace—not replace an analyst, but perhaps add capacity. Like, I’m a user of analysts, but I don’t manage those resources; I might want to try to offload some of this to Orion. Or is it really more going to be to a data team that wants to internally offload some of its work through technology? Who do you sell into, I guess is my question.

Lucas: Yeah, so we have two personas that we sell to. The head of data that has—exactly—either band—so either bandwidth, you know, as a challenge, or they have a grand vision of what they want to do. Like, the real data product manager, right? You have a vision of what you want to do, and in order to do it, you just, like, you can’t add ten more analysts to the team. So, with Orion, it would free up capacity so your team can focus on more strategic tasks, right?

The other persona is a product manager that actually wants to get more advanced capabilities to their customers. So, we do a decent amount of embed work, where you embed Orion into your product, and now your customers can get quite advanced analytics on your product, right, on what you’re offering and what you’re doing for them, so they can and they can have conversations with it, right, and all these things.

Brian: Got it. What kinds of objections do you get here? And the reason I’m asking this is we hear a lot—in the data product community and the space, we hear a lot about low user adoption, which, when data products involve AI, there’s usually a trust component there, not understanding where the data came from, did it look at all these special exceptions, I know about all these—I know where the demons are buried there. How can it possibly know about all this kind of stuff?

What kinds of—do you get those kinds of objections? And I wonder, how does the product either handle that, or maybe you have to handle this in a sales call. I’m not sure, but I’m sure there’s some suspicion here about how efficiently this can do this. And it’s like, yeah, these reports look amazing, but like what went into them, and how do I know, especially at scale, if it’s doing a hundred of these things overnight, like, how could I possibly validate that this information is right? And I could see the skepticism there, and I’m wondering how do you handle that, either in the product or the sales conversations? Like, that must be hard. I don’t know.

Lucas: Yeah. Oh no, absolutely. I mean all of the ones you mentioned, we get all the time, right? I think the first one is around hallucinations and being correct, right? And so, as we set out to build Orion, like, the number one thing, as anyone in data knows, right, you have to be a hundred percent accurate, or you lose the trust. And once you lose the trust, you’re done, right? Your business is done, your use case is done. It takes forever to get the trust back.

So, accuracy was the number one design pillar that we built around. So, we actually run any analysis multiple times. We have multiple quality assurance agents in there whose only job it is to ask, “Where did you get the data from? Walk me through the steps you took.” And so, we have these agents ask other agents right, to explain things to them.

And so, we have, then, full lineage, so then of course an end-user can also ask, “Where did you get this from? Walk me through the steps?” And Orion has it all right there. It has a spreadsheet that’s annotated with all the steps it took, right, and where it got the data from, every time with sources. So, that’s a super common one.

And then to your second point, it’s not magic. I think that’s another important one that everybody here should keep in mind, right? And anyone—like, the marketing always sounds like magic. You still need to onboard Orion like you onboard any kind of human analyst, right? So, if you have—a lot of organizations, give us their onboarding materials of what they use for their own employees.

Most of them actually don’t have it. Most companies do not have great onboarding materials, so what we do is we have a series of meetings, so we take the transcripts and actually give it to Orion. It can get a lot out of the metadata in your database, in dbt, in your business intelligence tools, but there is still so much, you know, institutional knowledge that a senior analyst has that we just have to get into Orion, right? Like, a glossary of terms, and this is, you know, out of the 16 date filters, this is the one we usually use, but sometimes we use this one over here, right? And so, Orion has to learn it, so we built an advanced memory capability into it, where it actually stores, runs, it picks something up, right? It stores it.

And so, it takes a couple of weeks, right? But compared to—I mean, and then the nice thing is here, so we have this large company. They have—it’s a massive company, they have 60 analysts, and they now actually use Orion to onboard their new analysts because Orion now has a really good understanding of their business, their acronyms, how they do things, how different department do different things. So, when a new analyst comes on board, Orion can explain how the organization does data and how different teams like to be served.

Brian: Oh, interesting. Got it. Got it. Tell me a little bit about how, like—your profile, when I was looking you up on LinkedIn and all of this, and your journey into product at Google, and you mentioned data product management in here, and how treating this, like, helped you get your job at how treating data as a product and this orientation helped you get your job at Looker. I think you were a VP of professional services, was it—I think—there?

Lucas: Yep, exactly.

Brian: Right. Where did all of that come from? Because, like, this, is a fairly new concept, this idea of applying product management and design thinking and what I would call software product management practices to, you know, internal data teamwork, or just data work in general, and analytics, machine learning and AI, where did that come from? Like, people just wake up and say that one day [laugh]. So, I’m always curious your origin, finding this space?

Lucas: Yeah, and this is where I get the chills because it really was a wake-up moment. I went to a talk. We had invited—I was working at a tech company, I was the head of data—and we had invited this CEO of this very successful company to come in and talk to us. And he was going through—it was during the financial crisis when he took over as a CEO—and he said, “What I realized is that we have two products: our main product, but then our second product is actually data. And once I realized that our second product is data and I started to treat it like a product, that was a pivot moment for our organization. And all of a sudden, like, things started to really change.” And he was able to sell it for over a billion dollars to Oracle a couple years later.

And I went out of this—I actually wrote him a thank you note ten years later because it was such a crucial moment in my career—I was like, “You know what? Data is the second product of my company.” Right? Our first product was that technology we’re selling, but the second product is data, so I should treat it like that. And I sat down and I wrote a plan. Like, this is how I’m going to treat data, this is our roadmap that we’re going to have.

I looked up at—I was not a product manager, right, so I looked up best practices for product management, and I presented it to our leadership team. I was like, you know, I know every day you’re asking me and my team to do certain types of analysis, but here’s how I see data as a product. This is how we’re going to tackle each department. We’re going to have, you know, six week cycles, and we’re going to move through it, and we’re going to have release notes and all these things. And they’re like, “Yeah, go for it.”

And so, I ran it like that. And then Looker, so I used Looker, they had just started, it was a BI tool that nobody knew about. They heard about this and they were like, “Hey, do you mind. Do you mind being a reference call for a couple of customers that we’re selling to right now?” And so, in one week, I was on the phone with Sony, with Yahoo, and with Uber, the ride sharing company. And each time they were like, this is insane. This is awesome. Do you have, like, a sample plan we could use? And so, I sent over my spreadsheet that I use for project planning and everything.

Anyways, it changed my career, right? I was, like, a very reactive data analyst team that always wanted to have aspirations to do more advanced analysis and some data science. This was 2013, right? Like, this was really early on. And so, I never called it data product management. But then, you know, I ran a consulting practice briefly where I got to set up analysis for, like, Amazon and Walmart and Uber actually became a customer, Disney, like, all these companies, and I just, like, I really felt like we should call it data product management.

And then I got to Looker, and they really liked the idea, too. I gave a couple talks on it. It didn’t really catch on at the time, but I kept saying it. I was like, you have to focus on the experience of the end-user, the business user, and how they actually—or the external, the vendor, right? The vendor might be a user, too. How are they experiencing it? What is the [drill 00:19:36] path that they’re going to have when they see the data? What’s the next question they’re going to ask? And what actions do you want them to take based on the data that they’re seeing, right? Maybe we put the action button right next to it, and we can close the entire loop. Anyways, I’m really passionate about this topic.

Brian: Yeah [laugh]. No, so it sounds like—yeah, it’s interesting. You kind of had this lightning bolt moment, it sounds like. It changed something for you. And you kind of dovetailed nicely into experience and user experience and design and thinking about the end-user and the persons that’s going to use these solutions.

And so, I’m curious if you could talk a little bit about, you know, you had the skeleton for Orion in your head. This LLM technology matured almost, like, overnight, it kind of emerged. You’re, like, “Oh, there’s the missing ingredient to possibly turn this company into a product.” But a product also has to be something someone’s willing to pay for, that they’re excited about, that they see the potential, and that the level of effort required to get the value out of it is not so high. These are all places where design typically can help, right, by reducing friction, reducing cognitive load, making things simple, making them exciting or delightful to use.

Tell me about your journey? Like, how did you even know that you needed to have experience design play a part of this product? And I’m asking you a leading question because I know that you have a designer on staff. A lot of data people don’t associate design as essential or user experience as being essential to their work. I think the first journey is data product management, and then eventually, when they mature there, they then realize, like, oh, there’s this whole other discipline that just works on that part. Maybe I get some help with that. How did you arrive at the fact that you needed this in order to make this product work?

Lucas: Yeah. And I think that’s where the vast majority, as I look around, of other companies that, you know, I just was at, Coalesce, at the dbt conference, and as I talked to other people in the data space, almost nobody has been an analyst before. Like, almost nobody ever—like, you know, you get a dashboard. Okay, you see the dashboard. You see a number moving, right? Like, what do you do next? What is your step, you know?

And so, having never done that yourself makes it I think—it’s a huge disadvantage to build any kind of product in this space, right? So, as I approach this, I try to put myself into the shoes of the business user, right, of the head of sales, or whoever it might be, an account manager, right? Like, what is there—when they see something, I want them to have a reaction of, like, “Okay, now I want to do this,” right? Like, if they don’t have a reaction that causes them to take an action, then I haven’t really done a good job of putting data in front of them, right?

And I think that’s where this adoption challenge and repeat usage challenge comes in, where dashboards die pretty quickly, right? You set it up, you—I think there is—okay, quick step back. I had a person come up to me at this conference and asked me about this. So, it was like, “Look, I just built this dashboard and we spent three months on it, and within a couple of months, nobody’s using it anymore. Like, what the heck? Like, I feel like, so… I failed at my job.”

I was like, “No, you did not.” Because the purpose of data, the purpose of insights, of analytics, is to change the business gradually, and so as you build really great reports, they cause business users to ask follow up questions and to evolve the business and slowly change it, you know, continue to optimize it to something new. So, when the questions, the next questions, are different from the ones three months ago, that is actually—that means that you did a great job because you moved the conversation, you helped the organization evolve. And so, that your dashboards die is actually a great thing. And so, as I—with Orion, I started talking about sentient dashboards, now, where it’s never really a static thing anymore; it continues to evolve with you, right? As Orion sees questions arise in different functions, it starts to modify the analysis it does, and as it sees, like, multiple people asking similar questions, right, its proactive approach also changes a bit in different directions.

Brian: How do you go about arriving at—like, I mean, you had these core templates that you talked about, but you’re also talking about how, over time, it sounds like that, literally, the design of these things may be evolving as questions mature in the organization, or as business needs change, those templates may also need to change. So, I’m kind of curious, how do you go about designing a product like this, in terms of you’re setting the scaffolding and it’s piping data into templates, versus Orion is making a choice about this should be on a chart and it should be compared to this chart, and probably they’re going to want to filter by this thing or not filter by this thing. This is where you can get really heavy, opinionated software. Or you could say, “No, we need to intercept that with deterministic planning.” Like, we’re going to decide when these templates should be rendered at runtime.

There’s a lot of design challenges here to try to make the outputs of Orion useful. And I’m kind of curious, how did you and your design and your product group think through these—this is probably—wow, I’m sorry, sorry. I hope you’re—I think you’re probably smart enough to track what I’m trying to ask you, but in terms of, like, leading the interfaces and the outputs that Orion provides by guiding it and saying, use these templates at these points, versus letting Orion decide how to—should this be a table? Should it be a chart? Should it be just a sentence with some conclusion in it? Tell me about that process.

Lucas: No, that’s a really good question. So, the design right now, if you go into the UI at Orion, you have the ability to control context at multiple levels. So, at the organizational level, right, here are things you need to know about our organization. At the project level, which we use for teams, or, like, any kind of special project you might have. So, you can say, you know, for sales, I want you to think about certain ways in these, and like, certain formats that you should be doing, certain tables you should be looking at.

And then you can go down as far as the user level, where the user can say, you know, I never want a scatter plot. I never want a heat map. I only want—like, you can—the user can be quite specific, too, on what they prefer. So, you have these multiple levels of where you can give Orion guidance. And you can choose to make it quite strong, right—never do this; always do this—or you can choose to give it some leeway.

Like, one of the points where leeway is important is smart alerting, right? If I want Orion to keep an eye on these eight dashboards for me and alert me if something jumps up or down, and tell me about the underlying metrics that actually changed and why they changed, it needs to have some liberty, right? Because you can’t define three-hundred different iterations of alerts, in essence. So, you need to work with Orion a bit on what the thresholds should be. So then, as Orion produces a deep insight, you know, like, does this deep analysis, it produces this insight.

And it’s very similar to a dashboard, in essence. It shows you visuals, and it shows you the narratives and things it thinks you should pay attention to. You then have the ability, as an analyst, to change these and approve of the final output. So, you can say, “Actually, over here you should do this, and over here you should do that,” and then lock that in. And if you say, “You know what, I really like this, put this on a schedule. Let’s do this every Sunday night so everybody has this on Monday mornings.”

You can define with Orion what elements should never be changed, you know, what should stay pretty static, and Orion is actually able to put that into just normal code. So, the AI will not run this again. The Python script will always be the same, right? Certain things will always be the same. And then the narrative for certain component, like, interesting tidbits, might change. But it’s a really—the question you’re asking is a really fascinating problem because we’re designing for something that has never existed before—

Brian: Right.

Lucas: —you know? And so, building for that, like, of course, we’re looking at all the other AI tools and how they do it. We had, over the last year, right, our UI became more and more complex and we took a step back, and we’re like, you know, it has all these bells and whistles, now. We need to take a step back and make it very simple-looking again, so it doesn’t look intimidating, right? But then you can go down and you can actually fine-tune it, if you want to.

Brian: Yeah. Do you have a way that you measure user experience such that you know that it’s not getting too complicated beyond just your own subjective opinion about whether it’s complicated, too hard to use, or it’s too many bells and whistles. Does your team have a way that they validate this so you know that you’re not, you know, shipping features and stuff that are actually going to get not used, either because they don’t understand the value of it, or it’s just too hard to figure it out, or I can’t get it to do what I wanted to do. Is there a way that you guys measure that at all, or think about that strategically?

Lucas: Well, so right now—and I’m taking this from the Looker playbook—we just stay super close to our customers. Like, we meet every week, right? We always we have Slack channels with all of them, and just having conversations about it all the time. And then, all the engineers we have are just so hungry and thirsty for, like, hearing the feedback from our customer engineers. So, we constantly like this, we’re sitting around the table, right, everybody’s talking about the customer. What did the customer say about this? How did they like that? And so, customer feedback is our product roadmap, in essence. Like, it’s constantly listening to the customer.

Brian: Got it, got it. Do you do any, like, observational research with them, where you’re actually going out and just observing them, or is it mostly, are you waiting for feedback or asking, quite, like, verbal-type feedback?

Lucas: We do do screen sharing, you know, where it’s like, okay, like, “Why don’t you just open up your screen and we just do it together,” but in just seeing where they click and how they click. And, like, “Okay, this just took four clicks. And it really shouldn’t take four clicks,” right? Like, or, you know, we do some—we want to be really careful also with, of course, respecting privacy, right, so any kind of tracking, of, like, people tend to prefer pie charts over here, or stacked bar charts, or whatever, it might be, so we’re really cautious with that. Like, that’s… data stays in the customer’s instance. It’s not leaving that.

But we do screen sharing and we hear what they’re saying. Any kind of more sophisticated analysis tracking we don’t do yet because, just we’re an early-stage companies, still, in many ways, until just having these conversations and screen sharing is very much doable, and it’s also, I think it yields a lot of results for us.

Brian: Yeah. I’m a big—generally agree with you that, you know, qualitative feedback as an initial point is generally where to start, not with analytics. On analytics has all kinds of [laugh] product problems, in my opinion because you’re actually measured—especially with user experience you’re tending—you’re actually measuring the product; you’re not actually measuring someone’s experience, which is basically a collection of feelings and opinions somebody has about how something went. And computers right now cannot actually track that information, so it can be a misleading metric. But that’s my opinion about it.

I’m curious, one thing that’s being I’m going to jump—put my real commercial hat on here for a second—there’s a lot of discussion right now with AI products that, especially given, you know, the cost to make calls, I don’t know what, if you’re using large language models, or, like, your own small language models are open-source, but cost and then therefore pricing. Are you doing outcomes-based pricing or report-based pricing, or is it just, like traditional SaaS, seat licensing? Like, tell me about your thinking there. Have you adopted some of these more modern discussions that are going on right now about outcomes-based pricing? And I’m really curious how that happens with, like, a report that’s being looked at monthly. You know, like, tell me about that.

Lucas: I would love to do outcome-based pricing because the, like, every time we do this with customers, we’re saving hundreds of thousands of dollars, sometimes millions in a week, right? Because I’ve just—I mean, there’s so many interesting ways this technology is being used where one way it really helps is that it allows you to step outside of your common patterns of analysis. “But like, well, we always looked at it this way. Why would we look at it a different way?” Right?

So, but Orion doesn’t, quote-unquote, “Care,” as much about that you have done it always like this, but it just checks, “Hey, like, maybe I should take a different look at cohorts, and all of a sudden, you know, there might be some interesting results in here.” It doesn’t complain that it has to run 40 permutations of a cohort analysis.

Brian: Right.

Lucas: Like I would, as an analyst, as a human analyst, right, after three or four permutations, I would stop and call it a day [laugh]. I can just, like, this is good enough, you know? Or, as a human, I also have this bias—and I mean, we all do right—we pay attention to the red numbers. We see a red number, and we do an analysis of, like, what went wrong, you know, what can we do so we don’t have this red number anymore? And it turns out, not many people pay attention to the green numbers of, like, what’s going well, and why is it going well, and what can I do so more things are going well in my organization.

And it’s interesting to see, like, Orion does not have this red or green bias. It investigates both equally. And so, you have this. All to say, I would love to do outcome-based pricing. It is a very challenging thing to do because then all of a sudden you meet with a, you know, finance team, the accounting team, and they’re like, “Well, you know, we, like, we also did other things, and so maybe that contributed to the ROI over here, so you can’t really say you got us 300% ROI.” You know [laugh]?

Brian: Right, right.

Lucas: So, really challenging. And so, what we’re doing—short answer—right now, you know, we have a platform fee plus usage. So, based on the amount of deep analysis that Orion does, it is expensive to run, right? Like we run a very sophisticated—we have over 30 agents working on this, and so it’s very, very thorough and accurate and insightful, but it’s not free, right? It’s not like—I mean, it’s super cheap, compared to, you know, having someone do 30 hours of manual analysis, but in the realm of computers, it’s not free.

And so, we have to—like, we have this one customer, they want to do, I think, about 30,000 reports every Monday morning, right? 30,000 unique reports? That’s a lot, and so we have to charge something for that. For most organizations, right? We can make it quite affordable because it is, like, the deep analysis, you know, we have to charge a little bit for that.

So, we do a platform plus usage fee. We are in conversations. You know, back at this conference, I just was at—everybody’s talking about, should we do it user based? Should we just do it consumption-based, right? We just charge a little fee on top of the LLMs, in essence. Everybody is trying to figure it out right now, I definitely want to have conversations with our customers on what makes the most sense with them as well, so we can learn, you know, what resonates, what doesn’t resonate, what makes sense for that budget, and we’ll make it work.

Brian: Got it. Got it. Are you supposed to be able to—like, if I bought this product today, it runs in the cloud, so you’re—and just your application, it runs in the cloud. It’s a software application runs in the cloud. It has, like, an LLM-style interface that then generates different reporting type artifacts—I’m just trying to paint a visual picture, since it’s an audio podcast, about what it looks like—is that something anybody can onboard themselves and, like, the expectation is, you know, swipe your card, we can get your team running today, or they can get themselves up and running, or is there, like, an onboarding, honeymoon process and professional services to wire up connectors and do all that kind of plumbing work that has to happen. Tell me about that whole—I call it the honeymoon experience, and for you, maybe there’s a period where there’s actually a demo or something, since it’s more enterprise software—but tell me about that process to get to the first value.

Lucas: [laugh]. Yeah, no. Awesome question. So, we do a three months—three weeks trial, sorry [laugh]—three weeks trial. And really, within the first week or so—

Brian: You heard it here first, folks, three months [laugh]. Just kidding [laugh].

Lucas: [laugh]. No, no, no. We do—exactly. And we—well, it’s actually, it doesn’t take all that much to set this up. We were debating if we should just do a do-it-yourself, but it’s such a new product that I want to be there. You know, our solution consultants, we want to be there with you.

So, we do the three week trial. We just need credentials, right? We need credentials to your BI tool or to your database, depending on how you want us to work with your data—you can do both—so once we have the credentials, we do a couple use cases, we set it up for you. And most of the work is really spent on tuning Orion, like the onboarding of a new analyst. That’s what we do, right? We tell Orion, “Hey, these acronyms mean this. This is how marketing likes the things. This is how account managers like to see things. This is how vendors want to see things.”

And we explain that to Orion because oftentimes it’s just not documented very well. And then right now, if you know, if you buy Orion, we actually say, we will stay with you. We have this jumpstart where we have our AI consultants with you, but we will stay with you for the first year. So, we have a Slack channel where you can always ping us, we always jump in, as any—right, like, that’s where the big difference between one of those behemoths, like Google or you have this very dedicated technology company right here, right? Ike, your success is more important to us than it is probably to you [laugh], you know, so we do everything to make this successful. I don’t think there’s actually all that much work to be done and that’s why I so freely offer these services because it is—but it is a new tool, right? It’s a new way of doing things, and so I want to stay close to it. We want to stay close to it.

Brian: Yeah, yeah. How much dependency is there on the organization having a certain level of data maturity? I’m talking about mostly about infrastructure and plumbing being in place, such that Orion—you know, garbage in, garbage out. Or maybe not. Is Orion good at, like, dealing with poorly labeled information? There’s five customer databases that over time, and three applications are still running because Australia is doing one thing, but UK hasn’t done it this way, and US has different regulations, and so we’re live three different, you know, e-commerce platforms and blah, blah, blah, you can see where this is going. How mature does a buyer need to be with their data in order to get value out of Orion sooner than later?

Lucas: I think this is the thing where every AI company, you know, it’s the same thing. A solid data foundation will get you very far. Like, the more you have descriptions on columns, right, or you have any kind of documentation, the better off you’re going to be. It doesn’t have to be perfect. Like, we have customers that have 500 star schemas, [laugh] and each one—like there’s a lot of overlap between the different schemas, right, 6000 tables and all these different things.

So, it doesn’t have to be perfect, but the more you have, like, Orion, can get a lot of information out of metadata and really quickly understand what’s what, but the better your foundation is, the better your descriptions, your, like, your documentation is, the more successful any AI tool, not just Orion, will be. Of course, we’re working, we’re constantly adding more features so it is able to handle all kind of messy situations, but you know, we have to guarantee a hundred percent correct results, and in order to do that, you know, the better the foundation is, the more likely we’re able to achieve, very quickly, a hundred percent accuracy.

Brian: Yeah. In terms of the results, and we talk about accuracy here, so there is a consumer experience here, in this case, the consumer being someone that’s receiving the insights that Orion is providing here. Are those people actually consuming this right from your interface, or is this kind of a raw material that’s then being translated into a PowerPoint deck, or whatever it may be? I’m curious how much the beneficiary user is actually using Orion directly versus is that an intermediate step?

Lucas: Yeah, so the business user, right, let’s say, an account manager, for example, they can get Orion’s, like, the outcome, the insights, into their email inbox, into their Slack channel, into a slide deck that they would, like, pre-populated with notes underneath, right? I really want to get data as close to the business as, like, where the business already is. I don’t want to force them to have to log into another UI. Of course, we have a UI, right? That’s what the data team will be. It’s beautiful. Business users can absolutely go there as well.

But [sigh] I don’t want that to be a hindrance, right? Everybody already has 16 different systems to log into, so can we make it as easy as possible for an average business user, to have data pushed to them, right, where it’s like you get an email on Monday morning. Hey, you have these meetings coming up today. I created some business review decks for you. Here’s probably what you want to focus on with this specific customer. You have an upsell opportunity there, right? Like, as proactive as it can be, as focused on actions and outcomes as it can be, and ideally in the place where the business use already is.

Brian: Got it, got it. So, like, here’s an example. You gave, sort of, like, a, like, maybe a sales use case there, or something where, “Hey, there’s this prospect. They’re ripe to buy. Now, we tracked something else happened. They just got investment. We saw this in a news article. It got consumed. Oh, they probably have buying power, now. Like, you might want to touch them again, dear salesperson.” Right?

Does this mean all, like, the sales team or these other business users, anybody can get access to the interface, or is it really like the analysts have access, you give them a small, like, maybe an email delivery, you can get that, but I could see a salesperson maybe wanting to go in and dig into, like, wait a second, like, these people were such a hard no on buying, very low propensity to buy. Orion’s telling me, like, they’re a ripe target. I don’t believe this. How did you come up with that? Click, I want to go see or show me the math. Like, does that sales person, for example, do they get—are they now in the Orion user experience directly? Or, how do you handle something like that?

Lucas: If it is an email, then yes, you click on it, and it gets you right into the chat with this insight UI. So, you land right there, you see the insight on the right side, you see the chat in the middle, and you can ask Orion right there: “Walk me through. How did you come up with this? Why do you think this is true? Walk me through your thinking. Have you considered this?” Right? You can even add additional PDFs or any blog post you might have found, right, and Orion can interact with all of that.

Brian: Got it.

Lucas: In Slack, you could just talk right there with Orion, and you can talk with the insight. A couple of customers have embedded Orion into their portal. So, let’s say, you know, you have large companies that you are servicing, they want the ability for that company to also talk to Orion. So, let’s take the example of Coca-Cola, right? You send Coca-Cola this report of, “Hey, this is how your ads are performing on our platform.” And then Coca-Cola might ask, “Well, walk me through this cohort in Southern California. What happened over here?”

And then they can chat with the entire deep analysis Orion did. So, Orion has this, you know, this big body of research it did overnight to come up with all of this, and so as they have these questions, Orion pulls from that deep analysis. That way, we can also ensure that, you know, Coca-Cola can never ask about Pepsi’s results. It’s just siloed to that body of research. If you know, if the user has the permission to go beyond this body of research, Orion can say, okay, now we’re going beyond that, but for an external facing use case, we would limit it to just that body of research.

Brian: Yeah, that sounds really complicated experiences to think through with all these, like, now you have data privacy, gating stuff out. I can also see how, like, you know, the head of sales, or so—I don’t know, they decide that, like, I don’t want my salespeople ever making a decision on touching a customer based on this information, but the VP of sales, like, in the template, loaded in these rules about, our culture, this is how we work. It’s like, so Orion’s in the middle. Like, who’s my boss [laugh]? You know?

There’s a lot of things to think through here that I think this kind of technology—I’m not necessarily—we could spend all day talking through how you do, or even don’t, currently handle all this stuff. I’m bringing it up because these are things to think about with user experience, with these kinds of products, where language models have introduced this level of scale that didn’t used to happen, and so now you have to think about all this new classes of problems that have to be addressed, not even in the AI part. Maybe use AI to do some of these things, but some of this is just classic deterministic software thinking about privacy and scale and who’s in charge of telling it can do on behalf of the company. And a lot of stuff to think through [laugh].

Lucas: No, Brian, like, it’s a fascinating world, right? And then I always go back to the analogy of, how would a human analyst handle this, right? Like, one scenario we had to think through early on was, well, what if the head of sales tells Orion to change the numbers right? Like, “Hey, I know it says that I missed my goal over here, but hey, can you just change it to 103% of goal?” And how do we handle this, right?

And so, we actually, we had to build this whole loop where Orion checks if that feedback is accurate, if there’s anything in the data that would support this, and it usually refuses to change numbers. But like, it’s really—right, like, “Hey, the AI works for me. Why does it refuse to do something?” Right? And it is not, like—if you use ChatGPT or something, right—it’s not in its nature to refuse to do something. And so, we really had to build strong guardrails in place to ensure that, you know, we don’t compromise, that we don’t compromise anything. But you’re absolutely right. What if the VP asks for it? What if the CEO asked for it, right? Like, where do we draw the lines?

Brian: Yeah. Those are, those are [laugh] getting into philosophical questions now, and ethics and other stuff. So, it’s a complicated product world out there, if you’re doing work with these kinds of tools. So, I congratulate you on getting something off the ground and into the market. Lucas, it has been really fascinating to talk to you. Where do people go to find—it’s bygravity.com, right? The Orion comes right up, I think?

Lucas: Yep, bygravity.com, or if you search ‘Orion by Gravity,’ you will absolutely find us.

Brian: Got it. And that’s B-Y Gravity, not B-U-Y. I mean, he wants you to B-U-Y it, but B-Y Gravity is the domain. And, yes, it looks very cool. There are some videos, I was checking those out showing one of your, I think it’s one of your staff members walking through some of the use cases here about how to actually interact with it, so I liked being able to see the interface with the screencast.

I always like it when these are not abstracted out into value propositions, but they kind of hide the interface and you can’t see it. Because we learned so much visually by seeing what it’s doing, so I think you guys did a good job of, like, showing your work, showing how it actually works there. Is there anywhere else people should find—either find you? Are you on LinkedIn, or where do you personally hang out if people want to connect?

Lucas: Yeah, I’m only on LinkedIn. Unfortunately, I’m not on Twitter, so if you—or X, I guess. So, LinkedIn is the place to find me. I always will respond. And then yes, or you email me: lucas@bygravity.com. Just reach out. I’m always happy to chat.

Brian: Cool. Lucas, any final last words you want to share with our audience, just about your kind of having this product mindset with data and creating value and creating outcomes, and any kind of—I want to give you the last word here.

Lucas: Well, I just think this is an incredibly interesting moment in time where you can really accelerate your career, right? Like for me in 2013, to say, “Hey, I’m a data product manager now,” really has changed my whole trajectory. And right now, the train is leaving the station, right? AI is happening. I mean, you might be skeptical, and rightfully so, but it’s happening, so I would just encourage you to jump on and learn about how you can be this conductor, how you can be the manager of AI and elevate your career into a more strategic position. Then it’s not a threat to you, right?

So, take this opportunity right now, right? Like, any month you wait, anytime you wait, right, it’s like it’s going to be further and further behind. Just imagine, you know, you were there in 2013 and just said, you know what? “Cloud is real, it’s happening, and I’m going to jump on this,” you would be a leader in this space right now. And very similarly, right now, we have this moment in time with AI analysts. You want to pick a really good one, right? You don’t—there’s so many out there. It’s very confusing. But use this for your career.

Brian: Awesome. Lucas, thank you so much for sharing your ideas and giving us kind of an audio demo of Orion. I really appreciate it.

Lucas: Thank you so much for having me.