2814: ThoughtSpot CTO on Why AI Could Make Products Worse

[00:00:00] Welcome back to The Tech Talks Daily Podcast, where every day we navigate the ever-evolving

[00:00:08] landscape of technology and also unravel the complexities.

[00:00:12] I'm Brinu Insights from the Frontlines of Innovation.

[00:00:16] Yeah, I'm your host, Neil C. Hughes, and today we're going to dive into a topic that's

[00:00:20] sparking debates across the tech community.

[00:00:23] And that is the paradox of AI in product development.

[00:00:28] Because joining me today is Ben Stanzel, field CTO at Thought Spot, a company that they

[00:00:34] vanguard of AI-powered analytics.

[00:00:37] And Ben's going to be bringing with him today a wealth of experience and a unique perspective

[00:00:42] on why in the rush to AI enable everything we might actually be taking backwards with our

[00:00:49] products.

[00:00:50] So we're going to explore the critical balance between innovation and utility and also how

[00:00:56] Thought Spot is navigating these waters to ensure AI genuinely and answers user experiences

[00:01:02] without falling into the dreaded hype trap.

[00:01:05] Quite a balance, isn't it?

[00:01:06] But before we get today's guest on, I need to pay the bills.

[00:01:09] We've got a huge podcast hosting fee to pay for when we're releasing 30 episodes a month

[00:01:15] and this month I've partnered with a company called Kiteworks.

[00:01:19] Now legacy MFT tools are dated and lack the security that today's remote workforce demands.

[00:01:25] So companies that continue relying on outdated technology though, they put their sensitive

[00:01:30] data at risk.

[00:01:31] And in a world where digital threats evolve daily, the need for a secure modern solution

[00:01:36] has never been more pressing.

[00:01:38] While Enter Kiteworks, the beacon of security and efficiency in managed file transfer

[00:01:42] at Kiteworks isn't just any MFT solution.

[00:01:46] With its FedRAMMMODRA authorisation awarded by the Department of Defence since 2017,

[00:01:51] Kiteworks sets a new standard for security and this certification is not just a badge.

[00:01:56] It's a promise of unparalleled protection for your data, offering a fast track to CMMC

[00:02:03] compliance and eliminating some of those cumbersome costly DIY authorisation processes that

[00:02:09] plague so many of the alternatives out there.

[00:02:11] So please step into the future of managed file transfer with Kiteworks you can find out

[00:02:16] more information at kiteworks.com to get started that's kiteworks.com to get you started

[00:02:22] today.

[00:02:23] So book a lot and hold on tight as I beam your ears all the way to New York where Ben

[00:02:29] is waiting to join us today.

[00:02:32] So a massive warm welcome to the show can you tell everyone listening a little about

[00:02:36] who you are and what you do?

[00:02:38] So my name is Ben Sancel, I am the field CTO of Thoughtsbot.

[00:02:43] Prior to that I was one of the founders of MODE.

[00:02:47] Both MODE and Thoughtsbot are BI tools, MODE was a BI tool that primarily served data

[00:02:52] teams and technical teams and then six months ago in July of 2023.

[00:02:58] MODE joined Thoughtsbot through an acquisition and so Nail, MODE and Thoughtsbot both the teams

[00:03:03] and the products are coming together is kind of combined BI tools.

[00:03:06] So at MODE I was like said one of the failures and focused a lot on basically product and

[00:03:11] marketing there for the roughly ten years that we were an independent company.

[00:03:17] And for people that are hearing about Thoughtsbot for the very first time, I'm understanding

[00:03:21] is a tech company that produces business intelligence analytics search software but what would

[00:03:26] you all innovate a pitch-bate for Thoughtsbot for anyone listening and just hearing about

[00:03:29] you guys for the first time?

[00:03:31] Yeah so it is in a lot of ways a classic BI tool and since you want to build a dashboard

[00:03:37] and reports and all the sorts of key analytics that you need to run your business.

[00:03:41] But the real innovation that Thoughtsbot has been developing quite some time is this kind

[00:03:45] of natural language search-based interface that obviously is also really expanded with

[00:03:51] the GNI developments over the last few years but Thoughtsbot was kind of pioneering a lot

[00:03:55] of that before the hype cycle of chat GPT and all of that.

[00:03:59] So essentially, the idea is that lots of folks around businesses have questions.

[00:04:04] Typically, those questions need to get answered through data teams or through BI teams or

[00:04:10] you know, computer folks that are asking a question they go pull a report for them, send

[00:04:13] them an Excel file, things like that.

[00:04:15] Over the years, we've developed a lot of solutions for people to be able to self-serve

[00:04:19] that stuff themselves through like drag and drop interfaces or sort of build your own

[00:04:23] report, wzwig type things and the sorts of things that make it easier to answer those

[00:04:27] questions without code.

[00:04:28] And Thoughtsbot's kind of real innovation on top of that was instead of having a UI where

[00:04:33] you drag and drop fields into different areas to build a chart, what if you could just ask

[00:04:37] natural language questions.

[00:04:39] And so that's been something that we've all been working on for quite some time.

[00:04:44] Again, obviously expanded a lot with some of the new developments in GNI but the core

[00:04:49] philosophy there is the same as can you just ask questions, the kinds of business questions

[00:04:53] you would be typically asking a data team of how many orders did we have in California

[00:04:58] of this particular product or can I drill into those orders and look at them by store

[00:05:02] as kinds of conversation questions that you would have to be able to help understand

[00:05:06] your business.

[00:05:07] If you could just ask them of the product directly and it'd be able to infer what you

[00:05:11] mean and go and fetch that data and help you build the charts in a port you want that

[00:05:14] way, that's kind of the core power of what Thoughtsbot provides in this space.

[00:05:21] And over the years we've had a couple of people from Thoughtsbot on the podcast.

[00:05:24] So me I think it was the chief development officer and Cindy Howson, are they still there

[00:05:29] as well?

[00:05:30] Yes, both of them are.

[00:05:31] Both of them are still very key parts of what we're building.

[00:05:35] Fantastic.

[00:05:36] Well, for the conversation today, of course I don't know all anyone's talking about out

[00:05:39] there is AI or started last year, he's continuing into 2024 but what put you on my radar

[00:05:45] was talking about the AI paradox in product development because I read that you mentioned

[00:05:50] that AI could potentially make products work.

[00:05:52] So can you elaborate on these paradox where the ease of building AI back products might

[00:05:57] not always equate to quality or improve user experience because I think this is a

[00:06:02] tale that we don't get to hear enough.

[00:06:04] Yeah, for sure.

[00:06:05] So there's a couple of ways I think that AI could at least for some time make things sort

[00:06:10] of messy.

[00:06:11] So one is obviously there's a big hype cycle around it now.

[00:06:15] There were AI commercials in the Superboy yesterday and so everybody's sort of

[00:06:20] focused on that.

[00:06:21] There's lots and lots of numbers and things you could look at about how companies that

[00:06:23] mention AI and their earnings calls and things like that or their stocks are doing better

[00:06:27] and startups are raising money on crazy valuations when they put AI in their pitch and all those

[00:06:32] kinds of things.

[00:06:33] And so lots of people are essentially trying to market themselves today, I businesses

[00:06:39] in lots of products that are already existing or wanting to be able to show that they're

[00:06:42] thinking about AI and thinking about how they're going to bring AI into their product

[00:06:46] and those kinds of things.

[00:06:48] And so I think what's happening is a lot of folks are essentially trying to tack things

[00:06:51] on that if you have a piece of software or you have something that you're running and

[00:06:57] you want to bring AI to it, it's relatively straightforward to kind of tack on chatbots

[00:07:01] and those kinds of things.

[00:07:03] You have a lot of these little chatbots to control the user interface or suppose you want

[00:07:10] to sort of like chat with our support docs and you can ask it a question and we'll go

[00:07:14] and kind of fetch information from their support docs and those kinds of things.

[00:07:17] And so there's a lot of places where people are trying to glue AI on to products, not

[00:07:21] necessarily because they're solving a direct problem with it but more to be able to say,

[00:07:25] hey, look at us.

[00:07:27] We have some AI feature and that kind of thing.

[00:07:29] And so I think there's going to be places where those additions aren't really necessary

[00:07:35] or they aren't sort of thought through is how are we actually going to help the end user?

[00:07:39] They're more there to be able to have a big AI banner somewhere to be able to say, hey,

[00:07:44] we're thinking about the stuff, we're thinking about the future, all that kind of thing.

[00:07:47] And so there will be places where I think we just get features that don't really work

[00:07:50] that well, that we don't really need, that people are going to build things that sort

[00:07:54] of fit into the chat GPT paradigm because that's the way that a lot of people have been exposed

[00:07:58] to generative AI initially.

[00:08:01] And so it's going to be some things that those will probably work really well at some

[00:08:04] of them, no, just because I think they're again people are rushing out the features more

[00:08:08] for the feature sake than really understanding whether or not it's a thing that their

[00:08:12] customers want if it solves a problem, that kind of stuff.

[00:08:14] So I think over time that will get sorted out.

[00:08:16] Some of those things will eventually go away, we'll move on to some other I'd cycle

[00:08:19] and not everybody will be chasing AI and we'll be chasing the next thing, we'll be chasing

[00:08:23] vision pro apps or whatever.

[00:08:26] And so then there'll be a little bit of a cool down phase where the people who really

[00:08:30] are building the lasting AI features will build something great.

[00:08:33] And some of the other ones may end up kind of fading away if they were just these kind

[00:08:37] of tacked on things that were chasing the hype cycle.

[00:08:40] And there's an XIT guy, one of the things I've always loved about Thoughts

[00:08:44] but is this no BS around shiny new technologies and hype cycles etc.

[00:08:49] It's always been about a strong emphasis that begins with real customer problems rather

[00:08:54] than technology first and it's so refreshing to hear that.

[00:08:57] So if we take that school of thought into the world of AI, how do you integrate

[00:09:01] this customer-focused approach in your AI powered analytic solutions?

[00:09:06] And why do you think it's so crucial in avoiding the pitfalls of AI in product development

[00:09:11] very, very early on?

[00:09:13] Yeah.

[00:09:14] I mean, I think so again, it's easy to chase these are the things that we want to build

[00:09:18] because AI features and stuff because they're hyped and because they will get attention

[00:09:22] and you can make splashy launches with them.

[00:09:25] There was a thing that came out I saw a couple days ago about Amazon built blockchain

[00:09:30] features and they launched like a hosted blockchain feature and all the people inside

[00:09:34] of Amazon were like, we don't actually think there's any value in this but people are buying it

[00:09:38] and so I guess we'll sell it just because it's a thing that we can sell on the hype.

[00:09:43] And so I think it's easy to do that sort of stuff and our focus is mostly on

[00:09:47] given that Thoughtsbot well before me or the Mode folks point has been about how do we bridge

[00:09:55] this kind of what is what was called best then?

[00:09:58] Like the generative AI type of experience where you can ask these now language question

[00:10:01] it can kind of infer what you mean.

[00:10:03] You treat it very much in the way that you treat a lot of these generative AI technologies

[00:10:07] but it was before that, that was the way we all talked about it.

[00:10:10] And Thoughtsbot was probably not going to be thinking about like, okay, what's the great

[00:10:12] experience?

[00:10:13] What's the way that customers actually want to use it?

[00:10:16] Let's not try to sell that technology.

[00:10:17] Let's try to sell like a better experience for what people want.

[00:10:21] Now is it a conversational interface?

[00:10:24] Is it something where it's more automatic alerts?

[00:10:26] Is it telling you things you should be looking at when you need to look at them?

[00:10:30] There's a lot of places where AI can help you better understand your business and those

[00:10:34] sorts of things but it's not just going to be like chat with your data exclusively.

[00:10:38] There are going to be other ways where that gets integrated.

[00:10:40] And so I think a lot of it is starting with the customer problem,

[00:10:43] starting with what things people want to do,

[00:10:46] starting with what's the ideal experience that people have in a technology agnostic way

[00:10:51] and then figuring out if the technology can actually solve that problem for you rather than saying,

[00:10:54] hey we have this fancy new open AI API.

[00:10:58] Let's forget how to plug it in.

[00:10:59] It's starting very much with understanding customers, understanding what they're struggling with.

[00:11:03] The place is where they really need this technology to help them and places where they don't.

[00:11:08] A lot of it is the cliche answer and these sorts of things is just like you got to talk through

[00:11:11] a lot of customers, you got to pay attention to what they're saying,

[00:11:15] you got to really listen to them and sit next to them and that's sort of the solution for all of

[00:11:19] this and I don't think any technology really changes that fundamental strategy and building stuff

[00:11:24] that's now a quote unquote customer-facus.

[00:11:27] And you're also a huge advocate for thinking outside the bot which I think is a great

[00:11:32] line. Can you explain that concept and how that mindset guides thoughts,

[00:11:36] thoughts, development and innovation in the field of AI and analytics?

[00:11:40] Yeah so this is one of the things to me that's interesting is like

[00:11:43] the thing that blew up,

[00:11:45] generative AI was Chatchy PT.

[00:11:47] Like obviously that was sort of the moment where it became a mainstream thing

[00:11:50] and suddenly the entire world was talking about it and stuff like that.

[00:11:53] And obviously Chatchy PT was a chatbot.

[00:11:56] You know you chat with it like you're texting a friend or whatever.

[00:12:00] And so I think a lot of people have started to quake generative AI with chat bots and like LLM's

[00:12:07] with chat bots and that's kind of the natural way that we now tie those things together

[00:12:12] and so a lot of folks I think when they think about how do we add generative AI capabilities

[00:12:18] to a product, a lot of it is centered around chat bots or experiences that are kind of

[00:12:24] chatty in that way. And I think there will be places where that works really well but

[00:12:30] kind of my earlier point about you know products getting sort of generative AI features tacked on.

[00:12:36] The easy thing to do is add these kinds of chat bots because you can build them on top of

[00:12:40] APIs that they're pretty straightforward to work with. But I don't think that's necessarily the

[00:12:44] right way of always using generative AI technology. One of the examples of this that I was talking to

[00:12:51] one of the designers on note about was if you're a designer, if you're like a web designer

[00:12:55] or you're trying to design a web page, a chatbot doesn't help you there. What you want is kind

[00:12:59] of a sketch tool where you can sketch something with pen and paper and then have generative AI

[00:13:05] basically fill in the details for you sort of do all of the tedious stuff that you typically have

[00:13:09] to do yourself where you can kind of express your idea through drawing and then have generative AI

[00:13:16] do the rest of like the kind of tedious work that you typically have to do to make it pixel perfect.

[00:13:20] And so there's no chat bot there but like generative AI is a key part of that experience. And so

[00:13:25] I think as an example of a place where like using generative AI in ways that actually fit into

[00:13:30] the workflow of a user fitted to how they think is really important because that makes a much better

[00:13:34] experience than having a designer trying to like describe the web page that they want to they

[00:13:39] want to design that obviously would be a much clunkyer thing than in this sort of sketch tool. And so

[00:13:44] for instance, an example is the difference side of inside of thoughts bot mode. We recently released

[00:13:48] a feature inside of mode called AI Assist. It's a and within mode there's a code editor for you

[00:13:54] to be able to write SQL queries to run against your database. In this case we don't we don't want

[00:13:58] to help people like in a chatty way for that because there are analysts who are writing code like chat

[00:14:03] bots don't really make sense there. We said when I followed this model that this designer described

[00:14:07] where like can you sketch out the thing you want to be able to build? Can you write kind of pseudo code

[00:14:11] and not have to do some of the tedious stuff about making sure all these syntax is exactly right but

[00:14:15] basically express the query that you want to to mode and sort of a shorthand where you can say okay

[00:14:21] join this table do this thing. I don't remember exactly how to join it but like figure that out

[00:14:26] basically try to outline the thing that you want to do and have it do all of the all of the tedious

[00:14:31] stuff and that's the sort of thing I think that can really accelerate those workflows for analysts

[00:14:34] and it's not chat with it and tell it how to write a query. It's kind of express the query in a way

[00:14:40] that's faster and more effective than having to type it all out yourself. And so I think all these

[00:14:44] sorts of ways are eventually how how generative AI appears and analytics or appears and even

[00:14:49] broader technology said is as places where it's not so obvious that you're chatting with a bot

[00:14:55] like it may not even be generative AI to a lot of users and may just be this the feature that

[00:15:00] just sort of works and I think over time that's what we'll see is less less focus on chat bots

[00:15:06] and things that are so clearly generative AI features and things that are just like experiences

[00:15:11] that suddenly seem to to just work super well in a way that that we're all comfortable with

[00:15:17] and kind of fit more naturally into into the regular workflows that we're trying to try and improve.

[00:15:22] I think just feel we're a slightly odd time at the moment there's so much economic

[00:15:27] uncertainty and at the same time everyone wants to go all in an AI and every tech project

[00:15:32] though is going to be placed on the screw any full ROI business value it might generate. So when he

[00:15:38] comes to investing in quality AI solutions what does real investment mean in the context of AI

[00:15:43] development for Thoughtsport and how do you balance those technological advancements would that

[00:15:49] need for creating genuinely useful product? I think a lot of this is recognizing that this is a

[00:15:54] long road that like yeah there's a lot of stuff that happened really quickly that one of the

[00:15:59] reasons I think these kind of tacked on chat bot types of features will will make some products worse

[00:16:06] is you can build kind of a basic version of these things pretty quickly you can build demo

[00:16:12] where pretty quickly. The one thing that analogy I use for this is like self driving cars so

[00:16:16] self driving cars were a thing that 10 years ago we had like it felt like we were on the cusp we

[00:16:22] had cars that could kind of drive on the highways and like are these things are getting pretty close

[00:16:25] and there were a bunch of companies and businesses that got started up around the expected patient

[00:16:30] that there were going to be self driving cars like any year now. And I think what follow up it

[00:16:34] realizes like okay you can actually do a lot of stuff pretty quickly with self driving cars but

[00:16:40] but there's a bunch of these strange edge cases and very difficult problems in itself driving cars

[00:16:45] case can become very dangerous obviously if you don't get it right. And so all of the all of the work

[00:16:50] is actually in that last little bit and they're not a functional product until you can actually get

[00:16:56] that last little bit. So it's sort of an inverted like 80 20 rule where 20% of the work can solve 80%

[00:17:03] of the problem but you can't actually solve any of the problems so you solve all of it because

[00:17:07] you can have a self driving car that like you know can't handle when it tries to pass a school bus

[00:17:12] and all the work isn't figuring out those those are details that that you have to get right to make

[00:17:17] this product functional. I think a lot of AI is along those same lines where it's pretty straightforward

[00:17:23] to be able to say I want to ask this chatbot a data question and it's a simple data question on

[00:17:28] top of relatively simple datasets and it will generate something that looks like a very good answer

[00:17:33] but a lot of businesses don't want to do that they want to ask much for complicated questions or

[00:17:36] they're asking questions and they need to be right because it's a question that they're reporting

[00:17:40] to their board or they're reporting you know for earnings calls or whatever and you have to make sure

[00:17:46] you're getting those complicated questions right too and there's a ton of stuff that goes into

[00:17:49] making that actually work and so I think a lot of ways that the real investment in this is

[00:17:54] just recognizing that this is at a oh we're going to go build something and it's going to be great

[00:17:59] and then we're going to move on to the next thing it's like if you want to build an AI powered

[00:18:03] product you have to be in that for the long haul and in some words of if you want to build a cloud

[00:18:09] based product that's not a decision that you can ship in six months and then move on to whatever's

[00:18:13] next that's like that's now your business your focus of your business will forever be making sure

[00:18:18] that you can run this thing in the cloud successfully and so I think a lot of AI products are going to

[00:18:22] be the same where you know to be able to build something that's great on top of this technology

[00:18:26] it's going to be a thing that you have to be committed to for for the duration of however long

[00:18:30] that you're supporting that product um and continuing to push it and make it better and solve these

[00:18:34] weird edge cases and all this kind of stuff so really a lot of this is just it's just the work it's

[00:18:38] a commitment to believing that that's the right solution and and continuing to push that technology

[00:18:43] forward instead of thinking it's it's a feature to build and then you move on to the next thing

[00:18:48] managing expectations in AI product offerings can also be difficult when we're

[00:18:54] right in the heart of this hype circle that we've mentioned a few times so regarding the

[00:18:58] resolution to avoid overselling and under delivering how do you at thought spot and show AI powered

[00:19:04] analytics platform that you offer how do you ensure that that meets and exceeds customer expectations

[00:19:10] because it must be difficult doing just that right now with so much expectation and hype around it

[00:19:16] yeah and I think I think you should have to be honest about it like I think that's that's the

[00:19:20] the important thing when you're talking to customers about this sort of stuff is there's people

[00:19:23] getting sold so much hype yeah um that that you have to be honest and say hey here are the limits

[00:19:29] of what this can do here are the places where it actually isn't going to work where we don't

[00:19:32] recommend you use it in this way um and and you know sort of have customers with you on that

[00:19:37] journey as you continue to get it better and say hey you know these are what we can do today this is

[00:19:40] what we think we're gonna do tomorrow but we want you to be a part of that process and enter

[00:19:44] really work with you to to make it good um you know you know it thoughts about has a number of customers

[00:19:48] that are that are exactly in this this camp that are kind of these innovation partners for us

[00:19:53] the leading gaming company who recently came on board for for doing exactly this some large financial

[00:19:58] services firms in the same way and looking to thoughts bought as basically a kind of partners

[00:20:04] in figuring out how they want to bring gen AI into their businesses that you know we obviously have

[00:20:09] a lot of expertise in that space um they understand their business is very well they understand

[00:20:13] the nuances than the like I said the sort of the difficult driving like traffic situations

[00:20:18] from the self-driving car analogy like they know exactly where those things are are difficult to navigate

[00:20:23] and so there's some recognition that hey we just got to work on this stuff together but I think

[00:20:26] I think the main thing you have to avoid is it's easy to get caught up in the in the hype and

[00:20:31] to sort of promise like hey yeah this technology is amazing it can do all these great things

[00:20:35] and there will be lots of people out there that say that um and and you sort of have to recognize

[00:20:40] like look that's that's not one not what's going to be true but two a lot of customers appreciate

[00:20:46] the honesty and the the sort of being part of the journey uh to figure this stuff out together um

[00:20:51] and if you know you sell something that's gonna sort of the magic wand that solves all their problems

[00:20:57] here's sort of headed for nothing but disappointment at some point

[00:21:00] an accessibility is a really huge talk keep point right now and I know it's thought spot you

[00:21:05] aim to make analytics accessible to everyone within an organization so what's the secret there

[00:21:10] how does your platform how does your platform enable users with completely very technical

[00:21:15] background some with very little to engage with complex data effectively because it's quite a

[00:21:20] balanced isn't it and and you know accessibility is obviously one of the sort of the key

[00:21:27] needs really any kind of be a tool uh that's the point in a way is it's your trying to solve

[00:21:32] these kind of self-serve problems you're trying to help people who who aren't necessarily

[00:21:37] accustomed to answering questions on their own get answers that they need are people who have like

[00:21:41] he said it's very degrees of technical capabilities um and those those degrees of technical

[00:21:48] capabilities often matter because some people really want to be able to write out uh it's

[00:21:52] important that if you're you know if you're an engineer and you're trying to build a website you

[00:21:55] probably don't want to use square space you want to be able to write code the build it because

[00:21:59] you have that capability and you can build a better website than you could with the dragon drop tool

[00:22:03] and this is a true for data if you're an analyst that knows how to write SQL or Python

[00:22:07] your ceiling is kind of much higher than what you can get out of most uh dragon drop kind of

[00:22:11] BI tools and so you actually want to have that that capability but the way we see this is is

[00:22:16] essentially trying to one provide a sort of range of different interfaces for people depending on

[00:22:20] where they are and and kind of their their technical ability or the kind of questions they want to ask

[00:22:26] you know well on one end there's the kind of conversational experience that we've talked about

[00:22:29] where you could just ask questions in English and thoughts about will help you figure out how to

[00:22:32] answer those um on the other ends you can write you know Python and R and do sort of advanced data

[00:22:38] science and and those sorts of tools are those sorts of uh solve those problems in inside of mode

[00:22:44] and that's one of the reasons that mode and thoughts about came together was to build a

[00:22:46] provide that kind of full spectra in like of technical experiences uh in in a single place um the other

[00:22:52] part though I think it's important and this is to be able to move between those different pieces so

[00:22:56] one of the things I think for accessibility especially in an AI or a BI excuse me that's tough

[00:23:01] is do you ask a question in natural language how many charts did we sell in California or whatever

[00:23:07] you get an answer you want to be able to know that's right like one of the things that's really tricky

[00:23:11] is if it just says we sold 50 shirts like how do you know to trust it um how can you verify it how

[00:23:17] do you make sure that that's actually accurate um and so there's a lot of things that we've built

[00:23:22] around that as well where you can see uh sort of stylized versions of the SQL queries that got written

[00:23:27] you can see basically how the the how thoughts about AI took your your English question

[00:23:32] the kind of tokenized version of that question where you can sort of read the essentially similar to

[00:23:37] like reading Excel formulas um so you can kind of parse your way through through the logic of it if

[00:23:41] you want to see the actual underlying SQL to do that you can get a step of take a step further down

[00:23:46] and so trying to give people different ways to to reason about the questions at the level that

[00:23:50] they're comfortable in and so a lot of that comes in these again these just ways to validate

[00:23:54] making sure that you're not you're not as black box that you're just like throwing answers at people

[00:23:58] um and so I know this is this is again kind of the point of like working with AI is real work

[00:24:04] all these things are what make make these products capable and functional and things like that

[00:24:08] like if you are a SQL chatbot where you ask a question and you just give an answer even if you get

[00:24:12] the answer right most of the time and people have no way to verify it uh they're not going to be

[00:24:16] terribly happy with that experience because into I want to go report to my boss some number that

[00:24:21] that I asked a bottle of and I'm 90% sure it's right but I don't know for sure and I don't have

[00:24:26] a way to validate it so there's a lot of those kinds of experiences that are really important in this

[00:24:30] if that part of the core experience you'd see in the 30s I can dinner uh but if the part of

[00:24:34] experience that makes the product something that people can really trust and rely on

[00:24:38] in these in these really important business cases and when we talk about things like ROI

[00:24:43] business value and decision making enhancement is anything you can share around or maybe some insights

[00:24:48] on how thoughts but AI powered analytics contribute to improving decision making processes

[00:24:54] in a business and also the kind of real world impacts that you've observed because I think

[00:24:58] this would be so helpful for people listening thinking how could all this work in my world what

[00:25:02] would it deliver is that anything you can share on the kind of impact that you've seen

[00:25:07] yeah so I can share a couple things from from you know folks and customers that have approved some

[00:25:11] some stories here um more generally uh a lot of the ways that you know thoughts about drives these

[00:25:17] impacts across most of our customers is providing this kind of accessibility to to folks

[00:25:21] trying to business to answer questions really quickly um the one of the core kind of value props

[00:25:27] essentially of both mode and and thoughts about a lot of ways is delivering answers really quickly

[00:25:31] this kind of time to value of like you have a question how quickly can you turn it around um how

[00:25:35] quickly can people get answers uh you know make can you make better decisions with with these really

[00:25:39] fast answers is really kind of the goal here and so I think again broadly when we make decisions

[00:25:45] just in our lives today Google makes that really easy it makes it really fast to get answers

[00:25:49] really quickly to things that may have taken forever but impossible before um the the kind of core

[00:25:54] philosophy behind behind thoughts about a lot of ways is is that same ethos of like how much better

[00:25:59] could we all be at the things we do if we could ask questions about what's going on in our business

[00:26:02] and get answers immediately um and so yes so there's there's things with with uh this company called

[00:26:08] MD audit um that's a premier health care organization provider for healthcare organizations um integrated

[00:26:15] with thoughts about sage which is which is thoughts about it's the the conversational tool uh to bring

[00:26:20] this kind of natural language uh a conversational bi to to their their employees um so they've been

[00:26:27] able to do this uh and and really minimize a lot of like billing risk because of the ways that

[00:26:32] people have been able to use it and get the questions that they need answer really quickly um the other

[00:26:37] big thing that mode does is are excuse me mode anthems about do to our embedded solutions where you

[00:26:42] can basically embed these these BI solutions to customers and the partners um and so uh come to go

[00:26:49] to act on which is a marketing automation tool uh uses thoughts about everywhere which is which is

[00:26:53] the embedded solution provided by thoughts but um to be able to to provide reports to their

[00:26:58] customers to provide this kind of exploration tools for for their end users um and gotten much higher

[00:27:04] adoption uh across those kind of analytics experiences for their customers um and so now people

[00:27:09] are spending uh basically twice as much time uh on on their pages to do things what they need with

[00:27:14] their data to answer these questions and stuff like that so there's a lot of things like that kind of

[00:27:18] kind of experiences uh where people are now just like engaging with data a lot more able to sort of

[00:27:22] be more generally aware of what's going on in businesses not just for the sake of entering individual

[00:27:26] questions they did answered but just also a this kind of ambient awareness that comes along with it

[00:27:30] when when you can answer your questions that you have to quickly and looking ahead if I was

[00:27:34] to ask you to look into my virtual crystal ball any predictions for the evolution of AI in

[00:27:40] the analytic space anything you see the excite shoe or how you see AI maybe shaping the way

[00:27:46] companies interact with day two in the next few years I realize it's an impossible question to

[00:27:50] rescue about with the speed of everything but anything that excites you will that you've seen

[00:27:54] taking shape. The one thing we haven't talked about here that I think is we'll see over time is

[00:28:00] basically unstructured data starting to take off a lot um so historically sort of broadly there's

[00:28:07] been two kinds of data there's structured data and unstructured data structure data is basically

[00:28:11] tables of numbers uh some kind of stuff you would open up an Excel and you know do math on and

[00:28:16] compute you know a list of of sales and compute the total sales by month or whatever with

[00:28:22] unstructured data is basically everything else its images it's documents it's it's text it's

[00:28:26] emails it's all this kind of places where we're some of the information and that stuff but it's really

[00:28:31] hard to get at it's hard to access um like if you were give me a giant trove of emails and said go

[00:28:37] figure out things that you can learn from this and be very hard for me to do it because you have to

[00:28:40] basically read the emails and then kind of aggregate them up in your head uh and think about what they

[00:28:45] mean like it's hard to summarize this stuff because we have to read it all um one of the things that

[00:28:50] that genai does and and lm specifically is they basically become these kind of aggregate functions for

[00:28:56] text where you can feed it I know a thousand emails or a thousand help docs and say can you tell me

[00:29:03] what these things mean can you summarize them for me can I give you a thousand support tickets from

[00:29:07] happy customers that thousand support tickets from unhappy customers and say like what are the

[00:29:10] differences between these two things and they can give you five bullets to explain those differences

[00:29:14] and so then to the like genai has the potential to make unstructured data much more accessible

[00:29:19] and much easier to work with and I think if that happens uh one of the things that we'll see is a lot

[00:29:24] more sort of technology and tooling and focus on actually finding value in that that unstructured

[00:29:29] data right now it's like there's information in those support tickets or those emails or those

[00:29:33] those documents uh but because it's so inaccessible we often kind of like treat that as a second class

[00:29:39] data set compared to the structured data once it becomes more accessible and we sort of have the

[00:29:43] the tools for mining it if you will uh I think there's suddenly a lot more focus on

[00:29:48] on trying to draw insights and and information and decisions out of all of these giants structure

[00:29:54] unstructured data sets in the same way that we've spent a really long time trying to help us all make

[00:29:58] sense of unstructured data and speaking of making sense of everything I think there's a

[00:30:03] pressure on everyone to be in a state of almost continuous learning now so I've got to ask a

[00:30:08] simple let's write in the heart of this space and there'll be a lot of business leaders listening

[00:30:12] thinking how do I keep up to speed with the pace of technological change where or how do you

[00:30:16] self educate anything you can share around how you keep up to speed with everything so

[00:30:21] that part of it is you know find new sources you can just read like I think that there's a

[00:30:25] handful of really good sources of news and things like that thing are the kind of common ones or

[00:30:31] things like Bentonson's blog is really good um there's if you finance ones out there uh

[00:30:36] honestly just reading tech publications uh things Wall Street Journal and and sort of tech crunch

[00:30:43] uh RIP a little bit um things like that are good for just following the news but honestly I think a

[00:30:48] lot of it is is it's not you have to sort of like saturate yourself and it's something that there's

[00:30:54] not gonna be here is the the five bullet summaries of everything I think it's it's it's very difficult

[00:30:59] to actually learn anything from things that are these really sort of distilled aggregations I think

[00:31:04] that this is one of the challenges of things like GNI is we don't actually learn that much if

[00:31:08] we just read these sort of like really succinct summaries I think actually a lot of the best stuff comes

[00:31:12] from longer form like I think it's more useful to read one longer form piece than it is to read 10

[00:31:18] little short here's the five takeaways that you should learn from the news kind of thing um

[00:31:23] you're not necessarily learned as many facts uh from the longer form stuff but it's the longer form

[00:31:27] stuff where you actually get much deeper uh like conversations about things you start to understand

[00:31:33] better ways to interpret what's going on you start to read much more interesting things and can

[00:31:36] kind of see things with more clarity when you're actually in these kind of longer form pieces so

[00:31:40] my general like recommendation for folks on these things is spend less time trying to find the short

[00:31:46] snippet like on the go aggregations that you can read and in a tweet um and spend more time

[00:31:52] finding a handful of good in-depth analysis type of pieces because once you have that sort of a

[00:31:59] uh you know teach a man to fish versus give a man to fish if you give a bunch of these little snippets

[00:32:03] you're kind of constantly just being given fish if you read these longer things you're giving

[00:32:07] fewer fish but but the way of thinking that's that's often presented on my think is much more

[00:32:11] valuable and and ultimately kind of teaches you about a fish and interpreting the newsers. No

[00:32:15] fantastic advice absolutely love that and for anyone listening wanting to find out more

[00:32:20] information about sorts but maybe they want to dig a little bit deeper on anything we talked

[00:32:24] about today where would you like to point everyone listening to find out more information

[00:32:28] not many different um domain names now where would you like to point everyone? Yeah um so

[00:32:33] thoughts bud uh if you want to find more about a thoughts bud it's just thoughts bud.com um

[00:32:37] and if you want to start a trial like better free trial there thoughts bud.com slash L uh

[00:32:42] no uh isn't there you can get through thoughts bud are also mild prompt uh has has more detail

[00:32:48] about about know specifically uh and then on link and and and twitter slash dex uh you can follow

[00:32:54] on for updates there uh which is just a thoughts bud so so easy to find thoughts bud uh no

[00:33:00] no credit domains or crazy uh handles just just thoughts bud uh everywhere basically

[00:33:06] uh is is the easiest way to get there. Well so I'll add all the links to the show now so

[00:33:11] people can find you and I see you and so many big takeaways around how AI will yes

[00:33:16] transform thousands of products some for the better some for the worse but if you start with a

[00:33:21] real customer problems think outside the box real results will require real investments and also

[00:33:27] of course avoid overselling and under delivering so many golden points there but just a big thank

[00:33:32] you for showing that with me today thanks again bud. Yeah for sure thanks very much. Now as we wrap

[00:33:38] up today's enlightening discussion with Ben Stantzill from Thoughtsbot I think it is clear that

[00:33:44] the journey towards an AI enhanced product is through with both opportunities and equally pitfalls

[00:33:51] because Ben and Ben's insights reminded me that the true innovation starts with the real customer

[00:33:57] problem and requires a thoughtful approach into how integrating AI will add genuine value to that

[00:34:04] and it's about thinking outside the box again absolutely love that line and making real

[00:34:09] investments and avoiding the all too common pitfall of overselling and under delivering it's something

[00:34:15] we see time and time again but for anyone eager to dive deeper into the world of AI powered analytics

[00:34:21] and learn more about Thoughtsbot's mission to create a more fact driven world. I could you to check

[00:34:26] out their website thought at Thoughtsbot.com and also uh the blog post that will accompany this

[00:34:32] episode at tech blog writer.co.uk go to podcasts you'll find a blog post associated with this episode

[00:34:38] which will include all the links that Ben mentioned and a few others to feed a jack out but over

[00:34:44] to you let me put the microphone in front of everyone listening this isn't just a case of passively

[00:34:48] listening what is your take on that AI paradox in product development you've heard from me you've

[00:34:54] heard from today's guests but I'll empty your experiences is so critical in this discussion so email

[00:35:00] me tech blog writer outlook.com twitter linked in instagram just at Neil's CQs let me know your thoughts

[00:35:06] on that but that's it for today so until next time keep innovating keep questioning stay curious

[00:35:11] keep pushing the boundaries of what's possible and I'll meet you here tomorrow same till

[00:35:15] place take care guys