AI Announcer (00:10)
Welcome to Prompt This, the ultimate podcast for business leaders who are done with the AI hype and craving the real deal. I'm your AI intro voice for this week. And let me tell you, it's going to be a game changer. Meet Greg, your seasoned guru for scaling sales teams and cracking that elusive Silicon Valley playbook. And then there's Clint, a startup veteran who has a knack for transforming massive ideas into thriving ventures.
They slice through the chaos to deliver you the analysis and playbooks you need to harness AI for launching new ideas, scaling your business, and most importantly, not getting left behind. And now, without further ado, here are your hosts.
Greg (00:58)
Welcome to the Promptus Podcast. This is Greg. And today we have a great episode. Our guest articulates an approach using AI that can improve deal coaching to maximize its effectiveness. Something, you know, in a way I haven't thought about it before.
Clint (01:15)
And this is Clint. Yeah, our guest did a great job of really kind of stripping away the hype and mystery of AI technology, giving really solid practical examples about how to think about AI in your business and where to get started. He made a really insightful point about how we are all still at the front end of figuring out how to use AI in business. He said, everyone thinks they are behind in adopting AI, which means no one is.
Greg (01:43)
Let's get into the episode.
Clint (01:45)
Let's do it.
Greg (01:50)
Today we're joined by Warren Cucker, founder of the software company Basic and creator of Topic, a sales intelligence product built to capture and unlock the value hidden inside customer conversations. Warren has spent years operating inside sales, product, and go-to-market teams and has a sharp point of view on why companies are still wasting their most valuable asset, what their customers have already told them.
His core belief is simple but provocative. AI isn't a feature and it's not a product category. We're going to dig into that idea more today with him. Warren, welcome to the show.
Warren Kucker (02:31)
⁓ Great to be on. appreciate you the time.
Clint (02:33)
You bet. It's great to have you here. So I'd love to just kind of dig into it right away and get to get to know you better. And let's start off with maybe just a little bit of a background on your career and what led you to building topics and starting your own company.
Warren Kucker (02:47)
I'll try to keep it brief. basically spent about 20 years in shipping and logistics. On the corporate side, that was like UPS and Apple. UPS side was like industrial engineer, process engineer, hat, clipboard, how do we save a minute on every driver's day kind of scenario. Absolutely loved that job. It was a good like foray into like what the real supply chain was like. Apple was logistics procurement. So they brought me in because they had a notion that they had hidden shipping costs throughout the company.
What that looked like was they had a big team shipping iPhones, iPads, MacBooks. They had no one looking at when we were buying servers. We might have the best IT procurement person on the planet. Can get down to the penny on a negotiation on what a server should cost. But when then it came to shipping, had absolutely no clue. And where that's important was that two and a half years there, it turned out they had about $500 million in hidden shipping costs. By internalizing that, bringing that in-house, we those costs by about 38%. So I saved by myself about $60 million my first year.
After three years, was hearing that people weren't doing this, so I launched a startup to do the same for Fortune 500 procurement teams. And that was the launch of my sales career. I never wanted to be in sales. I told former bosses of mine, if you ever put me in sales, I'll just quit. But start a company, you gotta be in sales. We were selling software to Fortune 500 procurement teams. That's where I kind of learned to transition from founder-led to actual growth sales. And this kind of become my niche over the past couple years.
Duct out of shipping for a year in dental practice management, which is really interesting and back to starting my own startups now as of about six months ago.
Greg (04:20)
You know, you have a unique angle when it comes to AI and how you describe it, how you describe it within your company and you know, it's not just AI powered. You talk about AI differently. Tell us a little bit about that.
Warren Kucker (04:35)
Yeah, so at its highest level, one, everyone on the planet thinks they're way behind in AI, which means that no one's way behind. Every executive I talk to thinks that they're the ones in their industry, their market, their niche, that are falling behind their competition. No one's really leveraging it across the board yet, which is great. There's a lot of opportunity to grow there. So that's one. But two, I think what we really have to get to, especially like, let's say, make the evolution now, is that there's no such thing as like an AI product anymore. Everything...
has AI built into it.
Clint (05:07)
So tell us about the problem you're solving with your new company. What is it that you're looking to make better?
Warren Kucker (05:14)
Absolutely. So I think the key is, from my perspective, there's two things that LLMs do really well. They allow us to get information out of unstructured data, and they write really well. And that's the scarier part of the two. So they do those two things. So when I talk to an exec, when you're looking for initiatives, what data sources should I be leveraging now that I couldn't leverage before? And for me, spending about seven years now on email technology, the communication part was the part that was most intriguing.
If you think about it, in a 30-minute meeting that you're having, whether there's two or ten or whatever many people there are, you're generating about 15 pages of content if you think about the size of a book, typical size of a book. 15 pages of content being generated in every single meeting we're having across the board. On top of it, meetings have increased by 250%. We've gone from two hours of meetings a day to five hours of meetings a day, right? So great, I'm in five hours of meeting a day. That means I was a part of 150 pages worth of content.
What does that happen in that five hours of meetings a day? I boil it down to maybe a one page document, a particular meeting I was in, if it was really robust, probably just bullet points in a sauna, maybe an email to a colleague, or most of the time it just goes out into the ether, right? That data is perfectly ripe for disruption because it's unstructured. I should be recording all of my meetings as an organization, and I should be using LLMs to ask key questions, not just what happened today.
But now I could be saying, what was my customer talking about three years ago that I couldn't solve for then, that I could solve for now? That unstructured data to me, the key is conversational data is really ripe for disruption. And so the problem we're solving is great. How do we leverage those conversations better? We have a broad mindset as to how we want to do this, but the best place to do it right now is in sales. Salespeople take meetings and send emails to close deals. We could make that much more efficient than we have in the past by leveraging LLMs right in the conversation.
Greg (07:10)
That's interesting. I think, you know, we've heard the beginning message a lot, right? You know, record conversations, pull sentiment out, you know, let's make some sales judgments. But this is one of the first times I'm hearing about, you know, keeping data around to make decisions and build product when you have the capability of doing it. So I like what you said, you know, hearing something three years ago is now actionable today.
because you have it and it's been preserved. this is not, so sales is like the entry point of listening to the customer and then you can use this data all around different departments of your company. Is that what you're looking to do?
Warren Kucker (07:56)
So there's that aspect to it, So let's talk about what's the knock-on effect. So product management in SaaS tech. The holy grail of product management is that you only build what your customers want you to build, and you work on the stuff that's most valuable to your customers and to the company. What's going to increase my revenue the most, right? What actually happens in product management is the squeaky wheel gets the oil. Whoever the loudest, the salesperson, the product person, the CEO, gets built what they want.
right? And it's because we don't truly do customer research like we'd like to. Every time there's a feature, product has to be going out to 20 customers, showing them Figma's, asking them their impact, and then scaling whether it's an important, is it going to prevent churn or increase revenue? And doing that every single time. We're barely doing that. Product managers are just putting out fires, making sure engineering isn't doing anything.
Greg (08:47)
being
told that that's happening all the time when I'm on sales side. Wait a minute.
Warren Kucker (08:52)
No, we have a prototype for product to do this for them. Based on all those conversations, we had a beta group of about 40 product managers. And what we learned was, if you know those products has no software, they use Google Sheets. Like it's not because they don't have budget. And it's because they're running around like crazy, just trying to put out fires and not truly talking to your customers. Well, great. Now with conversational data for products, you can have access to everything your customers and prospects are saying. Prospects are feature gaps.
Customers are what's really driving me nuts, but what value could be added. And really what should be happening is those features should be scored. From a CS customer success or sales perspective, how much revenue is tied to this particular feature, right? And then from a product standpoint, what's great is you turn that into a PRD, Where you product requirements document. You hand it over to engineering, but now the past is like at best, if you're really doing a good job, that PRD, that five page document, that's all engineering has to go off of, what to build.
You don't have to do that anymore. You can have that five page document really structured. But when engineering ⁓ has a question, they can ask all of the context of that feature from the conversations you had, both internal product and engineering and external with your customers. So the overarching point when I can't get away from the drain I keep on circling around is we need to be really focusing on using our conversations as an asset for the company that's been untapped. It's just like.
It needs to be continuing to be leveraged to drive value for the organization. You're paying your people to be in meetings for five hours and you're getting very little value out of that. If you attach it all together, you can really start to drive internal and external value for an organization.
Clint (10:33)
So aren't you getting that just by recording every meeting with Zoom and just having all that there, or are you going more than what Zoom will do for you or Teams will do for you?
Warren Kucker (10:44)
So you should be recording every meeting, which also from a standpoint happens less than you think. And even from a sales, there are less salespeople recording their meetings than you think. I'd say less than 50 % are recording their meetings. So one, stop that. Record all your meetings, internal and external. But two, you do get that. You get the transcript data, which Zoom has been providing for like five, six, seven years, right? We've been able to get the transcript. The difference now with LLMs, if you can connect that to, let's say, a vector database or some sort of database that allows LLMs to access that information.
in a robust way. You're going deep. ⁓ What you're talking about. And this would be helpful from like when you're vetting your software, when you're talking about AI, what's happening is when you're getting conversation in vector databases, you use an LLM to translate that conversation into numbers. You vectorize it, right? So and so what that is is the example I really like is.
Clint (11:14)
Vector database. Vector targeting. our audience. Yes, Cut up a level. Give us a little description of
Warren Kucker (11:41)
The vectorizing helps determine when someone says the word Buffalo, were they talking about New York or are they talking about wings? Based on the conversation, all the data around it, that's the core of what an LLM is doing. It's saying, great, this is Buffalo, New York, and this is Buffalo wings. And it gives it a number. Buffalo wings is 17,430 and Buffalo, New York is 84,321. And then when I ask a question about Buffalo and I'm talking about something in shipping, it goes to the 84,000 one instead of the one for 17,000.
So that's what you're doing. You're turning your conversations into context that you could use. And then you use an LLM again to ask that vectorized data questions. So then I'm like, great, I want to know about a shipment going to Buffalo, New York. It translates my question into those numbers, looks for all the data relationships, and then spits back and answer syllable by syllable or token by token. So point being is this. Your conversations used to just be flat file transcripts. You can search for a word, and hopefully you find it.
with no context around it. Now by pulling that data into an environment where you could ask the right questions and have all that data internal, you get to do a lot more with the information. So you change it from the index of a book to something that's supremely searchable with context.
Clint (12:52)
Okay, I'm following that. That's a good point there. not just words in a flat file that you're searching and hoping that you get the right meaning out of it, but you're actually turning it into searchable data with the right context around that. that's really kind of the space that you're unlocking is turning all those conversations. Is it email conversations, meetings, all of it? tell me where you're focusing.
Warren Kucker (13:16)
So for sales, the inputs are meetings, emails, and Slack messages or Teams messages, because a lot of selling is actually happening in these types of interactive Slack channels, basically. For product, so for sales, I don't actually really care about internal meetings that much. We record coaching calls and use those coaching calls to help drive what should happen next. Great example is when I'm coaching my sales reps, I never remember what the next steps are. They can come back, ask me about the same two deals, we could come up with the same next steps I would never remember from the prior week. We could stop doing that.
But mostly I don't care about other internal meetings. For product, internal meetings is where the weight is. So you may want to be recording those effectively. So those pieces of communication, if I was a CTO at any company, right now I'd say I need to find a way, it's not that hard, to get all of this information, record every single meeting, get the information, get it into a central database, and figure out how to use it later. That's what I would be doing from a CTO, CIO perspective.
Greg (14:12)
So if a company is doing none of this, where do you recommend the best place to start is? What side of the business?
Warren Kucker (14:20)
If we're like, we want to go like rudimentary, how do we get off the ground? Right? And let me take it like, ⁓ I'll be the VP of sales for a moment. And, but this example would be helpful for anyone. I'm not doing anything at scale. What I would be doing is I would be finding a way, maybe have a specialist go into zoom and take every transcript of every call. And I'd load them into GPTs and quad products and plot quad or GPTs and chat GPT.
And so a really good example is this, and I like this example because I think it allays some fears with where AI is going. I talked to a sales leader recently who has a custom GPT for every rep that they have. So they seven reps, they have seven GPTs. And before their coaching call, they take the 10 calls they were on, they push the transcripts in there. They take their coaching calls that they were on, and then they have a framework of where they are from a development standpoint, all in there, right? And from this, every coaching call they walk into, they have a customized...
because they have all of the context of the conversation they've had. They've had their last coaching. They have their current development timeline. And so now I'm actually, instead of using AI to become impersonal, I'm actually leveraging human relationships a lot better because of this AI. So if I was a VP of sales, this is what I would be doing. I'd have all my heads of sales, custom GPTs for every rep, pushing all that data in. You can then ask questions about deals. When you're doing forecast calls at a high level, with the executive team, could use all this data.
I would get really rudimentary, find a specialist, find a manager. Every week you paste this in, maybe you rotate it between all your employees, make it the chore, and just get them into GPT and start asking my data questions. And then don't need to go to my IT guy to pick it that up.
Clint (15:57)
So what questions would a sales leader be able to answer if their conversations were recorded and turned into this usable data?
Warren Kucker (16:04)
So as like a VP of sales when I'm working with a head of sales, when they come tell me something on a forecast call, right? The key question I come up all the time is did the prospect say that or are you inferring it? Right? So like, hey, we have a decision date. The decision date is December 30th, right? You need a decision date to put something into commit in sales, right? Like I'm going to tell my board that we have a 90 % closing. I need to what date they're going to make a decision and that we're vendor of choice. Two things. Almost all the time, people are telling me things that they think
and not what was in the rep's words. You don't have to rely on that anymore. I could perform my coaching call or my forecast call, my three key deals, be like, show me the quotes from my prospects that show that they've determined what the decision date is or that we're vendor of choice. And then when I'm inspecting this with my head of sales, my sales team, I have that data in the background. I know the quotes there. So everything we're doing in topic around like data analysis is like, is it important that the prospect said versus what the rep said?
and parsing that out and not relying on our own judgment. We can go out and find the actual quote. So that's a good way to like use that type of like data analytics, get the context and be much more prepared for your forecast and coaching calls.
Greg (17:18)
That changes the whole forecasting conversation too. If it's not just vetting of information coming up, but you've already have a lot of the information. Now you can talk through the deal at a much deeper level. I imagine at the beginning, putting all of this in and there's a lot going on for a team because this is different. Where do you see like teams over complicated or, know, try and boil the ocean too early? There's a lot of info coming at them.
Warren Kucker (17:46)
Yeah, I know. What I think, people are overly concerned about inaccuracy, right? Like, what's the... ⁓ And this is a dividing line for me. LLMs aren't going to be good at creating an invoice. One, they're bad with numbers. But two, I can't be wrong on 10 % of invoices. That's a problem, right? So like, people, do your invoicing. Cool. Sounds good. But you know what? When it comes to like, deal inspection, sales coaching, communication analysis, the human element is poor, right? Like, I could just be grumpy and hungry. And like, so my coaching's bad.
Right? like, or I could just be wrong. I misremember all the time. So like the level of error isn't a problem in most areas that you could use this. And that's an important dividing line. What is the consequence of having an error? If the human being is making those errors, then you should get over that hurdle really quick. Right? But two, over-complicating is don't try to systemize it. Just push your conversations into a custom GPT and start asking questions and see what comes out. And then you'll start to find really quickly more useful questions and more useful answers coming.
And I think when you make the transition to software is the difference is you can't overload a GPT with context. Something we have to do at Topic is figure out, great, I have a limited amount of space, context I can give when someone asks a question, how do I figure out what that context is? I have to parse that down and that's the role that that layer of software plays that you can't quite get from that direct. I can't put a thousand calls in there and get a great answer all the time because it only can look at data from 10 calls at a time.
Greg (19:12)
That's really good. You're the first one to kind of open my eyes to that. It's true. In coaching, if you're in a bad mood, you're going to be a bad coach. And if you're really into it and the other person's in a bad mood, it's going to be a poor session anyways. All that effort would go nowhere. But if you can take some of the emotion out or be able to have it on tap for when someone's ready to digest it, it's a much, much more efficient.
for productive ⁓ exercise, I guess is the way to say.
Warren Kucker (19:45)
Totally agree. Totally agree. And the sellers should prep that way, right? They should be in the same GPT and be like, how do I prep? What are the questions my VP of sales guy asked? And eventually, those coaching calls can get either, they'll get much more strategic over time. You'll stop having to repeat some of it. What is that? Hey, give me the three elements that make a champion. I have to have said this a thousand times. This will help us get past some of that base level, get to the more strategic. And then eventually, maybe the reps don't need a coaching call. They're like, we're kind of like.
pushing all the data in. He knows the answers to the questions. I know the app. ⁓
Clint (20:17)
Real-time coaching is just happening as part of the process when you're using the AI system. I love that. I love your whole train of thought there. This is good stuff. Hey, let's shift gears a bit. Let's talk a bit about your experience with AI tools. I'm always curious to hear what kind of solutions you're using inside of your business, maybe even things that you might be doing for fun outside of the business with AI. What are some of the hot names from your perspective?
Warren Kucker (20:43)
I think there's a I think there's a the evolution of how I've used AI There's a useful story here. So I started using clay that go-to-market kind of engine about two years ago at this point and I have like a reasonable technical background. I was using like Databases when I was younger, but I'm not an engineer by any stretch, right? So I get into clay. It's a little complex It didn't quite get what I wanted to I spent an hour on it Which is 59 minutes longer than I usually spend on something I didn't understand and I still didn't get anywhere
Right? So like, all right, sounds good. This isn't useful. I'm going to stop. Then a month later, like a new problem came up. I'm like, all right, maybe Clay can help me solve this. Two hours in there, nothing. This ain't useful. It's not helpful. It's too complex. It doesn't make sense. And I go through this a lot when I talk to people about this software. But then I read a LinkedIn post that talked about somebody else's evolution with these tools. That's like, great. Actually, took them, it actually said literally, first two times I had this problem, the third time I got it. Stick with it. And I got it.
The volume of time savings is uncomparable using a tool like Klay, which just connects to your go-to-market stack.
Clint (21:45)
Just for our audience who hasn't heard of clay before, what does clay do exactly?
Warren Kucker (21:49)
Absolutely. Clay is a go-to-market sales and marketing engine. It connects to a bunch of sales tools. That's what it did initially. It was like the Zapier for sales. But then when you started to layer LLMs in two years ago is when they really supercharged. They were at one million in ARR for six years, got to 100 million by year seven because of this supercharging. What it does is it allows you to enrich all your data really well, but also write sequence emails, do account research, scrape the web, find your contacts, create videos. It's like a big spreadsheet.
because your Salesforce data to cool outbound workflows. So that's what Clay does. It's slightly complex, but not that complex. And then it became a great tool. But in the same token, that's how I actually evolved with like everything I'm doing work wise. I try to look at it. If I've done something repetitively ever great. Can I solve this with AI? This is getting annoying. I've done this three times now. AI could probably help me. You always start like basically back at a napkin, something weird result was poor. But if you push the boundaries a little bit more, you always get to something more robust.
At this point, as a salesperson, I'm vibe coding on the side using quad code in a terminal, which I don't know what terminal is. I never audit the code because it's writing code. But I've made my own CRM because I like to do video prospecting in a way that's different than any other piece of software can do it. And so I was able to do video prospecting in a robust way. The key here is always just try to start with something. Can it solve this problem? Just start asking a question. Start trying to get a result.
and you'll end up in a spot where you're more effective and you're honestly, it's more fun. I'm enjoying what I'm doing now. I'm not tripping over minor issues with tools that I don't have to be anymore. can solve them.
Greg (23:28)
That's great. you read about this all the time. Systematize anything that you're doing in a repetitive basis. Here's a great example. Even when we connected with you to beyond the podcast, we can see every piece of ⁓ interacting with you, booking time, everything is ⁓ systematized, right?
Warren Kucker (23:53)
Yeah, and it's just like, especially in a lot of areas, the easiest way is like sellers hate any admin. Sellers would love to be on a call 50 hours a week and just close deals and don't send emails, don't update TRMs, don't talk to anybody about their deals. They just want to be on meetings. So like let's help automate the things that they hate to do the things that they really like, which is human to human relationships, problem-solving, being on calls and closing work and making money.
Clint (24:17)
All right, let's wrap up here with one last question for you. For that audience member who's a business leader, getting ready to start with AI, doesn't know where to begin, what's your one piece of advice?
Warren Kucker (24:31)
One piece of advice, what I tell business leaders that come to me and say they're way behind in AI is that what's happening now is boards are coming out and saying, you need to use AI and I want to see a 10 % increase in revenue and a 5 % reduction in cost because of AI. This is happening all over the place. Right. And why are they doing that? Because they feel it's an existential threat. Either their competitors are way ahead or a new disruptor will just come and do what they do right now really fast if they don't do it. So as a leader who doesn't really know where to start, what do I do? I buy software and I fire people.
Right? Like that's how I'm going to increase revenue using AI. Right? And decrease costs. Don't do that. The biggest thing that I say is that what we've done is an engineer should be working about three times faster than they were 18 months ago now using God code or copilot, whatever it would be. Don't get rid of them. Hire more engineers. Because the parallel that I look at is if my financial advisor came to me and is like, listen, in 2024, I got you a 7 % return on your portfolio. I started using AI last year and now you got 14%. I wouldn't be like, cool, let's take some cash out.
Right? I'm like, no, you're getting 14%. Let's put money in the pot, right? Engineers are three times more effective than they were two years ago. That's great. We should be moving a lot faster. Right? And so the way I look at it is like, where in my workflow are my people spending so much time when they don't have access to this unstructured data and they're writing all the time? So like, how can I solve those two problems? That's the two best things that LLMs do for me. Where my business flow?
So those two problems really helped move the needle for me internally, but also add more value to my customer base.
Greg (26:02)
great double down when it's working I love it
Okay, it's time for this week's AI Challenge. Now the AI Challenge is a takeaway assignment for our listeners to get their hands on some AI tools and do some exercises.
Clint (26:21)
This week's AI challenge is about eliminating product politics with evidence. We're running a product squeaky wheel audit, using AI to analyze sales calls, support tickets, and customer success conversations to separate loud opinions from real market demand. The goal is to rank feature requests by frequency, revenue impact, and churn risk, not by who complained the loudest or who has the biggest title.
If your roadmap debates feel subjective, this challenge shows how to let customer data, not anecdotes, decide what gets built next.
Greg (26:59)
Yeah, look down into the show notes and you'll find a link that goes right to the blog. It's got all the instructions that you're going to need. Now, if you've just finished an AI deployment within your business, we'd love to hear about your experience, the good, the bad, all of it. So go to www.promptthis.ai, go to the contact us page, fill out the form and we'll be in touch to talk about the story.
Well, this has been a great conversation. You've been thinking about this a lot and you articulate it very well. If ⁓ our listeners want to continue the conversation or learn more about what you do, where can they follow you?
Warren Kucker (27:40)
Yeah, anything I post is on LinkedIn. So that's where I'm talking most about these types of things. So follow me on LinkedIn, DM me, go, let's start the conversation. I'm always interested in what other workflows are out there that this type of data unlock could really help.
Clint (27:58)
Fantastic. Well, that was a great conversation, Warren. Appreciate your time today. Thank you very much for joining us.
Warren Kucker (28:03)
Thank you guys.
Clint (28:05)
And that's another episode of Prompt This.
AI Announcer (28:12)
Ladies and gentlemen, thank you for joining Clint and Greg today. You can catch all the electrifying prompt this episode and dive into some in-depth articles at www.promptthis.ai. And hey, don't forget to smash that follow button below. We can't wait to have you back for more action-packed content.
