[00:00:11] Andy Karuza: It's taking a while to go live. Here we go. Welcome everybody to another Nacho Tuesday, and good afternoon over there. I have the co-founder of Spiky here, and without further ado, we'd love if you can introduce your product and company and what you guys do.

[00:00:23] Burak Aksar: Absolutely, thanks for having me indeed. Hey guys, this is Burak, co-founder of Spiky. My background is actually in engineering—undergrad in electronics and computer science, and then I got my PhD at Boston University in computer engineering. Then I conducted research at MIT, IBM Labs, and Sandia National Laboratories on scalable and real-time AI systems.

So far, you're probably thinking, "What is this guy doing in the revenue space? This crazy engineer." I've spent a decade building AI that has to actually work in production, not just in a notebook. Spiky is basically a Cursor for sales conversations. Cursor today works because your code gives you instant feedback—it either runs or it doesn't. But sales has the opposite problem: a rep says something on a call, and then you find out maybe three months later whether it mattered at all.

We close that loop, actually. We analyze conversations to statistically prove which behaviors drive revenue and then deploy them as real-time feedback or real-time coaching during live calls—like when you're chatting with the client, not after the fact, not before. Then basically, we capture every decision that a rep makes, and this behavioral labeled data helps us to train next-generation AI agents to accomplish complex revenue operations. That's what we do in a nutshell.

[00:01:51] Andy Karuza: Very exciting. It's not your typical note-taker; it does a lot more on the analysis side and actually understanding what's being said and quantifying it.

[00:01:58] Burak Aksar: Yes, exactly. Because note-takers are good—it's like a good data source for us—but it's after the fact. Everyone is busy, no one wants to look at that. How can we make this more interactive and dynamic? That's what we are focusing on.

[00:02:09] Andy Karuza: I guess how is AI fundamentally changing the way GTM teams make decisions today compared to a few years ago?

[00:02:15] Burak Aksar: The GTM space has a lot of potential because that's where the money is, right? Two or three years ago, most GTM decisions were made from dashboards and gut feel, as a lot of the leaders we talk to today admit. But today, interestingly, most teams still make decisions from dashboards and gut feel—they just have an AI summary on top of the dashboard. That's what I would say.

But honestly, the real shift you're seeing—and I think that's the one that matters—is moving from descriptive to causal. In the old world, what we were seeing is, "Oh, our win rate dropped 10% last quarter." In the new world we are imagining with Spiky, it's, "Our win rate dropped 10% because reps stopped doing discovery on multi-threading after week three of the new playbook, and here is the specific behavior that correlates with the deals that did close." We are trying to learn the science behind your revenue, and you have enough data to figure it out. That's the actual change, and that's where we should be going.

Statistically valid answers in go-to-market data are still not there, but that's where we see the trends are going. A lot of the leaders we talk to, they ask us why they are losing. "If you can figure out why, how am I going to fix this, and how fast I can fix this?" Those types of directions we've been seeing.

[00:03:47] Andy Karuza: A lot of times, one sales rep might say one thing if you don't have this tool, and another sales rep might think it's a different problem than another one does. You're really just getting multiple points of feedback, and they all might be competing and conflicting with each other, if you will. You're not able to get the real picture.

Burak Aksar: Yes, because talking to a lot of sales people, we are all optimistic. I'm a salesperson myself right now. We are optimistic, so we just want to close that deal somehow. We need that data-backed approach in our lives.

Andy Karuza: A lot of times they'll say, "Oh, the product's... I need this from the product," or what have you. But with your tool, you could probably quantify that actually—like if customers are objecting on certain features not existing, rather than one sales rep just saying he couldn't close a deal because one feature is missing. You could probably find some value in that, too.

Burak Aksar: Absolutely. The aggregate insights and strategic insights, and then can we convert them into action.

Andy Karuza: I guess what are the biggest inefficiencies in traditional sales and marketing decision-making today that AI is now helping solve?

Burak Aksar: I think there are a few, but if we need to talk about the big ones, they all stem from the same root cause: there's no actual measurement between what we did and whether it's connected back to ROI in my company, because it's hard to measure. For example, if you look into a CRM, it will tell you that okay, this deal went from stage two to stage three, or it died—it's closed-lost. They tell you nothing about what happened in between or why this decision has been made. You don't know, basically.

There's this 90-day data cycle between two data points, and you're like, "Okay, what's going on?" As a CRO, you can't dig into each case. What you just see is which stage you lost in the funnel, but you don't know why.

The second one is definitely coaching. Right now, we see a lot of trends about, "Oh, let's automate human beings today," like AI SDRs. But honestly, what I've been seeing is we still haven't figured out how to coach a human being properly. If you can't make a human being 10x, and if you don't know the process there—the data process—how are you going to make an AI that will accomplish this in a human way? A lot of the time, managers have to listen to certain stuff and coach if they have time, but it's a very broken process.

The third inefficiency-wise is nobody knows what actually works. It might sound very weird, but let's say every sales organization today has a playbook. I ask them, "Guys, how did you come up with the playbook?" They're like, "Because I used MEDDIC in my old company." But how do you know it's a good playbook for your new company? Because they have MEDDIC, BANT, custom questions, etc. But maybe you don't even need MEDDIC. Maybe there are two questions that you need to ask, and that's where the success probability is the highest. That's the statistical confidence, and that's why I keep saying people don't know what they don't know. That's where the revenue leaders are today, and those are the biggest inefficiencies.

Andy Karuza: Sometimes you have to unlearn what you've learned because every new product or company that you're working with has a different need in terms of how you go about solving the problem or selling the product. There are some similarities you pull over for sure, but it's not the same every single time, whereas a lot of sales people treat it like it's going to be the exact same gig.

Burak Aksar: I think it's getting more competitive and more complex because now your buyers are more knowledgeable. You can't just say, "Oh, let me get back to you, let me search this competitor, whatever." They want you to know everything. The expectations are higher, so it has to be dynamic. Traditional sales enablement is already broken—where a person sits through or sifts through Gong recordings trying to find competitive stuff or what they did. It will never work anymore.

Andy Karuza: You need some help to go through all that and find where the deals are. I guess many GTM teams are really struggling with signal overload, right? Like you mentioned, going through Gong videos—there's a lot of data out there. How do GTM teams avoid that overload of data and focus in on what matters?

Burak Aksar: Great point. The problem is more data and more signals are not the same thing. Most GTM stacks right now are generating—I keep saying this—noise at an industrial scale, almost. The actual problem is 99% of these signals are not tied to an outcome. In all those enrichment tools, you are seeing a rep got this high-intent alert, like "Spiky is hiring SDRs, so I'm going to sell them a tool like Apollo or enrichment because they will need it."

The thing is, you don't know what they actually need and what their use case is. You just make an assumption based on the market. That's not high-intent selling or high-signal selling, so that's why it's a little bit noisy. You also don't know, when you have the signals in a similar company to Spiky, did you close the deal? Did they take an executive meeting? You don't have that information. You just make decisions out of your limited knowledge.

They are drowning in signals, but they don't know... the fundamental question they are not asking today is: which of these signals historically actually predicted revenue or tied back to revenue? If you can find those, those are the real signals. That's what you need to focus on.

Andy Karuza: That's a great point. How does AI-powered decision intelligence actually look like in a real-world GTM organization?

Burak Aksar: I would probably say it's less glamorous than the term sounds. Everything is AI-powered right now. In practice, there are a few things stitched together. The initial one is the AI piece watching every customer conversation, hopefully extracting the behaviors that actually happened, not just topics, by the way. I'm talking about behaviors.

For example, Andy, you came up with this pricing objection, and did Burak handle it well, or maybe I just anchored it, maybe I just deflected? Those types of behaviors are the key. In defining AI-powered decision intelligence, the second stage is when you figure out the behaviors at scale, can they get tied to outcomes? It could be closed deals, churn, expansion, closed-won. You can statistically find signals which help you with what.

The third one is the activation piece, which is a layer two levels above traditional enablement or traditional training, even roleplaying. The third step is these proven behaviors have to be pushed back to reps in the moment when they need them. It could be a new product knowledge, it could be how to handle a pricing objection when you are talking to a VP of Sales versus when you are talking to a CFO. All these matter. There's a lot of personalization opportunity here.

But that's the loop of AI-powered decision intelligence: Step 1 is figure out the patterns. Step 2 is tie these patterns into revenue. Then basically distribute, take action, and make profit.

Andy Karuza: Can your tool actually also read maybe behavioral cues, like visually?

Burak Aksar: We are looking into non-verbal cues as well. It's still early in the space, but we believe there's a huge portion or huge potential there. Short answer is yes.

Andy Karuza: Very interesting. Because you could see visually if there's an objection—maybe they say, "Oh, that's interesting," but their body language might show that they're not interested in it, even though they say that they are.

Burak Aksar: Absolutely.

Andy Karuza: I guess what are some of the most common mistakes teams make when implementing AI into sales and marketing operations?

Burak Aksar: They buy tools before defining the decision. They don't have a solid use case, or the solid use case they have is not tied to revenue. Most of the time they're like, "Oh, okay, what else can we use this for?" They try to do this backwards.

They should start with the decision or ROI that they want to improve and then find the tool. People love this, or people buy the brand, but honestly, I think the best product wins in the midterm and long term. Short term, this marketing stuff is really important, but people should be more focused on the actual ROI or the actual problem.

Second, they're too obsessed with AI output, even in a positive or negative way, because they are like, "Oh, AI gives me this answer." It's just a probability; it's not a verdict. You can't fire a rep based on AI scorecards without validating them against outcomes. That's why you should look into AI output in a way that you can use it positively, but people sometimes get too obsessed and say, "Oh, this sucks," or "This is really good." But they don't know actually whether it's going to help them.

The third one that we've been seeing is most AI stuff stays in the insight piece—they don't have a delivery mechanism. That's the biggest problem. When you bring a consultancy company to an enterprise, they will look at everything, they will make suggestions: "Andy, do this, this, this," in a 30-page PDF. Then you're like, "How am I going to execute on this?" You can have the best strategy, you know everything about the business, but if you don't have a delivery mechanism that people actually can take action on, it's going to be useless. The activation piece is really, really important, and people keep missing that.

Andy Karuza: I see a lot of people getting excited about AI, too, and they're just generating plans and content and just frankly having a lot of fun with it. But I don't think a lot of them really think through the implementation of it—like, "How am I actually going to put this piece of work to use?"

Burak Aksar: Absolutely. Strategy without a deployment is useless.

Andy Karuza: A lot of people don't think through all the way to the end, too. They'll just go about generating work for AI, but they don't actually know what the intended outcome is. How should GTM leaders think about balancing human intuition with AI-driven recommendations? It sounds like you're a fan of not removing human intuition and keeping the human in the loop as much as possible, too, right?

Burak Aksar: I would probably reframe the question slightly because I think the interesting balance isn't intuition versus AI—it is intuition informed by what. Let me explain what I mean by that. Let's say a senior Account Executive—the intuition that person has is incredibly valuable and it's built from maybe a thousand deals over their career, all filtered through their memory, which is also lossy and biased. AI's intuition is built from millions of conversations with outcome labels, hopefully, but it has no context for the specific human across the table.

That's why I think the right model is AI surfaces patterns the human couldn't possibly hold in their head, because our memory is very limited compared to AI, and then human applies judgment in the moment because they can read things AI still can't read. Maybe a hesitation, a relationship history—all that stuff is going to make a difference. If you can combine this, this is the holy grail there.

The mistake you're seeing is they try to make this binary. The best reps I watched, they use AI like a kind of co-pilot or they call it Jarvis, from Iron Man. In the mid-call, it surfaces a few data points, helping them: "Hey Burak, maybe you should ask this, too," but I can decide to do it or not. I still have my Jarvis, my companion, my assistant if I am stuck somewhere or if I want to pursue further. Neither of these alone is as good as both together, so it has to be intertwined.

Andy Karuza: That's great. Where are AI tools having the biggest impact today, especially when it comes to sales? Is it mostly in pipeline generation, forecasting, coaching, customer insights, or something else?

Burak Aksar: Pipeline generation got the headlines because AI SDRs raised billions at a billion-dollar valuation, and everyone is like, "Top of the funnel, top of the funnel!"

Andy Karuza: More business, more business! Figuring out how to close more business, actually.

Burak Aksar: Absolutely, so that's the interesting part for me, and I think leaders are also understanding this. It's fun to try this because it's easier, by the way. That's why people want to focus on that portion first in terms of, "Oh, I can get more deals, and technically I can put more humans in the mid-funnel, so then magically we should be closing more." But now, most AI SDR solutions fail or the personalized top-of-the-funnel generators work for specific cases, not let's say fail completely.

Leaders are realizing, "Hey, we've been doing this sales stuff for a very quantity-focused approach, but what about the quality? What about the mid-funnel, bottom funnel? How can we optimize?" Because honestly, most teams have enough pipeline today. The close rates are funny—we're talking about sub-10%, maybe 20%. That's really interesting.

You also mentioned this forecasting piece. Forecasting AI is improving, but the problem is to be able to forecast something, you need accurate data points. You need quality data in your pipeline. A traditional sales pipeline is built with, let's say, Andy thinks stage two is 30% probability, but I can come and say, "Andy, let's make this 45%." Now it's going to change the whole forecasting algorithm. The problem is we don't know whether 30 or 45 is the right one.

To be able to make this forecasting more accurate, people really need statistical models in this space, and to be able to work with those, you need labeled data. Sales is really interesting, so that's where we are seeing interesting trends and the biggest impact piece, but it's still really early. Also, on the coaching side, everyone wants to coach their people, but it's still broken. That's the fascinating part; I don't think anyone cracked the game yet.

Andy Karuza: Are you really giving people good quality insights and actual good coaching, or are you just talking, which a lot of people like to do in sales calls?

Burak Aksar: Absolutely. That's interesting, but it is what it is.

Andy Karuza: Same with social media metrics, too. It's like people obsessing over followers and not engagement, right? Same in sales—it's like we have a lot of leads, but are we closing all those leads?

Burak Aksar: Yes, that's why mid-funnel and bottom funnel, people are trying to optimize a lot.

Andy Karuza: How does conversation intelligence and behavioral data change the quality of decision-making across GTM teams? It sounds like it gives you superhuman capabilities because you could, like you said, see all these signals that you normally would miss just watching a sales call.

Burak Aksar: Even the definition of conversation intelligence—it's been around for the last 10 years, but the problem we've been seeing is I don't think the existing conversation intelligence from three years ago was intelligent at all. It was just a marketing term.

Today, every GTM debate is like an opinion fight because I'm saying, "Oh Andy, I think we shouldn't mention pricing earlier in the deal." And you'll say, "Oh, I mentioned pricing, it works." And then I'm like, "Okay, under which conditions, under what type of ICP?" It's just an opinion fight. But if you can instrument conversations and then tie behaviors to outcomes, those debates should end, or at least they should be minimal.

Now you can say, "Hey Burak, I looked into the last 3,000 deals with Spiky, and when we mentioned pricing before the second call or with a decision-making call, we lose 20% of these deals. When there's no decision-maker, we shouldn't do this." That's a fact about our business coming from our own data. When you change jobs, it might be different for a different company.

The downstream effect here is really huge. Technically, onboarding can go from months to much shorter because you are teaching new reps proven behaviors, not just opinions. Marketing knows with this behavioral data which messages actually land in real conversations—that's really valuable, and people don't utilize it too much. Product knows which objections block deals. The whole organization starts to make decisions from the same source of truth, and that's the real value of decision-making.

Andy Karuza: That's massive. You just hit the nail on the head right there because everybody's seeing the problem from their perspective. Product looks at it a certain way—they don't want to build certain features maybe because it's probably too much engineering effort. Sales looks at it a certain way, too—maybe they think this feature that to them seems like the lowest-hanging fruit to close deals might be the most work for engineers. Everybody will look at the problem from a different angle and perspective, but if you have a single pane of truth, you could really make better decisions because of that.

Burak Aksar: Absolutely. That's really important.

Andy Karuza: That works across the whole organization, too, which is key. What separates organizations that successfully operationalize AI from those that struggle to get value from it? We kind of talked about this earlier, but a lot of people are just really excited to throw AI out there and they're not really implementing it the right way.

Burak Aksar: The clear pattern we've been seeing is a clear owner of the feedback loop. The companies that fail treat AI like a tool purchase—they buy it, deploy it, run a pilot (everyone has a pilot budget), and get a report. There's nobody whose job it is to make the AI's output useful in the company. You need that type of AI champion, not just a regular champion.

The companies that succeed treat AI like a hire. Somebody owns it, somebody's responsible for asking: "Is it impacting anything? If it is not impacting, how can we make it impactful?"

The second thing—and this is a little bit more cultural—successful organizations let AI contradict their gut feelings. If the VP says, "Okay, we have to do this playbook," but then the data says, "Guys, this playbook is useless, you don't need those three questions in the playbook," and if the VP is open to these types of ideas, it's super powerful. They know now they can do better, and they're open to it. Culture-wise, you need to be more open-minded to be proven wrong.

Andy Karuza: In general in life, not just from AI, but from anybody you interact with in life, you need to be able to absorb feedback, take that objectively, and look at ways to improve the way you do things.

Burak Aksar: Absolutely. Especially in the AI era, that's more important because things move super fast.

Andy Karuza: We can tell, we see all these products every day, and the speed of business these days is just unimaginably fast. What kind of KPIs or business outcomes improve first when AI becomes embedded into the GTM decision-making process?

Burak Aksar: For us, it's mainly ramp time and productivity increase. I know it's a little bit of a boring answer from a ramp-time perspective, but a new rep takes maybe 12 to 18 months, depending on your domain, to ramp up. In a traditional sales organization, they hire more people than they need so they can fire fast. The KPI is that ramp-time perspective.

The second one is conversion, but it takes more time. Can we actually increase the win rate in certain cases compared to last quarter for a specific ICP? We try to go very pinpointed there in terms of the win rates.

As we were talking about this top-of-the-funnel lead volume, people are very obsessed with it because they think they can figure out the middle funnel later: "If I have enough quality leads, I will be closing." Some people talk about forecast accuracy in terms of KPIs, but honestly, I don't think it's easy to measure at this stage because the underlying data is more like a garbage-in, garbage-out type of stuff.

Andy Karuza: I see that happen a lot with people—they just assume that if they have traffic or volume, the sales can figure themselves out.

Burak Aksar: Absolutely.

Andy Karuza: Looking forward, how do you see AI reshaping the structure, roles, and workflows for GTM teams over the next few years?

Burak Aksar: Right now we have this pyramid in a traditional organization, and I think it's going to flatten. In this traditional management structure, let's say one manager per 10 reps, one director per three to five managers, etc. Most of that middle layer exists to do things AI is doing better now—it could be pipeline reviews, deal inspection, coaching, forecast rollups.

Within three years, the median go-to-market organization might have half as many managers and twice as many senior ICs or individual contributors—people who are part seller, part operator, working alongside AI agents. SDR, Solution Engineer, AE—that can be one person from my dream perspective.

The bigger shift most people right now aren't ready for is AI agents doing real selling—not just SDR outbound, but actual qualification, discovery, and actual follow-up. The job of an AE in the next two years is going to look more like a portfolio manager. They can orchestrate 10 agents jumping on the deals that matter and be the human in the room for the moments that require trust. Maybe I'm going to have coffee with you in New York so we can close a deal—something like this.

Andy Karuza: What's next for you and Spiky?

Burak Aksar: Probably two things in the near term: we are doubling down on financial services, banks, insurers, fintech. These are the hardest customers in the enterprise. They have regulations and the longest sales cycles, but there's no talent coming in. It's an unsexy field; no one wants to go to these fields after they graduate college, so our focus is there.

Longer term, we are actually building the world's first behavioral training layer for AI agents. Every company in the next two years is going to deploy AI agents into customer-facing roles. When I say AI agent, I know it's an overused term, but we are imagining agents that can accomplish complex revenue operations. Generative foundation models can't do that. I want someone selling, qualifying, and who knows how to manage a pipeline. That's where we are really excited.

Andy Karuza: Very exciting. Where can people learn more about Spiky and pick up your product today?

Burak Aksar: Definitely our website. Feel free to reach out to me if they want to chitchat about anything about go-to-market. I love this field because of the fundamental engineering problem or data problem they have. For any go-to-market leader listening: don't buy AI for the sake of AI. Start with the decision to improve and work backwards. The team gets real value with a clear problem, and then you can go from there. If you are in these industries, if you're trying to AI your revenue organization without blowing up compliance, come talk to us. Happy to chitchat. That's probably all on my end.

Andy Karuza: There's that saying, "If you write the problem down and understand it, you've halfway solved it already." Nice. Looking forward to seeing you hopefully out at New York Tech Week as well, Burak. Thanks for coming on today. That's spiky.ai—you can see the domain there in his name. You can also pick it up today in the NachoNacho B2B SaaS and AI marketplace. We have the best brands obviously because we have products like this, so check it out today. Burak, thanks again for coming on, and looking forward to connecting live in New York sometime.

Burak Aksar: Thank you so much, Andy. Thanks for having us, and looking forward to meeting in person as well.

Andy Karuza: All right folks, enjoy the Tuesday. Take care, bye.

Burak Aksar: Take care, cheers.