Episode 456AIRevOpsContent Strategy

Why the AI silver bullet isn't fixing your revenue problem with Matt Zelasko

Matt Zelasko, founder of growth agency Radish and self-described 'Tom DeLonge of RevOps,' joins Content to Close to diagnose the AI silver bullet problem: most revenue teams are using AI to do the same broken things faster, then blaming AI for the engagement drop that was already happening. He argues the gated-ebook playbook has been a case of diminishing returns since around 2018, that content is at peak saturation, and that LLMs are not stealing search traffic so much as exposing how tired the old motion already was. The unlock, he says, is realizing an LLM is just speculating what comes next — once you understand that, you shift from prompt engineering to context engineering ('what else do you need from me?') and stop trusting confident, wrong output. Matt shares why he turned down a client who wanted an AI agent to write case studies (no captured experience to feed it) but said yes to one who wanted an agent to write RFPs (the inputs are given, the format is standardized), and why creative people still need to write 50 to 100 bad taglines by hand before reaching for any tool — a Rick Rubin 'I know it when I see it' instinct that gets dulled when AI does the ideation for you. He also warns against the talking-points pitfall where buyers read six blogs, half-diagnose their own problem, and start treating symptoms instead of root cause.

Matt Zelasko

Matt Zelasko

Founder, Radish

18 min

Key Takeaways

  • 1Stop using AI to perpetuate the old playbook — the gated ebook that took months to develop and got read by nobody worked in 2013, 2014, maybe through 2018, but it has been a case of diminishing returns ever since; bolting AI on top of a saturated motion does not bring it back, it just scales the slop.
  • 2Reframe AI from prompt engineering to context engineering — the unlock is realizing an LLM is basically speculating and guessing what comes next, so the high-leverage move is to start every prompt with 'I want to do this — what else do you need from me?' instead of trusting a confident, wrong answer.
  • 3Diagnose the problem before you reach for the technology — buyers read six or seven blogs, get halfway to a solution on their own, then come to an agency treating symptoms instead of root cause, so the first job is to slow down, name the actual pain, and remember that 'blockchain will change all of marketing' and NFTs were the last two hype cycles you survived.
  • 4Use AI to buy back time, not to fake expertise — Matt uses it to draft things that do not need his time (a data security policy he then hands to a real proofreader) and as a creative ideation partner when he is stuck, but warns sellers without subject-matter expertise that they will not know when a confident LLM is wrong for their exact context.
  • 5Pick AI agent projects by whether the inputs and format are bounded — Matt told a client no on an AI agent to write case studies (you would need it on every client call for a year before it could capture the experience, no ROI in the near term) and yes on an agent to write RFPs (the inputs are handed to you, ten RFPs of training and it writes them the way you want).

About this episode

The old gated-ebook playbook stopped working around 2018, and slapping AI on top of it isn't going to bring it back. In this Content to Close episode, Matt Zelasko, founder of growth agency Radish and self-described 'Tom DeLonge of RevOps,' makes the case that most teams are using AI to do the same broken things faster, then blaming AI for the falling engagement that was already happening. Matt walks through why content is saturated, why 'intelligence' is the wrong word for what an LLM actually does, and why understanding how the technology works — it is speculating what comes next — is what finally unlocks its real use. He shares his shift from prompt engineering to context engineering, why he turned down a client who wanted an AI agent to write case studies but said yes to one who wanted an agent to write RFPs, and how creative people can use AI without losing the ideation work that makes the output good. If you are tired of AI hype and want a sharper view on where it actually belongs in revenue generation, this one is for you.

Topics covered

  • The AI silver bullet and the diminishing-returns content playbook
  • Prompt engineering vs. context engineering
  • How LLMs actually work (speculating what comes next)
  • AI in creative work — taglines, Rick Rubin, and writer's block
  • Choosing AI agent projects: case studies vs. RFPs

Notable quotes

I'm Matt. I am on LinkedIn, better known as the Tom DeLonge of RevOps.

Matt Zelasko(01:21)

That playbook worked in what? 2014, 2013, up until, you know, it's been a while, 2018 even, right? But I think it's a case of diminishing returns.

Matt Zelasko(03:35)

Instead of prompt engineering or something, whatever people are calling it, it's more like context engineering. Like, what else do you need?

Matt Zelasko(06:06)

The minute I realized that it was basically speculating and guessing what comes next, I went, all right, I get that.

Matt Zelasko(11:56)

Resources mentioned

  • Framework

    Context Engineering — 'What else do you need from me?'

    Matt's reframing of how to work with an LLM. Instead of trying to write the perfect prompt, start every interaction by telling the model what you want to do and then asking 'what else do you need from me?' This forces the model to surface the missing context — facts, constraints, voice, examples — before it generates a confident wrong answer. The mental model underneath it is that an LLM is speculating and guessing what comes next, so the quality of the output is the quality of the context you fed it. Use it especially when you are not the subject-matter expert and would not catch the model being wrong on your behalf.

  • Playbook

    AI Agent Project Selection — Case Studies vs. RFPs

    Matt's heuristic for deciding whether an AI agent project is worth building. Two real client requests illustrate it. A client wanted an AI agent to write case studies — Matt said no, because a high-quality case study requires captured experience the agent does not have; you would have to put it on every client call for a year before it could write one well, and the ROI is a year out. Another client wanted an AI agent to write RFPs — Matt said yes, because RFPs follow a standardized format, the inputs are handed to you by the prospect, and ten RFPs of training is enough for the agent to write them the way you want. The rule: build agents where the inputs are bounded and given, the format is standardized, and the ROI does not depend on captured tacit knowledge.

  • Framework

    The 50-to-100 Taglines Habit

    Matt's pre-AI copywriting discipline, still worth keeping. When he needs a tagline he opens a notebook and writes 50 to 100 of them longhand — anything that pops into his head, starting with deliberately bad ideas to break the impulse to write the smart one first. The point is the ideation muscle, not the final tagline. If you skip straight to 'AI, give me 50 taglines,' you lose the process that makes you a better creative — the Rick Rubin 'I know it when I see it' instinct gets dulled because you never built the reps. Use AI as a creative unstuck-tool when you have writer's block, not as the front door to ideation.

Full Episode Transcript

Benjamin Ard00:55Welcome back to another episode of Content to Close. Today I'm joined by Matt. Matt, welcome to the show.

Matt Zelasko01:00Thanks for having me, man.

Benjamin Ard01:01Yeah, now I'm excited. I think this is going to be a fun conversation. I think it's a conversation that people need to hear and I don't know how easy it'll be for people to hear, but I think it's something that people need to hear. But before we dive in, Matt, let's get to know your background, work history, all that fun stuff. Let's the audience get to know you.

Matt Zelasko01:21Yeah, man. Really appreciate you having me on here. I think it will be a little less painful than you're expecting based on our kind of back and forth. I think I may have been in kind of a mood about something while you were emailing me. So we'll get to that in a moment. But, you know, I'm Matt. I am on LinkedIn, better known as the Tom DeLong of RevOps. I've been in kind of the agency, marketing agency, rev-op space for longer than I'd like to admit. I have a lot of opinions on a lot of different things. I'm very willing to be proven wrong because I think that's how we all grow and we all learn. And I also think that, you know, far too many people take themselves far too seriously. It's okay to take the work seriously. Nothing that says that you can't have a little bit of fun while you work, while you do things. So, that's me in a nutshell.

Benjamin Ard02:13I love it. I love it. And where are you at now? What do you guys do? All that kind of fun stuff.

Matt Zelasko02:18So I run an agency called Radish. We are kind of a growth partner. We do a lot of different things for a lot of different people. But I don't believe in kind of like the vertical niche. I do believe in the horizontal of that for an agency. So ultimately, we like to say that we help folks do more rad shit, you know, we're kind of hell bent on helping customers grow, you know, take their next step in their growth. So that's that's what we're looking to do.

Benjamin Ard02:50And I have seen a lot of websites in my day. If anyone listening wants to see a website done really, really well, that is bold and fresh and just refreshingly cool. Check out the website. We'll link to it in the show notes. So I'm excited to dive in, Matt. What we're going to talk about is the disconnect between the AI silver bullet promise and the reality of what AI actually delivers. And how revenue is actually generated. So the AI silver bullet kind of starting there. What, what do you think about this traditional playbook of revenue generation and is AI actually disrupting that? Is it not like what, did the AI silver bullet term come from as we were kind of emailing back and forth?

Matt Zelasko03:35When we talk about kind of like the traditional revenue playbook or content playbook, I've been thinking about it a lot over the last two and half, three years. Like there's nothing worse than meeting with a client and finding that they still want to do like the gated ebook or the you know, whatever, and take months to develop it and then to have nobody read it. Like that playbook worked in what? 2014, 2013, up until, you know, it's been a while, 2018 even, right? But I think it's a case of diminishing returns. In fact, I don't think that. I know that. I watch it, right? Like we've literally stopped

Benjamin Ard04:01Spend a minute.

Matt Zelasko04:18Like, promoting the fact that we do content, we have content writers, we have copywriters, we have great people, but you know, two things are kind of happening there. People are asking for it less because they're using AI to generate whatever, whether it's slop or whether, you know, whether it's good. I've seen both, right? But I think that they're realizing that they don't need to spend months and months and months, but I think the problem that I have with of like the AI silver bullet and where this conversation started was they're just taking what they used to do and using a technology to continue to do the exact same thing. And also kind of like falsely attributing like their lack of engagement to the fact that AI is taking away from it. Like they're not searching for us on Google anymore. Like people are, right? Like LLMs aren't taking away that much right like the what was happening was it's saturated like we are at the peak amount of content that people need right now for the next few years like that playbook just isn't effective anymore and people are both blaming it on AI and also utilizing AI as the solution for it. And I think that that's wrong and kind of silly.

Benjamin Ard05:27I love it. So when it comes to the actual playbook to generate revenue, what is the ideal placement for AI in that process? How should companies be utilizing it? Is it in the ideation stage? Is it in the execution stage? Does it even have a role? I'm curious, like, as is your experience seeing, you know, a lot of different businesses come through trying to generate revenue, what playbook is working and how can AI appropriately be used in that playbook?

Matt Zelasko05:55I don't know that anybody should listen to me on this, you know? Good guess to have, right?

Benjamin Ard06:01The best answers come from people who usually say that right before they give an answer.

Matt Zelasko06:06Now I think I may disappoint you. No. I don't think that AI is without its use anywhere, right? I use it, right? And for me, it's more about maximizing my time that I can spend on things that are valuable to revenue generation. So I can only really speak for myself and what I see with our clients and that. I utilize it for things that don't need my time in my writing policies, right? Like for like data security policy, you know, and then taking it to somebody who can proofread it, right? And say, okay, yeah, that's good. Like I utilize it to maximize efficiency, essentially. I think that there is a lot of potential for it. I've utilized it to help me create like a content strategy, right? Where you're talking to it and, and, and, and helping yourself ask the questions you don't know to ask. Like that's, that's genuinely where I've gotten the most, use out of it. The most, impact that it's made on me is being able to say, in my prompt, like, hey, I want to do this. What else do you need from me? Right. And I think that that's the difference. I think it's, you know, instead of prompt engineering or something, whatever people are calling it, it's more like context engineering. Like, like what else do you need? Because otherwise it's, going to spit out a bunch of stuff and be very confident and say, here you go. And I mean, if, like, if you're in if you're in sales and you're not a subject matter expert, right. And you're trying to generate revenue and you're you've been given a tool that can, you know, query the all of human knowledge. Are you gonna know that that's right or wrong for your exact context? I don't know, right? I know I don't, you know. But I think that there's like, it's less about where it fits into revenue generation and how it helps you spend more time on revenue generation.

Benjamin Ard08:06I love that. I love that. And I love that you're talking about taking away the responsibilities of things that don't require expertise. And what I loved is I was doing my homework before this episode. You, again, I mentioned the website at the very beginning. You're an agency and correct me if I'm wrong, but it seems like you pride yourself on being really creative, thinking outside of the box, providing really cool opportunities for your clients. And doing things just like you said, rad stuff, like this incredible stuff that allows people to really market in a very fun and authentic way. Feels like with artificial intelligence, there's more and more of that that's needed because there is more noise. There is more content to be read. We have hit this apex of content consumption. How do you feel like creativity and AI work together? Now, obviously a lot of times people don't think those two things jive very well, but I would be remiss if I didn't ask you along this vein, because again, I get this strong sense from the agency. You guys do really creative work. I'd love to kind of hear how AI works in that or doesn't work in that creative process.

Matt Zelasko09:16So I think I'm gonna disappoint you a little bit and say that it's kind of more of the same from what I was just, you know, like it's helping you be more efficient, you know, like even in a creative space or in a creative role, which I mean, I am not in a creative role necessarily, right? Maybe strategically creative, but not like, I'm not a designer, I can't do that. I'm not an art director or creative director. Like I can't do that. I'm more like Rick Rubin. I know it when I you know, like I like what I know and I know what I like and I can identify it. But I think it certainly enables creative people to spend more time on creative pursuits. It also, I mean, you know, years ago I was a copywriter, right? And I need a tagline for XYZ, right? My first process or my, like the first thing I would do, which is open my notebook and just start writing taglines. And I would write 50 to 100 of them, anything that popped in my head, anything at all. That's, I think, something that could be lost if you only went to AI for that, right? Like, hey, I need 50 taglines. Like you miss out on that process, that ideation, I start by writing the worst ideas I can think of because that kind of you out of like that desire to write the smart thing first. Right? Like the, like the fun, you know, like the right one first. However, if you're stuck AI can be your best, best friend. You know, I mean, I, I, I do it all the time. Also creative people can't be on all the time. Genuinely cannot be on all the time. There's a, mean, it's writer's block. It's what, you know, like what creative blocks, I think it could help.

Benjamin Ard10:39Mm-hmm.

Matt Zelasko10:58There. I've used it to help there. I'm also getting a little older, so like I already have my own process for getting out of stuff like that, but I think it can help in that sense.

Benjamin Ard11:09So we all started this conversation really about this concept of the AI silver bullet and how just how, and I loved your answer to that, that this technology is really just perpetuating a lot of what you're already doing and expecting different results may not be the best approach here. When you hear someone come to you and say, hey, I think I found this technology. Got AI in it. I think it's going to solve all my problems. It can generate all this revenue, all that kind of stuff. Like what would be your advice? Like, I want you to pretend a listener has just come to you and said, hey, Matt, we have this cool new technology. We think it's going to solve all our problems. What do you think? Like, what would be your advice to that individual? I mean, obviously you don't know what the situation is. I'm really putting you in a weird scenario, but like, what, do you feel like off the cuff? Would be that kind of advice to have people look out for.

Matt Zelasko11:56My advice wouldn't necessarily be like wisdom or a one-liner that helps them find the right path with that technology. Would be, let's talk about the problem. Let's make sure you fully understand that problem. Because I think a lot of the time people, and this is a side effect of inbound marketing and all of the content that has been created. They read six, seven blogs written on what they think is their problem. And they get halfway to a solution and then talk to somebody like me or somebody much smarter than me even. Hopefully, and they're like halfway there and they've like kind of lost sight of what the actual pain point is, right? Like they're, they're, they know the symptoms and they're trying to treat that rather than the actual root cause. I would also then say, don't fall into the hype cycle because we've seen this before. Blockchain was going to change all of marketing, right? NFTs were like the, know, like, that was admittedly sillier and much less useful than, than AI. I also encourage people to really fully understand what the product is. In terms of AI, right? Because that's kind of a blanket statement. And 99 % of people that I have encountered who, and I was one of them for a long time, just, I don't jump onto hype cycles because I'm just kind of a jerk like that, right? Like I want people to be wrong for whatever reason, it's just how I am. I like get to know what an LLM is and how it works. The moment you realize how it works, that's when you can unlock its potential. If you just follow the marketing and the sales guy who's, I mean, he's got a wife and kids to feed or, you know, like he's got to put food on the table. He's often going to say things that he doesn't fully understand and kind of say, yeah, I can do that. Right. The moment I understood how an LLM works was the moment I realized that I could use it. Up until that point, was like, nope, this isn't good. This isn't working. It's not that useful because I wasn't prompting it right. Because I wasn't asking it what it needed from me. Because I wasn't giving it enough context. The minute I realized that it was basically speculating and guessing what comes next. I went, all right, I get that. I also think that that's going to help you avoid so many future pain points with it, right? Like when you understand how it works, you understand how to note it, like how to pinpoint when it's wrong, essentially. I also don't think that the word intelligence should be even a part of

Benjamin Ard14:00to.

Matt Zelasko14:20AI, because it's not intelligent. It's not self aware. Doesn't know the nuances that are required to be intelligent, right? It's, yeah, it's, I guess, and then, and then probably another hour of conversation that I would go down. I would go down that path of more, but you know.

Benjamin Ard14:39I love that. Well, I love focusing on the problem. And then I also love like really diving in understanding the technology, especially with AI, because I do think you're spot on where it can feel like a black box and to say, okay, I know what a large language model can do, which means I can understand how it's using the data to do what it's going to do and help the business things of that nature. And so it's kind of a fascinating way to look at it. I like kind of that nuance of saying, hey, let's talk about what an AI does. Let's talk about what a large language model is. Let's see if based off of that definition, you feel like it's really actually going to fill in some gaps that you feel like you have, which I think is really powerful.

Matt Zelasko15:07Mm-hmm. And I think it will. I think it may be different gaps. You know, um, just the other day I, you know, I had a prospect and a client asked me two very similar things about, hey, do you, do you do this? Right. It was content writing and, um, a client was asking about, can we have an agent, an AI agent? So, you know, whatever. Write our case studies for us. And I said, do you really want that? Cause like a case study, okay. You know, like maybe a really high level bullet point case study. It's great, but they didn't have the experience. Now, if you designed it and, and, and had it on every single phone call that you had with the client and that, you know, and knew everything. That's powerful, but also it's going to take a lot.

Benjamin Ard15:50Yeah.

Matt Zelasko16:03To develop that and you're not gonna see a return on that for a year because you need that much context. Let's start doing that. And then I had another, it was a prospect, ask me, do you guys do RFP writing? And I was like, no, like absolutely not. You could train an agent to do that. Absolutely, build an agent for that because RFPs are like, I mean, everybody follows pretty much the same thing. Go do that and have, you know, like, it would take you a couple of weeks, 10 RFPs. It'll write them for you the way you want them written and like you're given the inputs that it needs. You know, don't have to come up with what the inputs are because somebody's giving those to you. So like that's where you use it. So.

Benjamin Ard16:46I love it. That's amazing. Well, Matt, as promised, these episodes go by quick. We have run out of time. Thank you so much for the insights today. For anyone looking to reach out and connect with you online, how and where can they find you?

Matt Zelasko16:57LinkedIn. Yeah. Yeah. I think it's, you know, at Matthew Zelasko. You can look up probably the Tom De long of rev ops that that'll, you know, surface me that was something that I created many years ago and I'm kind of stuck with so

Benjamin Ard17:13Love it. Well, for anyone listening, scroll down to the show. Now it's regardless of what platform you're listening on. We will link to Matt's LinkedIn profile. Connect there, Matt. Again, I really do appreciate the time and insights today.

Matt Zelasko17:24Dude, I really enjoyed the conversation and thanks for having me.

Benjamin Ard17:28You bet.

About the guest

Matt Zelasko

Matt Zelasko

Founder, Radish

Matt Zelasko is the founder of Radish, a horizontal growth agency that helps clients 'do more rad shit' and take the next step in their growth. He has spent longer than he would like to admit in the agency, marketing, and RevOps space, and is better known on LinkedIn as 'the Tom DeLonge of RevOps.' Matt is opinionated, willing to be proven wrong, and believes you can take the work seriously without taking yourself seriously. Before running Radish he worked as a copywriter, and still leans on old-school habits like writing 50 to 100 taglines by hand before reaching for any tool.

Connect on LinkedIn

Continue Exploring

Sales Enablement Software Buyer's Guide

A detailed guide to sales enablement platforms, rollout patterns, and what actually improves rep performance.

Read the guide

Get new episodes in your inbox

Join listeners who get episode summaries, key takeaways, and content strategy insights every week.

Frequently Asked Questions

Matt's argument is that most revenue teams are using AI as a technology to keep doing the exact same thing they were already doing — the gated ebook, the months-long content production cycle, the inbound playbook — and then falsely blaming AI for the falling engagement. He points out the playbook had already been a case of diminishing returns since roughly 2018, well before LLMs were mainstream. Content is at peak saturation right now, and LLMs are not stealing as much search traffic as people claim. The silver bullet thinking is the trap: AI is being used as the cause of the problem and the solution to the problem at the same time, which is silly.

Context engineering is Matt's reframing of how to work with an LLM. Instead of trying to write a perfect prompt, you tell the model what you want to do and then ask 'what else do you need from me?' That flips the dynamic — the model surfaces the missing context before it generates a confident wrong answer. Matt's underlying insight is that an LLM is basically speculating and guessing what comes next, so the quality of the output is the quality of the context you fed it. This matters most when you are not the subject-matter expert, because you will not catch the model being wrong on your behalf.

Matt uses two recent client requests as the contrast. A client wanted an AI agent to write case studies — he said no, because a good case study requires captured experience the agent does not have, so you would need it on every client call for a year before it could write one well, and the ROI is a year out. Another client wanted an agent to write RFPs — he said yes, because RFPs follow a standardized format, the inputs are handed to you by the prospect, and ten RFPs of training would be enough to write them the way the client wants. The rule of thumb: build agents where the inputs are bounded, the format is standardized, and the ROI does not depend on captured tacit knowledge.

Matt says AI helps creative people spend more time on creative pursuits by absorbing the work that does not need them — drafting policies, generating filler, breaking writer's block when they are genuinely stuck. He is careful, though, about giving up the ideation muscle. As an old copywriter he still writes 50 to 100 taglines by hand starting with the worst ones he can think of, because that process is what made him a good writer in the first place. He compares himself to Rick Rubin — 'I know it when I see it' — and warns that if you only ever ask AI for 50 taglines, that taste-making instinct never gets built. AI is best as the unstuck-tool, not the front door to ideation.

EP 45217 min

How to build a sales and marketing feedback loop that actually works with Nat Norris

with Nat Norris

Most marketing teams hand sales a stack of brochures and never hear back. In this Content to Close episode, Nat Norris, VP of Marketing and Customer Success at Model 1 Commercial Vehicles, breaks down how his team gets out of the trophy case of unused white papers and into the rooms where deals are won and lost. Nat walks through how he embeds marketers inside the company's three sales segments — public, commercial, and retail — why he forces his team into weekly quote review and deal loss meetings, and the data hygiene work he had to do in Power BI before any of it could function. He also shares the FAB framework (Features, Advantages, Benefits) he carried over from his catalog days, with a simple rule: push the benefit, self-serve the feature. And he closes with the two governors he uses to decide when there's enough data to act: the 80/20 rule and the 'front page of the newspaper' worst-case test. If you've ever wondered how to turn marketing collateral into something sales actually uses, this one's for you.

May 15, 2026Listen
EP 44414 min

How to use AI across sales, marketing, and customer success with Richmond Taylor

with Richmond Taylor

Most people use AI like a chatbot: one short prompt, a back-and-forth, and a mediocre output that gets worse the longer the thread runs. In this Content to Close episode, Richmond Taylor breaks down a smarter way to think about AI across the whole go-to-market motion. Richmond uses the Feynman technique to simplify go-to-market into three connected functions, sales is how you speak, marketing is how you look, and customer success is how you get the second date, and explains where AI can take over 80 percent of the work in each. He digs into why prompt engineering is the single skill that determines whether AI helps you or hallucinates on you, walks through the four prompt categories (system, user, developer, assistant), and explains why one big detailed prompt beats twenty short follow-ups every time. If you want a practical view of where AI fits inside a real business cycle, and how to stop wasting tokens on prompts that contradict themselves, this episode is worth your time.

May 1, 2026Listen
EP 43621 min

How to Operationalize Discovery So Buyers Sell Themselves with Nick Lopez

with Nick Lopez

Most discovery calls fall apart for the same reason: the seller asks surface-level questions, bounces between topics, and then defaults to pitch mode the second things get quiet. In this Content to Close episode, Nick Lopez walks through a discovery framework that fixes all of it. Nick teaches the 'pillar' approach, where every question you ask drills deeper into one topic before moving on, so you actually uncover the pain instead of skimming past it. He explains the 80/20 listening rule, why personal pain matters more than company pain, and the specific questions that get prospects to sell themselves on your solution. Nick also shares how to build a clean handoff from sales to customer success so the rapport you built in discovery doesn't get lost the second the deal closes.

April 17, 2026Listen

Get new episodes in your inbox

Join listeners who get episode summaries, key takeaways, and content strategy insights every week.