# You keep using that word. I do not think it means what you think it means. *I'm so tired of talking about AI. But since we can't get away from it, let's see if we can move the conversation forward.* --- For the last three years, my professional life has been fully inundated by three types of AI content, loosely bucketing out to “Here's why AI is good/bad” and “Here's a guide to do niche thing X with AI” or “Here’s why you don’t need X function because now we have AI.” I find most of this content isn’t very useful. We’re at the point in the technological hype cycle that I know a few things beyond a shadow of doubt, namely: - Feel how you want to feel about AI, but it’s here, and that bell isn’t likely to be unrung - We’re in the App Store Flashlight stage, everyone’s building “AI-powered” products and flooding the zone with niche solutions or workflows. - Despite what you hear about agents, multi-context protocols, etc., etc., humans are very much still required Lately I’ve been exploring a more useful (to me) line of inquiry. I don’t need a litany of things I _could_ do with AI, or another hot take on why it’s good or bad. What I need is some informed first principles to help govern my relationship with AI. As a GTM partner that lives in the space where strategy and technology intersect, this line of inquiry is very much [alive for me right now](https://damrunner.com/thinking/questions/what+remains+human). I have some deep philosophical, ecological, and practical concerns, but I’m not a luddite. Technology is _cool._ It’s a core pillar of my practice as an independent strategist, and I also enjoy eating food and having a job and such, so I've started to develop a few practical rules that help me navigate my day-to-day. First, let me share a few fundamental beliefs: We need to understand that we’re still very early in the game. If your feed is anything like mine, AI is going to come for me tomorrow. But in the real world this just [isn’t the case](https://sherwood.news/tech/ai-experts-think-everyone-uses-ai-all-the-time-we-dont/?utm_source=snacks&utm_medium=email&utm_campaign=snacks_20250407&utm_content=b3f5fccd7114b2f9d7fcd459bcce98b7). When AI experts were asked in a recent Pew Research survey how often they thought Americans used AI in their daily lives, 79% said they were totally sure that regular Americans were using AI “almost constantly" (several times per day). Less than one-third of survey respondents said they used AI that much, and one-third of people said they had never used a chatbot at all. What you should read from this is that there is a small minority of *very loud* talkers that would have you believe that you are behind the curve, and you should examine why they would have you believe that. Second, you need to understand that there is some validity to the idea that AI is coming for your job. Setting aside judgment for a second, let’s be clear that ***knowledge work is forever changing as a result of artificial intelligence***. Zoom out far enough and this is true of all great paradigm shifts in technology. I've seen enough to know that no amount of hand-wringing will send this genie back to the lamp, and so it falls to us to both define what the future of work should be and, possibly more importantly, ***make it worth it.*** How you think about your self and your work and what makes you uniquely valuable needs to adapt. Thirdly, all of these things being true, the most practical way to navigate the moment is to fully engage with the exercise to develop your own deeply personal heuristics that guide your relationship with technology in general, and AI specifically. I wouldn't dream of telling you how to think or feel about it, but I'll gladly share my own perspective. ## Thinking is your moat At the end of the day AI is a statistical incarnation. It’s very good at spotting and matching patterns, and for vast categories of knowledge work this is exactly what’s required. But I also believe that there’s a difference between [thinking, computing, and simulating](https://damrunner.com/thinking/concepts/thinking%2C+computing%2C+simulating), and that it pays for us to be mindful of our own cognitive logistics as we navigate our day-to-day lives. For many years, we've defined "specialization" through functional [[skill premium|skill premiums]], but much of that work is at risk of replacement. All of this is extremely automatable, if not replaceable: - following or executing commands - curating and synthesizing existing knowledge in an objective way - pattern spotting, matching, or recreating But we also know that [[Ai Models Collapse When Trained on Recursively Generated Data - Clips|AI models "collapse" when trained on recursively generated data]]. So the more we feed AI models with content _they_ created the less capable they are of understanding the real world. Thinking, *real thinking*, is [embodied](https://damrunner.com/thinking/concepts/embodied+cognition) and cognitively expensive, which is why it's our most defensible and deepest moat. Here's an example for my work: I like to think strategy is primarily about [anticipating how ideas move through the real world](https://www.are.na/block/36901021). Doing this well is deeply contextualized in a lived experience. Everything required to succeed—generating or synthesizing net-new knowledge, applying judgment, testing and breaking patterns—is uniquely human domain. The tools, frameworks, and methods I use can and will evolve with the technology available to me, but thinking above and beyond my capacity to technically execute something is ultimately what determines my skill premium. ## Evaluating Value x Risk In 2023 Cody Doctorow introduced a [2x2 grid for evaluating AI applications](https://pluralistic.net/2023/12/19/bubblenomics/?__readwiseLocation=) with “value” and “risk tolerance” as the primary axes. This is a fairly evergreen way to evaluate most things in life: is the juice worth the squeeze? The devil's in the details, of course, and how we parse risk and value is highly subjective. My position is that there’s plenty of inherent risk with automation and AI, and a lot of it isn't new. Consider the myriad ways in which technology obscures, obfuscates, and ultimately controls. AI introduces some new flavors, but much of this is progression on an existing arc of [[enshitification]]: - our inboxes overflowing with orders of magnitude more cold outreach and [phishing attacks](https://www.techtarget.com/searchsecurity/tip/Generative-AI-is-making-phishing-attacks-more-dangerous) - [vibe coders nearly bankrupting themselves](https://nmn.gl/blog/vibe-coding-fantasy) - [would-be mushroom foragers putting their literal lives at risk](https://www.theguardian.com/technology/2023/sep/01/mushroom-pickers-urged-to-avoid-foraging-books-on-amazon-that-appear-to-be-written-by-ai). Working with AI, even in mundane ways, often introduces “invisible complexity gaps” where a combination of hallucinations, an overabundance of slop content, [[Towards Understanding Sycophancy in Language Models - Clips|institutionalized sycophancy]], and a lack of subject matter expertise creates costly outcomes. Here we can define “costly” in a broad way—people may form negative associations with your brand or they might die because the book you co-wrote with AI told them to eat a highly poisonous death cap mushroom. Cost is also a dimension of risk to consider. Some of our personal choices for how we use AI may be [[A Cheat Sheet for Why Using ChatGPT Is Not Bad for the Environment - Clips|relatively inconsequential]], but we might also consider how these small choices shape and determine collective action. [One MIT study from January 2025](https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117) found that the electricity of data centers rose to 460 terawatts in 2022—which would have made data centers the 11th largest electricity consumer in the world. Yes, this is only partly driven by artificial intelligence. Yes, closed loop cooling and renewable energy investment could address this ([[How AI Demand Is Draining Local Water Supplies - Clips|at the expense of more carbon]]). Yes, we can [argue for how explosive economic growth](https://arxiv.org/pdf/2309.11690) will, in aggregate, offset the costs. All of these things could be true, but given that they’re all components of immensely complex geo-political and social systems, they're all [hyperobjects](https://damrunner.com/thinking/concepts/hyperobjects) to me. I'm not going to pretend I know which levers to pull and what happens when I do. The point is this: these are all value judgments that can help you understand the cost and complicity of a practice. Understanding how you and your work fits gives you a firm grounding to decide what is or isn't worth the squeeze. ## Technology is a spectrum Here we arrive at the eponymous nexus of this article. As an independent GTM strategist, roughly 1 in 3 conversations I have boils down to: we want you to help us integrate AI into our product/how we work. Each time, I feel myself slowly morph into Inigo Montoya. <iframe width="560" height="315" src="https://www.youtube.com/embed/dTRKCXC0JFg?si=_9cnen6KBjPiPRRh" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe> Firstly, when it comes to product development I'm reasonably convinced that [[Composability will define the next SaaS era|the traditional SaaS ecosystem will progressively fold in on itself]], and layering AI into your standard "UX on top of a generic database" product isn't a viable long-term strategy. Secondly, when it comes to business operations, most early and growth stage companies are scrambling to "adopt AI" because they keep hearing it's what they're supposed to do, not necessarily because it's the best solution. Automation has been a part of GTM for a relatively long time now. One of my very first projects out of college (nearly 20 years ago) was onboarding Pardot and designing marketing drips. At the time it was bleeding edge stuff, but it has since become pretty pedestrian in B2B and SaaS, spinning off massive cottage industries, whole disciplines, and thousands of products. [Statista estimates](https://www.statista.com/topics/10768/marketing-automation/#topicOverview) the global marketing automation industry’s annual revenues around $8B in 2024, and projects it to reach $21.7B by 2032. This kind of technology [exists on a spectrum](https://www.linkedin.com/posts/akantjas_automation-ai-workflow-or-ai-agent-to-activity-7318265649863045120-TKTL/?utm_source=share&utm_medium=member_ios&rcm=ACoAAAK2eMcBGE3emCSS0NyUYAcGgnvtx38fnXg), from simple automation on the left toward complex AI agents on the right. Thinking about GTM solutions this way might seem overly pedantic, but it's consequential: the scope and cost of a "good" investment scales exponentially as you move to the right. Far be it from me to tell companies that want to run to walk, but it is my job to help them make good and impactful choices. Despite what the marketing says, the vast majority of specialized software tools aren't really agentic AI applications (and that's okay). What we call it matters, though, because the real question we want to answer is, "what do I need to unlock, and what's the most efficient way to do it?" In a lot of cases, technology isn't the answer and when it is, navigating those choices is a whole lot easier with good working language and a mental model. My perspective to solution-building here is pretty straightforward: we should almost always work our way from left to right on this spectrum to maximize value and minimize risk. For one, all of these systems are built on the same fundamental context. Before you can create an AI workflow you need to understand your broader data model, set boundaries, and understand the important inflection points for human oversight and intervention. Before you build or oversee an AI agent you should understand the workflows, decision points, and outcomes you're offloading. Secondly, the scope of the investment (in people, complexity, time, expertise, or straight up cash) compounds. I've been thinking on a term for the particularly silly tech behavior where we do [[updumping|costly re-engineering of perfectly workable solutions]] under the guise of innovation. This mad dash to fully replace GTM with AI everything fits the bill. The size and maturity of the marketing automation segment is a good indicator of why many GTM operators are reluctant to adopt the “GTM Engineering” language and joie de vivre. Anyone who's been operating in this game in a meaningful capacity over the last 15 years has long accepted the idea that GTM is an “engineering challenge.” The evolution of the tool kit doesn’t fundamentally change that. ## Some closing thoughts I'm not a luddite by any stretch of the imagination, I'm a big believer in the power of technology. But technology, like all tools, reflects our intent. If the last decade of growth was about how much we *could* do, the next one is about what we *should* do. I'm not a prophet, and I don't think there's an immutable playbook here, but I do know that we have to think carefully and act deliberately. --- ![[footer#^a63e8e]]