# Writing

Notes on AI strategy, AI engineering and GTM automation by Colin Gillingham.

- [Use technographic data to qualify accounts before your reps waste a single call](https://colin.gillingh.am/writing/gtm-technographic-qualifier) (2026-04-17): Firmographics tell you the company fits on paper. Technographics tell you whether they'll actually buy your product. Here's how to turn tech stack data into a qualification signal that routes the right accounts to your pipeline.
- [The Three Kinds of AI Work Your Company Is Actually Doing](https://colin.gillingh.am/writing/the-three-kinds-of-ai-work-your-company-is-actually-doing) (2026-04-16): Exploration, implementation, and operations are three different jobs. Most companies staff them as one — and that's why the AI team is always behind.
- [Why the ROI Framing Kills Good AI Projects](https://colin.gillingh.am/writing/why-the-roi-framing-kills-good-ai-projects) (2026-04-15): Demanding ROI at the pilot stage selects for safe, measurable, low-value work. Here's what to demand instead.
- [Job postings are your best free source of buying intent — here's how to use them](https://colin.gillingh.am/writing/gtm-job-posting-analyzer) (2026-04-14): A company hiring a VP of Revenue Operations is telling you something. Here's how to build a workflow that reads those signals automatically and writes them to HubSpot before your competitors do.
- [What Responsible AI Actually Requires Operationally](https://colin.gillingh.am/writing/what-responsible-ai-actually-requires-operationally) (2026-04-14): Responsible AI isn't a principles document. It's a set of operational decisions about who reviews what, when, and what happens when something breaks.
- [The AI Literacy Gap Is a Leadership Problem](https://colin.gillingh.am/writing/the-ai-literacy-gap-is-a-leadership-problem) (2026-04-13): AI literacy programs aimed at individual contributors miss the point. When leaders can't evaluate AI claims, everything downstream breaks.
- [How to Scope an AI Project Before You Quote It](https://colin.gillingh.am/writing/how-to-scope-an-ai-project-before-you-quote-it) (2026-04-12): AI projects blow their budgets because the scope was never real to begin with. The framing questions you need to ask before a number is ever written down.
- [Build a competitor analyzer that keeps your CRM honest](https://colin.gillingh.am/writing/gtm-competitor-analyzer) (2026-04-10): Most competitive intel lives in a slide deck from last quarter. Here's how to build an n8n workflow that monitors competitor signals automatically and writes structured analysis to HubSpot.
- [The Six Things Every AI Strategy Document Should Answer](https://colin.gillingh.am/writing/the-six-things-every-ai-strategy-document-should-answer) (2026-04-10): Most AI strategy docs answer questions nobody asked. Here are the six that force real decisions — and what good answers actually look like.
- [How to Think About AI Risk Without Paralyzing Your Team](https://colin.gillingh.am/writing/how-to-think-about-ai-risk-without-paralyzing-your-team) (2026-04-08): Risk-aversion is real and valid. But there's a version of risk management that enables action instead of blocking it. Here's the framing I use.
- [What I Look for in the First Week of an AI Engagement](https://colin.gillingh.am/writing/what-i-look-for-in-the-first-week-of-an-ai-engagement) (2026-04-07): The first week of an AI engagement tells you almost everything. Here are the five signals I watch — and what they actually mean.
- [Agents vs. Automation: How to Know Which One You Actually Need](https://colin.gillingh.am/writing/agents-vs-automation-how-to-know-which-one-you-actually-need) (2026-04-06): Most companies deploy agents when they need automation, or automate something that requires judgment and wonder why it keeps breaking. Here's the test that tells you which one you actually need.
- [How to Write an AI Job Description That Attracts Real Talent](https://colin.gillingh.am/writing/how-to-write-an-ai-job-description-that-attracts-real-talent) (2026-04-05): Most AI job descriptions fail before the first candidate reads them. Here's how to write one that signals you've actually done the thinking.
- [Build an email sequence writer that actually sounds like you wrote it](https://colin.gillingh.am/writing/gtm-email-sequence-writer) (2026-04-03): Most AI-written outbound reads like AI-written outbound. Here's how to build a sequence writer in n8n that pulls HubSpot data and generates openers your prospects will actually reply to.
- [What a Good AI Proof of Concept Looks Like](https://colin.gillingh.am/writing/what-a-good-ai-proof-of-concept-looks-like) (2026-04-02): Most AI POCs prove nothing useful. Here's how to design one that actually tells you whether AI will work in your context.
- [How to Train Your Team to Use AI Without Making It Mandatory](https://colin.gillingh.am/writing/how-to-train-your-team-to-use-ai-without-making-it-mandatory) (2026-04-01): Mandating AI use creates compliance, not adoption. The companies getting real results are building pull, not push.
- [Build an objection handler that knows the deal before your rep does](https://colin.gillingh.am/writing/gtm-objection-handler) (2026-03-31): Most AI sales tools serve canned responses to objections. Here's how to build one that pulls deal context from HubSpot and generates responses specific to the prospect, the stage, and the actual concern.
- [Judgment Work Is Safe (For Now)](https://colin.gillingh.am/writing/the-job-titles-that-will-disappear-and-the-ones-that-wont) (2026-03-31): AI isn't replacing jobs wholesale — it's moving the line between assembly work and judgment work. The problem is that line keeps moving.
- [What Good AI Governance Looks Like for a 50-Person Company](https://colin.gillingh.am/writing/what-good-ai-governance-looks-like-for-a-50-person-company) (2026-03-30): Governance sounds like enterprise bloat. It's not. Here's the minimal AI strategy every 50-person company needs before something goes wrong.
- [How to Run an AI Sprint in a Non-Technical Company](https://colin.gillingh.am/writing/how-to-run-an-ai-sprint-in-a-non-technical-company) (2026-03-29): You don't need engineers to run an AI sprint. Here's a five-day format that works for ops, marketing, and product teams with no ML background.
- [The Difference Between an AI Feature and an AI Product](https://colin.gillingh.am/writing/the-difference-between-an-ai-feature-and-an-ai-product) (2026-03-28): Most 'AI products' are just products with a GPT call in the middle. Here's how to tell the difference — and why it changes everything about how you build.
- [Build a data cleaner that stops your GTM workflows from eating garbage](https://colin.gillingh.am/writing/gtm-data-cleaner-normalizer) (2026-03-27): Your ICP scorer, personalization writer, and routing logic all downstream of the same CRM data. When that data is wrong, everything built on top of it is wrong too. Here's how to build a normalizer in n8n that fixes it at the source.
- [Stripe Projects: Agents Can Now Provision Their Own Stack](https://colin.gillingh.am/writing/stripe-projects-agents-can-now-provision-their-own-stack) (2026-03-27): Stripe just shipped a CLI that lets agents sign up for services and generate credentials without any human involvement. That's not a time-saver. That's infrastructure.
- [Why Smaller Companies Have an AI Advantage Right Now](https://colin.gillingh.am/writing/why-smaller-companies-have-an-ai-advantage-right-now) (2026-03-27): Enterprise AI adoption is slow because of legacy systems, politics, and process. Small companies can move faster — if they're intentional about it.
- [The Questions to Ask Before Any AI Vendor Demo](https://colin.gillingh.am/writing/the-questions-to-ask-before-any-ai-vendor-demo) (2026-03-26): Every AI vendor demo looks great. That's the point. Here are the questions that tell you whether the product matches the pitch.
- [What "AI-Native" Actually Means for a Product Team](https://colin.gillingh.am/writing/what-ai-native-actually-means-for-a-product-team) (2026-03-25): Most products aren't AI-native — they're normal products with AI features bolted on. Here's the test that tells you which one you're building, and why the distinction matters.
- [Build a LinkedIn profile analyzer that knows who to prioritize](https://colin.gillingh.am/writing/gtm-linkedin-profile-analyzer) (2026-03-24): Your SDRs spend the same five minutes on every LinkedIn profile. An AI analyzer spends two seconds and tells them which ones actually matter.
- [How to Get Your Exec Team Aligned on AI](https://colin.gillingh.am/writing/how-to-get-your-exec-team-aligned-on-ai) (2026-03-24): When everyone in the room is 'pro-AI' and nothing is moving, the problem isn't the technology. Here's how to close the four gaps that are actually blocking your AI strategy.
- [The Hidden Cost of Not Having an AI Strategy](https://colin.gillingh.am/writing/the-hidden-cost-of-not-having-an-ai-strategy) (2026-03-23): The absence of an AI strategy isn't neutral. It's a series of compounding decisions being made without anyone in the room.
- [When to Build vs. Buy AI (The Framework I Actually Use)](https://colin.gillingh.am/writing/when-to-build-vs-buy-ai) (2026-03-22): Build/buy decisions in AI aren't like SaaS decisions. The framework most companies use produces bad answers. Here's the one that doesn't.
- [Why AI Projects Need a Product Manager, Not a Project Manager](https://colin.gillingh.am/writing/why-ai-projects-need-a-product-manager) (2026-03-21): Most AI projects are run like software migrations. That's why they produce artifacts instead of outcomes.
- [Build an AI-powered ICP scorer that handles the fuzzy stuff](https://colin.gillingh.am/writing/gtm-icp-scorer) (2026-03-20): Traditional lead scoring counts page views. ICP scoring asks a harder question: is this company actually a good fit? Here's how to build one in n8n that uses AI for the judgment calls HubSpot's native scoring can't make.
- [The Stack a Modern AI Strategy Actually Runs On](https://colin.gillingh.am/writing/the-stack-a-modern-ai-strategy-actually-runs-on) (2026-03-20): Not a vendor ad. A real breakdown of the tools, models, and infrastructure patterns that are working right now — and the one thing most companies consistently get wrong.
- [AI ROI: Why Most Companies Measure It Wrong](https://colin.gillingh.am/writing/ai-roi-why-most-companies-measure-it-wrong) (2026-03-19): Most companies measure AI ROI with a cost-savings lens. That lens was built for a different kind of automation. Here's what you're missing.
- [How to Pick Your First AI Use Case](https://colin.gillingh.am/writing/how-to-pick-your-first-ai-use-case) (2026-03-18): The best first AI use case isn't the most impressive one. It's the one that teaches you something. Here's the filter I actually use.
- [Build a personalization writer that actually sounds human](https://colin.gillingh.am/writing/gtm-personalization-writer) (2026-03-17): Most AI-written outbound reads like AI-written outbound. Here's how to build a personalization engine in n8n that pulls HubSpot data and generates openers your prospects will actually reply to.
- [What I Tell Every Company in the First Meeting](https://colin.gillingh.am/writing/what-i-tell-every-company-in-the-first-meeting) (2026-03-17): The mistakes are identical even when the companies aren't. Here are the five things I say in every first AI consulting conversation.
- [The Difference Between an AI Strategy and an AI Plan](https://colin.gillingh.am/writing/the-difference-between-an-ai-strategy-and-an-ai-plan) (2026-03-16): Most companies think they have an AI strategy. What they have is a project list with 'AI' written at the top. Here's why that distinction costs you.
- [The Rise of Agent-Centered Design](https://colin.gillingh.am/writing/the-rise-of-agentx) (2026-03-16): A new kind of user just showed up. It's intelligent, it scales instantly, and it doesn't need your UI. This is the high water mark for human-centered design — and the start of something faster.
- [Why Your AI Pilot Failed (and How to Fix It)](https://colin.gillingh.am/writing/why-your-ai-pilot-failed) (2026-03-15): AI pilots fail for predictable reasons: wrong metric, wrong team, wrong timeline. Here's the pattern — and how to design one that actually produces something.
- [Build a company research agent that replaces 12 open tabs](https://colin.gillingh.am/writing/gtm-company-research-agent) (2026-03-14): Your reps spend 30-60 minutes per account copying data between browser tabs. Here's how to build an AI agent that does the same research in under a minute and writes it straight to HubSpot.
- [The AI Audit: What to Do Before You Build Anything](https://colin.gillingh.am/writing/the-ai-audit-what-to-do-before-you-build-anything) (2026-03-14): Before writing a line of code or buying a single tool, there are five questions every company needs to answer. Most skip all of them.
- [From SDR to GTM Engineer: The Career Shift That's Already Happening](https://colin.gillingh.am/writing/sdr-to-gtm-engineer) (2026-03-13): 90% of the SDR function is being absorbed by AI. That's not a prediction — Vercel already did it. Here's what the people who understand this move do next.
- [How to Evaluate an AI Consultant (Without Getting Burned)](https://colin.gillingh.am/writing/how-to-evaluate-an-ai-consultant) (2026-03-13): The AI consulting market is full of people who read one paper and built a Linktree. Here's what actually separates good from bad.
- [Your First 90 Days with an AI Strategy](https://colin.gillingh.am/writing/your-first-90-days-with-an-ai-strategy) (2026-03-13): Most AI strategies die in slide decks. Here's the 90-day arc that actually moves from strategy to shipped product.
- [When Agents Outnumber Humans](https://colin.gillingh.am/writing/agents-outnumber-humans) (2026-03-11): 51% of internet traffic is already bots. The agent-addressable market is larger than the human one. Every SaaS assumption built on the seat is quietly breaking.
- [Business Flywheels: Use AI to Autonomously Improve Performance](https://colin.gillingh.am/writing/autoresearch-business-flywheel) (2026-03-09): Andrej Karpathy's 'autoresearch' pattern is a secret weapon for business: AI that improves itself every night. Here are 8 business use cases for this loop.
- [Humanoid Robots for Software: Why Agents Don't Need Your API](https://colin.gillingh.am/writing/humanoid-robots-for-software) (2026-03-08): Computer use agents inherit every interface humanity ever built. They don't need your permission, your API, or your integration roadmap.