// BLOG
Notes on turning technology into a company.
Writing on AI economics, marketing measurement, search, commercial strategy and the Chief AI Officer role. The articles separate research findings, my operating advice and the work described in the case studies.
Start here · Guides
Chief AI Officer: What the Role Is, What It Owns, and When You Need One A Chief AI Officer guide to the mandate, operating scope, hiring questions, first-quarter work, and fractional option. A commercial view, with role and evidence limits. AI Commercialization: How to Turn a Working Model Into a Business AI commercialization: pricing, account-level costs, go-to-market and a story buyers can repeat. A guide with dated funding and margin research.All posts
62% of Marketing Agencies Offer AI-Related Services. Why Hire Yours?A review of 350 agency brands found 62% offering AI-related services. What clients should pay for beyond AEO reports, software access and AI labels.Stop Paying People to Copy Numbers Into SlidesAutomate the collection, keep the decision in human hands. What a CEO-ready weekly brief needs, where AI helps, and which tools may be worth buying. Your Marketing Agency Sells AI. Where's the Proof?A review of 350 selected marketing agencies found 217 advertising AI services or tools. Named people, dated projects and proof to include in proposals. The Marketing Learning Loop: How to Record What AI Actually LearnedRecord what changed, preserve the outcome and make the next AI marketing recommendation explain what it learned. Seven fields and a dated review procedure.AI Can Find Demand. It Cannot Create Capacity.Before AI marketing recommendations change spend, check sales capacity, inventory, service hours and cash. Name the owner and set limits the business can meet. Sometimes your competitor should get the AI recommendationChoose AEO prompts around buyers worth pursuing. Separate page corrections, acquisition decisions and product opportunities before funding the content plan. Who Owns an AI Marketing Decision?Name the human owner, define what AI can recommend, draft or change, and record the approval before it acts. Why Your CRM, GA4 and Ad Platforms Disagree About the Same CustomerReconcile CRM stages, GA4 events, ad-platform credit, sales activity and financial outcomes before an AI workflow recommends a budget change. The Test Before the A/B TestPre-test validation helps teams decide what to A/B test, define the control, audience, primary metric and stop rule, and reject weak tests before launch. Jev AI: What It Can Do, What It Can’t, and What the Viral Demos Leave OutCan Jev analyze ads or predict results? Its capabilities, limits, security risks, and how I would use real campaign data to build a better system. Build an AI Marketing Operating Model Around the Cross-Team Decision Define ownership, evidence, permissions, Sales capacity and outcomes for recurring marketing decisions, with practical criteria for renting, building or refusing AI software. Marketing Agency Predictions for 2027: Show Me the Contract A public review of 50 agency brands asks six procurement questions about AI platforms, client control, measurement, portability and pricing. The AI Audit Log: Reconstructing Why the Model Said That An AI audit log for commercial teams: what to capture for AI decision traceability, why model aliases break reproducibility, and how to log LLM outputs for compliance. What to Connect and What to Refuse MCP connector security and Model Context Protocol governance for commercial teams: what AI agents should read, what needs a human approval, and what should stay disconnected. The Human Gate That Isn't a Rubber Stamp Human-in-the-loop AI review needs evidence and authority to change or stop the next action. Approval rate alone cannot show whether the review works. What Actually Enters SOC 2 Scope When You Automate With AI Scope AI provider use against the actual data flow, audit boundary, and active configuration. Retention settings support evidence but do not decide the audit scope. The 2027 Marketing Stack Separate the martech product census, dated utilization survey, and AI-maturity categories. Evaluate each renewal by its own use, value, quality, and control requirements. AI Governance for Commercial Teams AI governance for commercial teams: enforce access boundaries, test reviewer authority, capture decision traces, and define audit scope. IBM findings and their limits. The AI Budget Trap AI readiness and budget share are associated in Gartner's survey. A practical funding sequence meters workflow cost, tests review, and measures accepted outcomes. The Last CMO Dated CMO tenure, title, and AI-leadership studies show different roles and samples. Define commercial authority, technical dependencies, and budget before setting targets. Proving AI Pipeline to Your Board Preserve observed AI-source evidence through CRM handoffs, separate attributed bookings from revenue and cash, and expose missing or conflicting records. Personalization After the Cookie Reversal Chrome cookie choices, retired Privacy Sandbox APIs, scoped AI transparency duties, and the arrival-time personalization problem. Demo figures are modeled projections. AI Referral Value by Engine: Revenue, Conversion, and Conflicting Figures Compare revenue per visitor with defined conversion events, check conflicting vendor arithmetic, and measure each AI assistant in your own funnel. AI Search Traffic: What Queries, Clicks and Pipeline Can Tell You Search queries, assistant-platform visits, website clicks, and revenue measure different things. Use dated evidence and property-level outcomes to plan search work. Why GA4's AI Channel Still Undercounts GA4 classifies recognized assistant referrals but groups Google AI search clicks with Organic Search. Check source coverage and reporting periods before interpreting growth. Marketing to Machines: When AI Agents Become the Customer Separate assistant discovery, merchant checkout, and completed agent transactions. Check the dated evidence, llms.txt limits, and capability requirements before planning scale. Dark AI Traffic: What the 70 Percent Estimate Can Prove Check the 70.6 percent vendor estimate, distinguish later AI-influenced visits from missing click referrers, and keep crawler activity separate from human demand. What to Ask When Hiring a Chief AI Officer Four Chief AI Officer hiring questions, a production artifact review, and reference checks. Test contribution, judgment, business results, and the authority the role needs. The AI Attribution Gap: Known Referrals and Missing Sources How AI referrals lose source information, what GA4 can classify, and how to separate observed traffic from vendor estimates and attributed revenue. The 2027 CMO: What Marketing Leaders Must Prepare For Six completed articles separate survey findings, forecasts, and operating recommendations for AI budgets, agent commerce, search, personalization, leadership, and renewals. AI Gross Margins: Rethinking the 90% SaaS Benchmark AI gross margins: dated executive projections and reported cohorts, the Bessemer and a16z debate, and account-level cost, routing and pricing decisions. Metering Cost Per Customer: The AI Unit-Economics Playbook Build account-level AI cost metering: request tags, provider charges, serving-cost allocation, cohort margins and alerts. An instrumentation playbook. The First 90 Days of a Chief AI Officer A Chief AI Officer first-quarter plan: meter spend, set review boundaries, agree budget and decision rights, and evaluate one commercial workflow against a CFO-owned metric. Why Chief AI Officer Hires Fail Three Chief AI Officer hiring risks: an undefined mandate, accountability without authority, and a profile that does not fit the work. Survey evidence and its limits. What Investors Look For in an AI Startup at Seed and Series A Seed and Series A investor questions on demand, revenue, margin, retention and efficiency. Dated Carta, SVB, ICONIQ and Bessemer benchmarks with cohort labels. Who Should Own AI: CAIO vs CTO vs CIO A proposed division of internal AI operations, product engineering, and commercial outcomes among the CIO, CTO, and Chief AI Officer. Survey evidence and mandate limits. The Commercialization Gap: Why AI Pilots Need an Economic Test MIT NANDA's preliminary 2025 study describes weak enterprise AI returns. A separate framework for ownership, pricing, narrative and unit economics. Fractional Chief AI Officer: Scope, Cost Questions, and When to Hire One (2026) When a fractional Chief AI Officer fits: define the decisions, authority, availability, and fee basis before comparing a part-time engagement with a full-time seat. Chief AI Officer vs Chief Data Officer: Who Owns What A proposed division of AI commercialization and data stewardship between the Chief AI Officer and Chief Data Officer, with overlap, mandate, and reporting questions. Usage-Based vs Seat-Based Pricing for AI Products Compare seat, usage and hybrid AI pricing against delivery cost and buyer predictability. Seats can work with stable usage and a small compute share. How to price an AI product before Series A. Prepare AI pricing before Series A: account-level delivery costs, a tested usage or hybrid rate, retention and correctly labelled fundraising benchmarks. What a Chief AI Officer Actually Owns A commercial Chief AI Officer mandate: project decisions, pricing, cost to serve, review boundaries, and continuing production work. Define authority before hiring. Why Your AI Startup Isn't Making Money Yet A working AI product still needs pricing, account-level delivery costs and repeatable sales. What to check before the next raise, with dated research. Category creation is a proof system, not a positioning exercise. Category creation needs customer language, visible adoption, pricing buyers can evaluate, and evidence behind the name. The crypto-mining case shows the contribution boundary. AI pricing has to respect compute reality. Price AI against variable delivery costs, customer value, support and expansion. Flat seats can work with stable usage; variable costs may need a hybrid. The market does not buy complexity. It buys a story it can repeat. Help a buyer explain a complex product to their CFO, committee or partner. A retelling test and four narrative layers, with the technical detail preserved. AI demos are easy. AI businesses are hard. A useful AI demo still needs a buyer, pricing that covers delivery, data and workflow advantages, and a repeatable path to accepted revenue.