Thinking out loud.
Notes on building with AI, creative technology, and the things we're figuring out as we go.
Automate Invoicing: Tools, No-Code, or Custom (a Build-vs-Buy Decision Guide)
"Automate invoicing" is really two jobs — the invoices you send and the ones you pay — and the honest answer is a dial with five settings, not a purchase. Where most businesses should land, with the real per-invoice math.
Answer Engine Optimization Services (AEO, GEO & AI Search)
AEO, GEO, and AI-search optimization services from an agency that runs one on its own site — the deliverables, the public rate card, and when you shouldn't buy.
Slop as a Service vs. Service as a Software
There are two ways to sell AI-powered marketing, and they're opposites wearing the same clothes. One sells you the output with the human optional; the other sells you the outcome, supervised. The difference is who checks the work.
Automated SEO Tools vs. a Managed Agent: How to Actually Choose
Most "best automated SEO tools" lists rank the wrong thing. The question that decides whether a tool helps or hurts is simpler: who owns the checking? A straight three-way comparison, from a team that runs the third kind.
Which Claude Plan Should You Actually Buy? Pro vs Max vs Team vs Enterprise (Most Teams Over-Buy)
Most teams buying Team don't need it yet, and most reaching for Enterprise aren't ready. The honest decision guide to Claude's plans — Pro vs Max vs Team vs Enterprise — from an agency that runs them.
How to Use Claude Loops for Business: /loop vs Scheduled Tasks vs Routines
Claude has three ways to run a recurring job — Cowork scheduled tasks, Claude Code /loop, and cloud Routines. A practitioner's map of which to use when, from an agency running five loops in production.
Which Claude Model Should You Use? (Most of You Don't Need Fable)
An honest, cost-first guide to choosing a Claude model: what each costs, what the July 7 Fable change means, and why most people don't need the expensive model — from an agency that runs Claude in production daily.
The Future of Work Is Finding New Problems to Solve
As AI absorbs execution, the skill that appreciates is the one machines structurally lack: finding problems nobody has named yet. The conviction piece behind the automaton agency model — and why problem-finding is a learnable practice, not a gift.
Can AI Build My Website? An Honest Answer From an Agency That Builds Them
Yes — AI can build you a working website today, and for a simple site you should let it. The real question is which of the three honest paths fits your business, and what the cheap path costs you later. From an agency that builds with AI daily.
What Is an AI SEO Agency? An Honest Taxonomy, From an Agency That Runs One
Most "AI SEO agencies" use AI tools inside a familiar retainer; a few run agentic SEO programs where a system ships fixes daily with humans on strategy. We run one — here's the honest taxonomy, the costs, and the questions that sort the market.
llms.txt and How to Make Your Website Agent-Ready (the Honest Version)
llms.txt is real but barely read by AI engines — we measured. Google doesn't use it, crawlers fetch it on ~0.1% of visits, and 300,000 domains show no citation effect. The full agent-readiness stack, ranked by what actually earned our AI citations.
Is SEO Dead? What We Measured on Our Own Site
SEO isn't dead — the destination changed. When Google's AI Overview took over our head query, exact-match impressions briefly compressed 83%; six weeks on, the query's impressions have tripled and its CTR sits near 0.2%. Visibility up, clicks gone. The field report, with data.
AI Web Design: What It Actually Is (and What It Can't Do Yet)
AI web design is genuinely good at the production layer — layouts, copy, imagery, speed — and still can't do the judgment layer: distinctiveness, accessibility, taste. What it does well in 2026, where it fails quietly, and how to use it without shipping a site that looks like everyone else's.
What is an agentic website? (And what we did to make ours one)
An agentic website runs its own growth and lets AI agents operate it directly — not a site AI built once. What it is, how it differs from the agentic web and AI builders, WebMCP, and what's hype. From an agency running one.
Service as Software: The Next Great Companies Sell the Work, Not the Tool
Service as software is the model where AI delivers the outcome, not the tool — and the next great companies are built on it. An honest field report from a services firm already running this way, including where the popular thesis is wrong.
Claude Cowork for Business: The Operator's Guide (2026)
The operator's hub for businesses adopting Claude Cowork: what it's actually good at, what it costs, how to use it, when to buy versus build, and where it fits by function and vertical — routed to the deep-dive on each, from an agency that runs Cowork in production every day.
We pointed an autonomous agent at our own SEO for 90 days. Here's the citation data.
For 90 days an autonomous agent has run our own SEO and AEO program — daily. Here is the data nobody else publishes: the citation ledger, what actually moved (and what didn't), and the five things running the engine taught us about ranking in the AI-answer era.
How to use Claude Cowork: a daily-driver's field guide (2026)
Past the setup tutorials: how to actually use Claude Cowork day to day — the real setup (local folders + Drive + an OS that knows where things live), connectors and their limits, the autonomy settings, the recurring-task loops we run in production, and the three Claudes we work with. From a one-person agency that lives in it.
Generative engine optimization (GEO): a practitioner's field guide (2026)
Generative engine optimization (GEO) is the practice of structuring content and authority so AI answer engines cite you inside the answer they generate. What GEO is, the levers that actually earn citations, how to measure it with a citation ledger, and what doesn't work — from an agency running the program on its own site.
How to share an HTML file (the one your AI just made you)
Claude or GPT handed you a working HTML file — now what? How to turn it into a shareable link in 60 seconds, where to actually host it, why it shows as code in Slack and Docs, how the person you send it to opens it, and the one thing none of it solves yet: collaboration. From an agency that ships these every week.
What Is an Automaton Agency? The Definition We Created, Because We're Human.
An automaton agency automates its own production so the humans do only the creative work — solving hard problems and finding the ones nobody's named yet. The definition, from the agency that coined it — and why it's not the same thing as an AI automation agency.
How much does an AI automation agency actually make? (the economics, from inside one)
A well-run AI automation agency runs 50-70% gross and 20-35% net margins — not the 80% the hustlers promise. Where the money actually goes, why retainers exist, and the three ways to get automation capability: build in-house, hire a traditional agency, or embed a partner. Honest economics from inside one.
Answer engine optimization: the practitioner's definitional guide (2026)
Answer engine optimization is the practice of structuring content so AI Overviews, ChatGPT, Perplexity, voice assistants, and the featured-snippet box can extract a direct answer and cite you. What AEO covers across five surfaces in 2026, the format pattern that earns citations, and what we measure — from an agency running the program on its own site.
Best AI for accountants in 2026: the five categories that actually matter
There is no single best AI for accountants — there are five categories doing five jobs, and the right answer is a small stack, not one tool. The five categories that matter in 2026, including the one every listicle skips: general-purpose LLMs. Honest costs and field notes from an agency that runs them on its own books.
Zapier vs Make vs n8n: which automation layer your business actually needs in 2026
Zapier vs Make vs n8n decoded by an agency that ships client systems on all three: they're not competitors, they're three settings on one dial. The honest cost curves, the per-task and per-operation traps we've hit on real builds, and how to pick the setting your business is actually at.
AI proposal generator in 2026: buy, hybrid, or build (an implementer's decision guide)
Three paths, not two. Most teams should buy off-the-shelf or run a hybrid Claude/ChatGPT-plus-template setup; six conditions justify a custom build. Honest costs, real vendor names, and the decision rubric — from an agency that builds these.
AI tools for accountants: categorized by what they actually do
AI tools for accountants need a taxonomy, not a list. By mid-2026 there are 30+ tools claiming "AI for accounting" — categorized by the job they actually do, from AI-native ledgers to AP automation to the general-purpose LLMs (and the new Claude for Small Business skills) now closing the month. The map of the territory, from a firm that runs the stack.
AI Automation ROI: What to Realistically Expect in 2026
The real ROI of AI automation isn't the number your vendor quoted. It's what's left after implementation cost, change management, and the things automation quietly breaks. Here's what the 2026 data actually says — and what we've seen from the inside.
GEO vs SEO: What Each Surface Actually Rewards (A 60-Day Field Report)
GEO and SEO solve adjacent problems but reward different content shapes. After running both programs in parallel for 60 days on the same content, here is the operational view: five foundations stay identical, five tactical layers diverge, three measurement surfaces require different cadences, and the decision rubric is simpler than the LinkedIn debate suggests.
AI agency vs traditional agency: why the comparison is wrong
Every "AI agency vs traditional agency" comparison frames it as a choice between speed and creativity. The real question is whether your agency builds systems that compound — or deliverables that depreciate.
How to Rank in AI Overviews: An Implementer's Playbook (2026)
Three independent data sources point the same direction: Google's own May 2026 AI Search guide, Cyrus Shepard's Zyppy 23-factor analysis, and 60 days of our own field data. Five things that actually work in AI Overview citation — and four that don't move the needle despite what most guides say.
AI marketing agency vs traditional agency: the ROI comparison nobody publishes with real numbers
The real 2026 cost comparison — AI marketing agency vs traditional agency — with the five cost lines the headlines never show, where each model actually wins, and the third model that beats both. An implementer's ROI breakdown from an agency that publishes its own rate card.
What Is Claude Cowork? An Implementer's Field Guide (2026)
Claude Cowork explained by an agency that runs it in production — what it actually is, how it works, the seven real use cases, the honest limits, and how it differs from Claude Code, Claude chat, and the API.
AI Overview citation: 30-day field report, Part 2 — Google says FAQ schema isn't required. We keep getting cited.
Three days after Google published its first official AI Search optimization guide, our pillar on what a creative technology agency does showed its strongest AI Overview citation density yet — five inline cites across six sections, with no re-crawl in 24 days. The schema isn't doing the work. Here's what is, and what we're changing in the format-pattern operating system.
AI Due Diligence with Claude Cowork in 2026: An Implementer's Field Guide for M&A and PE Work
A practitioner's take on AI for M&A and PE due diligence in 2026. The four sub-workflows, the May 12 Claude Cowork Corporate Legal plugin treatment, honest comparisons with Thomson Reuters CoCounsel and Harvey, and a build-vs-buy framework for a 3-to-5-person M&A or PE team.
Claude Cowork for law firms: an implementer's take on the legal plugin (2026)
A practitioner's take on Claude Cowork's legal plugin for small and mid-market law firms. The five use cases that work in a 5-attorney firm, the three predictable failure modes, an honest comparison with Spellbook and Harvey, the will-AI-replace-lawyers elephant, and a 90-day deployment plan that satisfies state-bar AI rules.
The five-layer framework for business systems
Every system we build follows the same architecture. Here's why layer five is the only one that matters.
How much does a creative technology agency cost in 2026? (And why everyone else is vague about it)
The real 2026 price tag for a creative technology agency — written by one. Project, retainer, and sprint rates with actual numbers, what's inside each tier, and how the category compares to traditional creative and AI agencies.
How three of our pillars broke into AI Overview citation in 30 days (a field report)
In thirty days, three of our ten target queries for Answer Engine Optimization flipped from zero citations to AI Overview citation. Here's exactly what shipped, what indexed, what got cited, and the format pattern we extracted — including how the cadence runs and the human-versus-Claude division of labor that produced the work.
AI for wealth management: what actually works in 2026 (an implementer's field report)
A practitioner's field report on AI for RIAs and financial advisors in 2026. Five use cases that work, four that don't, the SEC and FINRA examination focus, the Apoorva Mehta autonomous-agent hedge fund as a frame for the asymmetry traditional firms face, and a build-vs-buy framework for a 5-to-25 advisor firm.
Claude Cowork vs Claude Code: An Honest Practitioner's Build-vs-Buy Decision (2026)
Claude Cowork is the version of Claude that runs your work for you. Claude Code is the version that lets you build software. Here's the honest framework — with real numbers from an agency running both, written so a non-developer can actually use it.
AI receptionist for small business in 2026: off-the-shelf vs custom (vs keeping your human)
Three paths, not two. Most small businesses should buy off-the-shelf. Some shouldn't use AI at all and should keep their human receptionist with workflow automation. Six conditions justify a fully custom build. The honest decision framework, with real costs and real pricing.
AI for Accountants: What Actually Works in 2026 (Implementer's Field Report)
"AI for accountants" in 2026 is not one product — it's five distinct categories doing five different jobs. Tooling for a 10-person firm runs $300-$1,500 per month. A custom agent build runs $15K-$80K and 60-90 days. The bigger story isn't tools — it's what AI is doing to the billable-hour pricing model. An implementation services firm's honest field report.
Stop hiding your pricing
If your ideal client can't afford you, you want them to know that before the call, not during it.
We rebuilt a law firm's entire intake in 3 weeks
A build log: decisions, dead ends, and the system that emerged.
5 real challenges small businesses hit with automation (2026)
The five problems that actually break small-business automation in 2026 — tool sprawl, no system of record, automating chaos, AI review overhead, and integration brittleness. Honest field notes from builders.
How we build: the Automaton stack (and why it's different from an "AI agent stack")
The exact tools, layers, and trade-offs behind how one creative technology agency ships client work in 2026 — including the client portal we built on our own stack and where Claude Managed Agents fit next. Not an AI agent tutorial.
What a creative technology agency actually does
A creative technology agency builds intelligent systems — AI agents, automated content, integrated revenue operations — where creative, technical, and strategic thinking happen simultaneously. Here's what the work actually looks like from inside one.
Why your AI chatbot sounds like everyone else's
The problem isn't the model — it's that nobody bothered to give it a real voice.
The creative technologist is the new agency
The traditional agency pipeline is collapsing. One person with taste, judgment, and AI tools now delivers what used to require a team of fifteen.
Claude as a CMS: how we manage websites through conversation
We replaced WordPress with Claude. No dashboard, no admin panel — just conversation. And the teams using it aren't developers.