The Working AI Playbook: Everything an AI Agency Knows, Published Free

Most of what gets sold to small businesses as "AI transformation" is an afternoon of setup behind a markup. Not all of it — some of this work is genuinely hard, and I'll show you exactly which parts. But the gap between what AI services cost and what they take to build has become the most profitable confusion in small-business software, and nobody selling to you has a reason to clear it up.

So I'm publishing the whole thing. The Working AI Playbook is everything an AI agency knows about putting AI into a small business — free, with receipts, no email gate. Not a lead magnet with the good parts held back. The actual playbook.

Why give it away

Three honest reasons.

First, the information wants to be free anyway. Every technique in this playbook is learnable from documentation, forums, and trial and error. What I'm selling was never secrecy — it's the hundreds of hours you don't have to spend, and coverage for the parts that genuinely can't be DIY'd. Pretending otherwise is the hype tax, and I'd rather compete without it.

Second, some of you will read a chapter, do it yourself in a weekend, and never pay me a dollar. Good. That was the point of the chapter. You'll also know exactly who to call when you hit the parts that break — and you'll trust that person, because they told you the truth when it was free.

Third, I want the standard to move. If a $99/month "AI package" is something this playbook teaches in twenty minutes, that price should not survive. The businesses in my town deserve to know the difference.

How it's organized

Three levels, so you can enter wherever you actually are:

• Level 1 — Use AI today. Copy-paste prompts for the admin that eats your week: review replies, estimate follow-ups, no-show reminders, invoice chasers, job posts. Nothing to install. Works in the free tiers of ChatGPT or Claude.
• Level 2 — Set up systems. The working-afternoon projects: measuring what missed calls cost you, choosing between an answering service, an AI receptionist, or a hire — with real published prices — and checking what AI search engines say about your business.
• Level 3 — Build it yourself. For owners who like to tinker: running local AI models on a $600 Mac mini, and building your first real business automation with Claude Code. Real tools, real costs, weekend-sized.

What's live today

Two chapters, both shipped with working artifacts rather than advice:

What missed calls actually cost your business. A 30-second calculator that runs your numbers — not a vendor's slide. A plumber missing five calls a week at a $400 average job and a 35% booking rate is losing about $3,031 a month; the chapter shows the formula so you can check the math, then compares the three fixes like-for-like. It also says plainly when the right answer is do nothing — if you miss two calls a month, no product pays for itself.

The daily-ops prompt pack. Ten complete prompts with the guardrails each one needs. The negative-review prompt includes the line most owners have never been told: if you run a healthcare practice, confirming that a reviewer is a patient in your public reply is a HIPAA violation — practices have paid five-figure settlements for exactly that sentence. The guardrails are half the value of the pack.

The rules this playbook follows

• Receipts on everything. Every number comes from a stated formula, published pricing, or my own data. When a chapter uses my numbers, you'll see the actual data.
• No email gate, ever. There's an optional list if you want new chapters by email. The content never hides behind it.
• Where DIY breaks, the chapter says so. After-hours phone coverage that can actually book appointments, healthcare and legal compliance, deep integrations — every chapter is explicit about which walls are real. That honesty is also, frankly, the business model: the hard parts are what we sell.
• Maintained, not archived. Chapters are living pages, refreshed as prices and tools change — not dated posts left to rot.

What's coming

In progress, in roughly this order: the 2026 AI receptionist pricing index (real numbers across vendors, including the per-minute fee traps), real call audio and transcripts so you can judge an AI receptionist with your own ears, the 20-minute AI-search checkup, the where-AI-is-not-safe chapter for healthcare and legal, the $600 Mac mini local AI build, and a receipts chapter auditing my own 92 blog posts — including the 50 that turned out to be invisible in search. The failures go in the playbook too.

Who's writing this

I run WildRun AI solo from Bend, Oregon, and until recently I drove a delivery truck full-time — every system in this playbook exists because I needed my business to run while I was on a route. The content engine that publishes here, the analytics loop that grades it, and the AI receptionist that answers for it are the same systems the playbook teaches. When a chapter says "this works," it means it's running right now, unattended, on my own business.

Start with the calculator if your phone is the problem, or the prompt pack if your evenings are. The rest of the map is at wildrunai.com/playbook.