Basics Find where AI pays in your business, and bank your first measured win
- The opportunity mapSpreadsheet XLSX
chapter 2
Score each task in your business by time, frequency and AI-fit, with risk as a gate, and it flags your first bets — so you back the two or three places AI actually pays, not the flashiest demo.
- The one-page data ruleDownload TXT
chapter 5
What your team may put in, what they never put in, which account to use — the page everyone signs, so AI never becomes the thing that costs you a fine or a customer. Plain-English, not legal advice.
- The "what good looks like" standardDownload TXT
chapter 6
Five headings that make AI sound like your business, not a generic stranger — plus the five-question review gate, so nothing off-brand reaches a customer, whoever on the team is driving.
- The guardrailDownload TXT
chapter 7
AI can be confidently, fluently wrong. The four-line rule that catches a made-up price or date before it goes out: trigger, owner, a real source to check against, and a hard stop.
- The keep/kill saving calculatorSpreadsheet XLSX
chapter 8
The honest maths: type your old and new times — with the checking counted in — and it shows the real saving a week and a year, against a keep/kill line you set first. Then it tells you: bank it, or bin it.
Intermediate Put AI across your operation — run by your team, governed, and proven to run without you
- The rollout planSpreadsheet XLSX
chapter 1
Rank the functions of your business by what AI would actually return — the hours a function eats, times how well AI fits it — then fill the adoption contract for the first one (operator, approver, the target, the review date) behind a launch-readiness gate that opens only when you're truly ready.
- The operating brainDownload TXT
chapter 2
The one shared source of truth the whole team feeds the AI from — who you are, who you serve, how you sound, what you never do, the facts that must be right — plus the provenance rule that makes using it non-negotiable, so the work sounds like one company whoever's driving.
- The function deployment kitDownload TXT
chapters 3–7
One repeatable template to stand up any function on AI so it runs to standard, operated by your team, without you: the loop, the publish gate, the never-decide line and escalation, the customer-service authority cap and severity handoff, the back-office definition-of-done and sampling rule, and the conditional sunset that makes it stick.
- The monthly management packSpreadsheet XLSX
chapter 8
The picture your team prepares to a fixed spec — six things, month on month — so you arrive to decide, not to assemble. Plus the decision register: every call (or a justified "no change") logged with a baseline, an owner and a review date, carried until it's closed.
- The govern & prove kitDownload TXT
chapters 9–11
The layer above every function: a one-page governance policy, a kill switch with a named holder, the cascade drill, the weekly operating cadence, and the owner-absent verification that proves — on evidence, not a hopeful "all good" — that the business runs without you in the room.
Advanced Turn AI from a copyable cost-saver into an advantage rivals can’t easily copy
- The edge selection gateSpreadsheet XLSX
chapter 1
Four candidate edges in, one out. Stage A asks whether each is a real, durable edge at all — the mechanism in a sentence, uniqueness, copy cost in months, permissions, and the ownership gate a named person has to accept out loud. Stage B asks which one can show something observable in ninety days. The gates fail closed: a blank reads as out, never as through. Plus the pre-registered baseline, signed by someone with no stake in the answer.
- The governed outcome recordDownload TXT
chapter 2
Your archive records what happened. This records what worked — the one thing a competitor can’t buy, because the only way to produce it is to have done the work. Name the asset, define “worked” so a second person can check it, capture at two moments the business already has, four fields, four roles on four desks you already have, and the boundary that says what you may and may not record.
- The decision-rule cardDownload TXT
chapter 3
Take one recurring decision out of your best person’s head and put it on a page where it gets sharper every quarter. Trigger, default, who may override and on what grounds, and the citation rule that stops a confident machine inventing reasons: no cited case from your own record, no recommendation. Plus the override log — because overrides aren’t failures, they’re the most valuable entries you have.
- Pilot, price and commitmentSpreadsheet XLSX
chapters 4, 5 & 10
The priced pilot with a three-clause stop rule (delivery failure, service burden, rule reliability), then the cohort test that measures what it’s really worth in realised margin per delivery hour — the only honest unit at this scale — against a four-branch decision rule you write before any result exists. Ends with the commitment page and the five-condition exit proof.
- The durability auditSpreadsheet XLSX
chapters 8, 9 & 11
One question, four gates: if your main tool vanished tomorrow, could you keep the promise you published? Supplier dependency, reuse rights, the fallback, and a timed deskilling drill — thresholds written first, scored by two people separately, and anything you can’t evidence scores zero. Plus the disposition table that decides kill, narrow, replace or de-risk before you look, and the quarterly review with the copy-cost log.
Kept current AI tool prices, checked by hand
The fastest-changing figure the Basics book touches — what the main AI tools cost. Checked by hand and dated. The book's method doesn't move; these numbers do.
⚠️ Prices change monthly and vary by country and plan. These are a shape to expect, not a quote — always check the maker's own current pricing page before you pay, never a comparison article.
- Individual paid plan (ChatGPT, Claude or Gemini) — roughly £16–20 a month per person (Google's is about £19). Any of the three will do your first bets well.
- Team / business plan — priced per seat, from about £20–25, sometimes with a small seat minimum. This is the version to use for real customer or business data: business tiers typically don't train on your data by default and can come with a data-processing agreement — but check the maker's own terms, and the admin usually has to accept them in the settings.
- Free tier — every one has a free version. Fine to try, and often enough to run your first test; upgrade to a paid seat only when the free limits get in your way.
Checked at source 2026-10-12; next review due Jan 2027. Dated copy, not a live feed — confirm at each maker's own current pricing page before you rely on a figure. Nothing here is legal advice.
Free calculators Live tools you can use right now — no download
- AI ROI calculatorCalculator Live
Payback period, first-year net and return on an AI or automation project — before you commit the budget.
- AI time savings calculatorCalculator Live
The hours and money a repetitive task reclaims per year once AI handles most of it.
- True cost of an employee calculatorCalculator Live
The fully-loaded annual, monthly and hourly cost of a hire — the other half of the hire-or-automate question.
- Which task to automate first — scorerScorer Live
Score any business task on frequency, time, repetitiveness and error-cost to see what to automate first.
Take them free — that's what they're for. When you want the argument for using them, AI for Your Business — Basics takes a non-technical owner from "is AI even real for a business like mine?" to a single, measured, banked saving on one real job — safely, and sounding like your business, not a robot. Intermediate and Advanced follow, when one proven win needs to become how the whole business runs.
See the AI for Your Business series →