Free guide · 2026 edition · no sign-up
Most small teams don’t have an AI problem. They have a guessing problem. This free 25-page guide shows you how to start with one workflow, measure it, choose the least autonomy that works and run a short pilot before you buy anything else.
Free and ungated. No email needed. Written for owners and managers of small, often remote teams.
What is in the guide?
The guide keeps the ten practical steps of the original metalaxis guide and adds what changed with bots and AI agents: what information a system can read, what it is allowed to do, who approves it and how you measure the result.
- Know your goldmine. Find the revenue streams that matter before you automate anything.
- Map your operations. Draw the real workflow, including owners, inputs and what happens when a tool fails.
- Run a time audit. Measure net human effort per accepted result, including checking and rework.
- Find opportunities. Sort tasks into drop, delegate, automate and do.
- Prioritise. Treat risk as a release gate, not an afterthought.
- Plan in stages. A demo is not a deployment: discovery, prototype, evaluation, pilot, release.
- Budget for operation. Count review time and running costs. Capacity is not cash.
- Build buy-in. Name who owns the result when the AI output is wrong.
- Prepare your team. Agree data rules, review and when to stop and ask a person.
- Scale evidence. Retest when models, prompts or connected tools change.
Plus: bots and agents explained, a safe-actions checklist, four illustrative workflows, a fill-in pilot charter, a budget-and-measurement page, a 30-day plan and a glossary.
Read, draft, approve, act: the rule behind the guide
Read → Draft → Approve → Act. Start with AI that reads and prepares. Let it draft. Approve the exact action (the real recipient and the full message, or the actual record and change). Only then let it act, and only within limits you can undo. A draft and a sent message are different business events.
Who is it for?
- Owners and managers of small businesses and teams of roughly 5–50 people, many working remotely.
- Teams that already tried a few AI tools and can’t say what they saved.
- Anyone asked to “do something with AI” who wants a plan they can defend.
Want a second pair of eyes?
Start with the free AI readiness check. If you want a second opinion on your result, bring one mapped workflow and a baseline.
Want a second opinion? Email [email protected]
Related reading
- Which tasks should you delegate to AI? A 4-way sort for small businesses
- How to use AI in your business without the chaos: a 5-step roadmap
- How we work: the delegate-first method
- Free tools: revenue focus and delegation sort
Frequently asked questions
Is the guide really free? Do I need to give my email?
Yes, it is free and there is no form in front of it. Click the download button and the PDF opens. If you would like a personal read-out, the separate AI readiness check gives you one, and you can email [email protected] for a second opinion. The guide itself stays open to everyone.
How do I start using AI in a small business?
Pick one workflow that matters to revenue or service, such as proposals or support replies. Record how long it takes and how often it needs fixing. Try the simplest fix first, often a rule or template. Use AI for one bounded step, keep a person approving anything that leaves the business, and compare the results with your baseline after about 30 days.
What is the difference between a chatbot, a workflow and an AI agent?
A chatbot answers or drafts. A workflow follows fixed steps you define. An AI agent chooses tools and next steps toward a goal. Each needs different controls. The guide recommends using the least autonomy the task needs and starting read-only where you can, then expanding authority only after testing.
How long does a first AI pilot take?
The guide suggests an illustrative 30-day sequence: understand the workflow (days 1–5), prepare sources and test cases (6–10), build and evaluate (11–20), then pilot and decide (21–30). It is a planning aid, not a guarantee. Extend testing when integrations, risk or information quality need it.
How do I measure whether AI saves time?
Measure net human time per accepted result: the time to brief, review, correct and approve, not just how fast a draft appears. Track errors and exceptions too. Saved hours only become cash savings when spending actually falls, so record released capacity and real financial benefit separately.