A playbook on the four barriers that stop organisations scaling AI. The first is compliance paralysis, solved with a “two zones” framework that separates safe testing from restricted use, plus a readiness checklist. The second is shadow AI, solved by making the approved route the easiest one to use. The third is capability lockout, where the playbook shows that organisations needn’t choose between quality and control: frontier models suit early exploration, and smaller models often deliver the same results once solutions mature. The fourth is invisible AI spend, solved through visibility of usage, cost and outcomes, plus a simple “ROI lite” approach. Each chapter ends with practical checklists and a key takeaway.