Session 1, Context and research: Your private workspace and a durable context file on your businesses and priorities (40 min); Deep research (ChatGPT deep research, Claude Research) on a live question of yours, with the method for checking every claim against its source, saved as a reusable agent skill (80 min); A standing market and competitor brief for your sector, refreshed on one instruction (60 min). Between sessions: You run the research skill on two live questions. Session 2, Documents, connections, multimodal: Reports, contracts, tenders and management reports read and cross-checked, then organised into a knowledge base (60 min); Connecting the AI agent to your mail, calendar and files, with read access by default and your approval before it sends, edits or deletes anything, plus the security settings that keep company data private (60 min); Voice notes, scans, photos and drawings, applied to your sector (60 min). Between sessions: You build your first knowledge base from your own material. Session 3, Agents, automation, decisions: Three or four of your recurring routines turned into agent skills, including drafts in your own voice (60 min); Automation with approval points: inbox triage, follow-ups and a daily brief that run on a schedule, with the AI agent asking before it sends, books or changes anything (45 min); Stress-testing a real decision with the AI agent (45 min); A cost map of your recurring work ranked by saving, the plan for what comes next, and a short handover so your office can support what was built (30 min). The arc is a template, customised for you.