The agenda, hour by hour.

• Hour 1 · AI foundations: what Claude can and can't do today, where it fits your work, and safe, responsible use under your firm's policy
• Hour 2 · Prompting that works: context and instruction patterns for reliable, high-quality output, practised on your own documents
• Hour 3 · Claude Projects: you set up your own, with your firm's templates, SOPs and tone, so every conversation already knows your organisation
• Hour 4 · Claude Cowork: hand Claude a folder and a task and it reads, drafts and edits your real files end to end while you do something else
• Hour 5 · Your real work: each person applies it to their actual daily tasks, set up on their own device, with Soch at their shoulder
• Hour 6 · Live demo, Q&A and what to build first: a run on your firm's real work, every "would it work for…?" answered, and a plan for the next morning

What people use it for from day one:

Working together as one organisation, your team picks the highest-value, most repetitive work across the business, then Soch coaches everyone to build real AI agents for it, hands-on over live video. Soch coaches; your people do the building, so everything is theirs to keep and extend.

• Hour 1 · Opportunity map: list where AI saves the most time across the business, then pick the 3–5 shared jobs worth building
• Hour 2 · Agent or workflow? Design each one: decide the right shape, map each to Project, Skill, Cowork and connectors, and define its goal, guardrails and human checkpoints
• Hour 3 · Build shared Projects & knowledge: the "what your agent knows" layer, loaded with your organisation's context, templates and knowledge base
• Hour 4 · Build reusable Skills: capture your output formats, tone and standard operating procedures as Skills, so quality is consistent for everyone
• Hour 5 · Build your first agents: assemble Cowork + Skill + Project + connectors into agents that run on your real files, then steer and verify
• Hour 6 · Connect & test end to end: wire connectors and MCP into the tools your firm already runs, Google Workspace, Slack, your CRM, and run everything on real work

• Hour 1 · Harden the build: edge cases, checks and a "check before you send" step baked into each agent, then stress-test it so it holds up
• Hour 2 · Advanced agent patterns: where the work earns it, add subagents, scheduled autonomous agents, or the Claude Agent SDK for your technical staff
• Hour 3 · Governance & safe use: what must not be automated, data handling, approval gates, and a one-page AI-use policy the team signs off
• Hour 4 · Package & document: assemble the handover pack, every prompt, Project, Skill and agent, written so anyone can run it
• Hour 5 · Demo hour: your team demos what it built to the whole firm; every agent gets a named internal owner, accountability not enthusiasm, and one simple success metric
• Hour 6 · Rollout & enablement: a 30/60/90-day plan to roll the agents out to the rest of the firm, onboard new joiners, and keep building without us