Most teams rent AI and re-explain themselves every morning. In this hands-on workshop, your people build a personal AI workspace that knows them, does their real work, and compounds week over week - and they leave with it running, not with a folder of notes.
Most people use AI like a vending machine: one prompt in, one answer out, starting cold every time. The fix was never a better prompt - it's a better workspace.
Most software is best on day one and decays. This is the opposite: a little wobbly on day one, noticeably stronger every week as it learns you and takes on more of your work. The threshold to solve a problem with software just hit an all-time low, and a workspace that compounds is how a team does 10x what it could before. The value was never the setup - it's the slope.
The setup stands on three legs. Take one away and it tips over - you're back to a chatbot with amnesia. Everyone leaves having built all three, then pointed them at their own real work.
Delivered as two half-days (ideally a week apart, so the setup gets real use in between) or as one full day back-to-back. The taught material is the same either way - the majority of it hands-on.
| Time | Segment | What we do |
|---|---|---|
| 0:00-0:30 | The shift | Why most AI use plateaus, and the model that compounds instead. |
| 0:30-1:30 | Leg 1 · Brain | Stand up your AI workspace and give it an identity.Hands-on |
| 1:30-1:40 | Break | Any setup snags get fixed here, not mid-build. |
| 1:40-2:55 | Leg 2 · Agents | Build your first custom agent - a skill plus one real tool.Hands-on |
| 2:55-3:55 | Leg 3 · Context | Give that same agent persistent knowledge - and see it improve.Hands-on |
| 3:55-4:15 | First run | Point your assembled setup at a real task, end-to-end.Hands-on |
Between sessions: use it on your actual work and note what breaks - that friction becomes the raw material for Session 2. (On a single full day, this is the lunch break.)
| Time | Segment | What we do |
|---|---|---|
| 0:00-0:25 | Debrief | What worked, what broke - we turn real friction into fixes.Hands-on |
| 0:25-1:05 | A day in the life | Worked examples across real recurring tasks - inbox, meetings, reports, drafts. |
| 1:05-2:15 | Build clinic I | Build an agent for a recurring task you brought.Hands-on |
| 2:15-2:25 | Break | |
| 2:25-3:35 | Build clinic II | A multi-step job - connect a tool and chain all three legs.Hands-on |
| 3:35-3:55 | Graft, don't adopt | Staying ahead without chasing releases, the value-vs-risk lens, and keeping it maintained. |
| 3:55-4:10 | Share-back | Live demos, the team-multiplier effect, and Q&A. |
| Format | Hands-on build, on-site or remote. Mostly doing, not listening. |
| Length | One full day, or two half-days a week apart (recommended - the setup gets real use in between). |
| Group | 6-10 people, working in pairs. Mixed ability welcome; this is not a coding course. |
| Stack | Claude Code plus whatever tools your team already uses, tailored to your stack. |
| Prep | A short checklist two days before - about 15 minutes to install a few tools and bring some real work. |
| Language | English (Hebrew on request). |
| Pricing | On request - depends on format and group size. Get in touch. |
Lior Meshullam is the founder of Revgineer and a Head of Product. He has spent the past year living inside an organically evolving AI setup that now runs a portfolio of live products - and consistently finds himself ahead of the tooling roadmap. He builds in the open, and this workshop hands your team the exact way of working: the disposition plus a seed, not a frozen toolkit.
Tell me a bit about your team and what you'd want out of it, and we'll find a format that fits.
Host this workshop