One department.
Half a day.
Agents that keep working.
Most teams already use AI to write. Very few use it to work. We sit with one department, on its real data and its real weekly tasks, and by the end of the day the team has built two or three agents it runs itself. No slides. No demo data. Hands on the keyboard.
Departments that already use AI, and are ready to build with it
The best candidates are teams with a clear repeating task and data that lives somewhere reachable. A few we run this with:
R&D and data teams
Agronomists, analysts, researchers. People who sit on a database full of experiments and still export to Excel to ask it a question.
Operations
Inventory, shipments, vendors, Monday boards and spreadsheets. The department where the most manual hours hide.
Marketing and sales
Content, CRM, call summaries, follow-ups. Teams that already use AI to write, and are ready to have it work.
Finance and admin
Invoices, reconciliations, reports that get rebuilt every month by hand. Repetitive, rule-based, perfect for a first agent.
Three steps. One scoping call, one build day, one week of follow-up.
Scoping call with the department lead
What the day-to-day looks like, which tools are already in use, where the data lives, and the three to five tasks that repeat and eat the most time. You get back a precise agenda, a shortlist of use cases, and a technical prep list (data access, accounts, laptops).
Build with the department
Four blocks: what is possible (45 min, real examples from other companies, not a slide deck) · choose two or three use cases from your actual week (30 min) · hands on the keyboard, everyone on their own machine, connecting Claude to your data and building the tools (2 to 2.5 hours) · wrap-up: what works, what still needs engineering, who owns each tool, and what comes next (30 min).
Follow-up until the tools settle
A written summary of every skill and agent that was built, with its definitions and our recommendation for the next step. Questions by email or WhatsApp for a week after, until the tools are stable in daily use.
Tools, a map, and people who know how to build the next one
Six rules every workshop follows
Hands on the keyboard
Nobody watches a demo. Every participant builds on their own laptop, on their own real task.
Your data, not sample data
We connect to the actual database, the actual sheets, the actual inbox. The tools work on Monday morning because they were built on Monday's data.
One department at a time
Company-wide AI talks are nice. Results come from sitting with one team on its specific pain points.
Nobody is forced into the terminal
Some people want to build. Some want a tool that just works. Both get what they need: the builders build, and we build for the rest.
Claude Code, Claude Desktop, MCP
The tools we use every day in production. We teach the difference between a chat, a skill, an agent and a connector, and when each one is the right call.
Humans where judgment is needed
Every agent we build has a person approving what matters. Speed from the agents, decisions from your team.
Start with one department. Grow from there.
Department day
The core format. One scoping call, one half-day build session with one department, one week of follow-up.
Multi-department series
The same process, department by department, over a few weeks. Each team gets its own day and its own tools; the company gets a map of where AI actually moves the bottom line.
Ongoing accompaniment
After the day: one-on-one sessions on a specific use case, tools built for people who prefer not to build, maintenance when something breaks, and extending to the next team.
Before you write to us
Do participants need to know how to code?
No. We work in Claude Code and Claude Desktop with plain language. Someone who can describe their work clearly can build a skill by the end of the day. The technical setup is handled in the prep list before the day.
How many people per workshop?
One department. Small enough that everyone builds and gets feedback on their own tool during the day.
Which tools do you use?
Claude (Claude Code, Claude Desktop, MCP connectors) on top of your existing systems: databases, Google Workspace, Monday, CRM, spreadsheets. If a tool has an API, we can connect to it.
What about our data?
Everything runs inside your accounts and your infrastructure. Nothing is copied to us. The prep call covers access, permissions and what must stay out of scope.
What happens if we want a full system built afterwards?
The workshop is where those needs surface. Anything that turns out to need real engineering becomes a separate, scoped project, priced on its own after we know exactly what it is.
Hebrew or English?
Both. Workshops run in Hebrew for Israeli teams and in English for international teams.
Which team should go first?
Send us a WhatsApp with the department and what it does all day. We will come back with a first read on where an agent pays off, and we take it from there.
Talk to us on WhatsApp