Workshops

We train founders and GTM teams on Claude Code.

Hands-on build workshops: bring one real workflow, build on your own data, and leave with a working system you run yourself.

Book a Workshop

3 hours from workflow map to working system

4 hours saved per project, in one builder's own words

22 steps the biggest manual workflow a builder brought in

The method

How a session runs.

Every session runs against one metric: time to first win.

Map the real workflow

Before anyone walks in, each builder documents their actual process: every step, every tool, hours per step. One founder's map came back 22 steps across seven tools. Another timed his bottleneck at 1-4 hours per project. That map is the raw material for the day.

Scope to the step that pays

A session that tries to automate a 13-step pipeline ships nothing. We find the single step that eats the most hours or blocks the most revenue, and the whole day is about that step.

Build on real data

Real exports, real past emails as voice samples, real prospects. When data can't leave the company, we mirror it with a synthetic dataset that matches the real export field for field - so the system transfers straight onto the real thing.

Ship in ranked order

Every build starts with a must-ship, build, and stretch list. The decision rubric ships even if everything else dies. Polish comes last, not first.

Leave running it

The build leaves the room as a skill plus a runbook the builder runs Monday morning without us. And nothing sends itself - every output stops for the owner's review and approval.

Case study

From first win to a motion built from scratch.

The Dealer Network runs Ontario's peer-to-peer sourcing network for independent used-car dealers. They're standing up their first GTM motion from scratch - no lead list, no campaign history - on Headless GTM, our open-source Claude chain.

No list. No campaign history. A first campaign, running.

Instead of buying a database, Claude starts from a question: where are these buyers already listed?

01 Source of record Ontario's public used-car dealer registry. Not a purchased database.
02 Qualify Every dealer against the ICP, before a dollar of enrichment is spent.
03 Rank Who to contact first, and the reason why.
04 Draft Outreach written from the signals, in the founder's voice.
05 Owner approves Nothing sends without a human sign-off.

That is the same system the workshops teach: discovery from sources of record, a qualification gate before any paid step, signals with provenance, and an owner approval on everything that ships. Told with The Dealer Network's permission.

From the room

Builders from the room.

The GTM & outbound track we ran at AI Labs Toronto: three hours, one build each. What they walked in with, and what they left running.

A four-person conference media agency

Walked in

A 13-step content pipeline and the step blocking it: 1-4 hours of manual footage review per project, standing between the team and a new content revenue line. Self-described "not extremely technical" - ChatGPT was the whole stack.

The build

A review system that turns long-form recordings into a structured content map - themes, candidate pieces, angles per platform - that the team can act on without re-watching footage.

“Incredible experience! I built something in Claude that saves about 4 hours of work per project... because of the output, I can now analyze my work on a deeper level that pretty much equips me with the firepower I need to identify & pitch future sales opportunities.”
Javon Martin, CEO

Enterprise B2B lifecycle marketing

Walked in

Nine years in B2B marketing and a weekly grind: 30-80 at-risk accounts flagged by a health-score system every week, 4-6 hours each Monday triaging them by hand - and still never covering the full list.

The build

A two-skill triage system. A categorizer applies their decision rubric to the raw export and sorts every account into a risk type with a recommended intervention. A writer drafts the re-engagement email in their voice, trained on their own past sends, with every email stopping for their review. Built against a synthetic dataset matching the real export field for field, so it transfers straight onto the real thing.

A marketing consultancy

Walked in

20-25 hours a week of manual prospecting: scanning job boards for companies hiring marketers, checking ICP fit, hunting decision-makers, writing personalized outreach. 22 steps across seven tools, all by hand.

The build

Her first AI agent, blueprinted against a real use case: an outbound engine built on her best buying signal - active hiring. Companies hiring for marketing roles in; qualified companies with decision-makers, context, and drafted outreach out, logged where she already works.

“As a marketer and non-engineer, I walked in intimidated. I walked out realizing the barrier to building tech is shrinking faster than ever... Watching my idea go from concept note to execution in real time felt really exhilarating and empowering.”
Prachi Salvi, Director

An AI job board, built off a 30,000-follower audience

Walked in

A job-search audience built in seven months, and the problem it hits every day: good roles buried across job boards, found too late.

The build

A hidden-jobs system: pulls hard-to-find roles from multiple sites, filters them to her exact criteria, keeps only postings from the last 24 hours, and delivers them daily. She kept building it after the session for her job-seeker audience.

A web design studio for professional-services firms

Walked in

A studio whose best buyers - small law and accounting firms - can't be found in standard B2B databases. The tell that a firm needs the studio is the state of its website, and no database filters on that.

The build

A website qualifier: point it at a firm's site and it returns a verdict and the outreach hook - the first skill of a four-skill outbound pipeline that finds firms, qualifies each by its website, finds the owner's verified email, and writes the sequence.

“Learning with Yusuf for just a few hours opened my mind to literally a world of possibilities of implementing AI into my business to super charge it. It totally feels like a proper cheat code.”
Suhail Akhtar, Founder

Three ways teams run this.

One session

Build lab

One scoped build, first working system. A founder or operator takes their highest-leverage workflow from map to working system in about three hours. The format behind the builders above.

Weekly sessions

Build-alongside

Weekly working sessions for a founder or small team standing up a motion from scratch. We lead the architecture and the hard parts; you build alongside and own the output - every skill documented so the next one can be built without us.

4-8 weeks

Team enablement

The full program: embedded with your GTM team for 4-8 weeks, two tracks (Cowork for most of the team, Claude Code for the builders), and an independence test at the end - each person ships a new workflow with us out of the room. Applied AI for GTM teams →

The stack we teach on is open source.

Headless GTM is our outbound chain, published as Claude Code skills: company discovery, a free qualification gate, website research, hiring and funding signals, judgment with provenance, verified email resolution. Cost gates and approval steps are built into every paid step.

Workshop builders leave running the same chain we run for clients. No black box, nothing to license, nothing that breaks when we leave the room.

Try it yourself →

Bring the workflow that eats your week.

One session. Your real data. You leave running the system.

Book a Workshop