[Services / 02.2]
Put it to work.AI training and custom workflows for creative agencies, on the tools your team already pays for.
◆ machine · generated · 17 tokens · 1380 ms
I train agency teams on tools like Claude Code, then build the custom skills and workflows that make those tools fit how the agency actually works. The first example is always your own material.
- Sessions
- Hands-on
- Material
- Your own
- Review
- Human in the loop
- Handover
- Guides + owner
- 01
Training
Hands-on sessions on the tools your team already has, starting with Claude Code. Your projects, your files, not a slide deck about the future.
- 02
Custom workflows
Skills, agents and MCP connections built around how your team works: your templates, your tone, your data.
- 03
Consulting
Where AI is worth it in your agency, where it isn't, and which tools to trust with client work.
Image SEO that runs every day
- Before
- Every image on a manufacturer's WordPress site needed research, a new file name, alt text and a caption. By hand, one at a time.
- Workflow
- A framework I built on Claude Code, with skills, agents and subagents. It reads Search Console through MCP, checks the keyword niche, looks at the image and at the page it sits on, then writes the name, the alt text and the caption. A person reviews and confirms before anything goes live.
- After
- Hundreds of images a day. Nobody renames them by hand any more.
Start small, prove it on real work, then scale what pays back.
- A
Start with a session
A hands-on workshop on your team's own projects. We find the jobs worth handing to AI, and the ones that aren't.
- B
Build what's missing
Skills, prompts and tool connections for the jobs that came up in the session.
- C
Leave it owned
Written guides and a named person inside the agency who looks after it.
- Claude Code
- Codex
- Skills & subagents
- MCP
- Search Console
- WordPress
Do we need a technical team?
No. One curious person is enough to start, and training begins with the tools you already pay for.
What happens in the first week?
Once people see how the harness, skills and MCP fit together, they want to do everything with AI, and forget permissions, tokens and where the model stops being smart. A good part of the first week is learning where to stop.
Do you still build n8n automations?
When monitoring matters, yes: n8n still gives you a place to watch every run, which a CLI agent doesn't. Most agency teams are better served by skills inside the tools they already use.
What about client data?
Where the models run is decided with you, based on the data involved. Every workflow keeps a person in the loop for reviews and confirmations.
Can you work white-label for our clients?
When the project needs it, yes. I also build AI features that agencies offer their own clients: chatbots and semantic search grounded in the client's content.