Turn real work into agent-powered operating systems.
Every team has workflows, judgment, context, shortcuts, exceptions, and institutional memory. Alchemy is the premise: turn that raw material into visible, reusable, agent-usable systems.
Two pieces, designed to compound.
Most companies buy AI tools before they understand how work should change. The better sequence is training plus infrastructure: map the workflow, teach the operating model, then wire the systems.
I teach teams how to work with agents
Training and operating playbooks for turning real workflows into project folders, durable context, memory, compaction, skills, subagents, review loops, and artifacts.
- 01Map undocumented process
- 02Hands-on agent fluency
- 03Role-specific playbooks for how work actually gets done
I build workflow infrastructure
Implementation for the data, tool, and document layers that make agent work useful inside a serious company: MCP servers, internal agents, workflow automation, and deck/report factories.
- 01Agent-accessible data layers
- 02MCP and tool-call surfaces
- 03Repeatable workflow, memo, and deck systems
AI adoption is an operating problem, not a software-license problem.
Work is scattered across Excel, email, slide decks, dashboards, meetings, shared drives, and people's heads. Agents only become useful when context, tools, review, and ownership are designed on purpose.
Name the work
Map the real process: decisions, shortcuts, handoffs, exceptions, review criteria, and the judgment that usually stays undocumented.
Teach the interface
Train the team to brief agents, preserve context, supervise work, and avoid turning AI into unmanaged chat.
Wire the substrate
Expose approved data, files, and internal workflows through typed tools, clear review paths, and durable artifacts.
Ship the loop
Turn the highest-value workflows into reusable project memory, skills, subagents, artifacts, and operating rhythms.
Why me?
Because I am using this stuff for real work, not teaching it from the sidelines. I can show teams the behavior, then help build the parts that need to exist.
Operator, not theorist
I use these tools every day to build companies, websites, decks, workflows, and internal operating systems.
Teacher who can build
The work starts with education. When a team needs infrastructure, I can help create the practical tools that make the behavior stick.
Plain-language adoption
This is not prompt theater or AI hype. It is a clearer way for teams to capture how work gets done, supervise agents, and improve the workflow over time.
For regulated or sensitive teams, the implementation starts with boundaries: approved data only, firm-approved tools, no confidential material in unmanaged systems, reviewable outputs, and clear ownership of artifacts.
The training sequence has three moves.
The point is not to teach abstract AI concepts. It is to move a team from the why, to technical confidence, to a repeatable operating model they can use on real work.
The Work Surface Has Changed
Why work is moving from app-centered workflows to project folders, agents, memory, artifacts, and review loops.
Technical Basics for Agentic Work
Files, folders, paths, terminal, Claude Code, Codex, and the confidence to open an agent inside real work.
Memory, Skills, and Subagents
How context, memory, compaction, reusable skills, specialist lanes, and artifacts turn a folder into a project system.
The deliverable is a working project system.
Training creates the shared language. Implementation turns one real workflow into a project folder the team can reuse: context, instructions, sources, tools, skills, subagents, review loops, and artifacts.
Workflow map
Inputs, steps, decisions, exceptions, handoffs, review criteria, and outputs.
Project memory
Folder structure, instructions, source docs, examples, templates, and durable context.
Reusable methods
Skills and subagents for repeated work that should not be reinvented every session.
Artifacts
Dashboards, memos, reports, decks, QA checks, and reviewable outputs the team can keep using.