Strategy, content, creative production, lead research, outreach, prompt design and reusable skills were trying to share one doorway at Notorious MBA and ExitGuru. In one undifferentiated chat, Kevin Hucker would remain both dispatcher and institutional memory.
Evidence at publication: six agents and 36 bindings were observed on 30 May 2026, with dedicated German Discord routes, gbrain memory, local voice transcription and Kie.ai tooling configured in the deployment. The source retains no attributed commercial result and no end-to-end outreach volume.
The company
Kevin runs Notorious MBA and related projects including ExitGuru. The operating agenda spans brand and offer strategy, written content, email marketing, asset creation, lead research, cold outreach and sales-pipeline work. Each lane benefits from different context and standards, yet the resulting work still belongs to the same business.
The workspace also needed to feel native to a German-speaking operator. Channel names, agent behavior and progress surfaces could not read like a generic English-language demo pasted onto the business.
The constraint
A single assistant would collapse distinct responsibilities into one queue. Kevin would still need to restate venture context, explain whether a request was strategic or tactical, and keep creative work separate from sales research. Separate standalone bots would create a different failure: departmental capability without shared memory or safe coordination.
The bottleneck was the routing layer between useful tools and the work Kevin actually delegates.
Why existing tools were not enough
Make, Drive, Calendar, Composio, Kie.ai and lead-workflow tooling could each perform part of the job. They did not define which specialist owned a request, what Notorious MBA or ExitGuru context should be retrieved, where an output should appear, or what tool execution was safe.
That required departmental roles, channel bindings, durable context and an allowlisted execution posture. Without those controls, adding more tools would increase the number of places Kevin had to supervise.
Transformation map
Before
Growth request → Kevin supplies context and chooses tools → generic chat or manual workflow → Kevin moves the result onward
After
German business channel → named specialist → venture memory → permitted tool path → review channel or next department
One workflow, end to end
This is the implemented content-to-creative route; the source verifies its components, not a published volume of completed campaigns.
- Input: Kevin places a content requirement in the appropriate German channel rather than briefing a general chat from scratch.
- Routing: The binding assigns the task to the content lane, with strategy or prompt work separated when required.
- Context: gbrain retrieves retained Notorious MBA or ExitGuru material from the shared business memory.
- Draft: The content specialist produces the written direction within its own workspace and operating instructions.
- Creative handoff: When an image or video asset is required, the creative lane uses its installed Kie.ai path rather than broad shell access.
- Review: Output returns through the asset-review or business channel for human judgment before external use.
- Reuse: Prompt or skill patterns that should recur can be delegated to the prompt-engineer or skill-creator lane.
What changed
- Built: Six separate agent workspaces for command, content, creative, strategy/outreach, prompt engineering and skill creation.
- Connected: German Discord routing, gbrain memory, Kie.ai image/video tooling and local Whisper transcription were configured around those roles.
- Tested: Channel bindings, memory paths, creative tooling and strict allowlist behavior were checked during implementation.
- Runtime verified: Six agents and 36 bindings were visible on 30 May 2026 in the Hostinger-hosted environment.
- Actively used: The retained case record does not contain activity totals or campaign receipts sufficient to claim routine adoption.
- Provider or client dependency remaining: Composio, Apify, Make, Google and outreach actions depend on their corresponding authenticated routes and campaign approval.
- Financial result reported by client: None retained.
Proof
- Agent receipt: Main, content, creative, strategy, prompt-engineer and skill-creator were configured as separate workspaces.
- Routing receipt: 36 bindings were observed at the verification point.
- Operating-surface receipt: Named channels covered direct command, tasks, text, email, content planning, creative review, lead research, cold outreach and sales pipeline.
- Control receipt: Tool execution used strict allowlist behavior rather than granting the agents broad shell access.
- Claim boundary: The deployment receipt does not establish booked revenue, qualified-lead volume or hours saved.
Technology - revealed last
The system runs OpenClaw 2026.5.22 on a Hostinger VPS with Discord, billing-proxy routing, Ollama, gbrain, Kie.ai, local Whisper, Composio checks and Apify CLI support. Each agent has its own operating files and workspace. That separation matters because the prompt engineer should not behave like the outreach operator, while shared memory keeps both from becoming context-free bots.
Client quote
No source-backed verbatim client quote is published for this case. The public claim is limited to the system and runtime evidence retained in the project record.
Next transformation
See how Scaling Webinars coordinated eight specialist departments through one command center. It shows the same departmental principle in a webinar-delivery environment with a larger retained activity receipt.
Get your AI blueprint
If your growth functions need specialists but still share one founder's context, the first design decision is routing—not model choice. Book a free AI audit to identify the department whose handoffs are costing the most attention.
Source record
- OPENCLAW-DEEP-CLIENT-CASE-STUDIES.pdf, page 3.
- RECAST_CASE_STUDIES_2026_UPDATED.pdf, page 9, document RA-CASES-001, version 2.0.
- RECAST_WEBSITE_MASTER_BLUEPRINT.md, sections 10.4, 11 and 58.