The method
Four sequential phases, tested at Albus Factory (95 agents in production) and replicated at Boomrang Events in two weeks. Audit, Memory, Blueprint, Deployment: a complete system shipped, transferred to your team, self-optimizing.
A complete system in production, transferred to your team, with measurable ROI from the first quarter.
Full mapping of workflows, tools and time leaks. The standalone AI Audit (6.5K EUR) gets you this step in 2 weeks. The 3-5 high-ROI projects are identified and quantified before any Mission commitment.
Building the company's AI second brain. Without structured memory, agents stay generic. It's the most decisive phase for the quality of results.
Custom AI-First system architecture. Every agent specified, every KPI defined before the first commit.
Agents built and shipped to production. Short iterations, first measurable results before the end of the project. By this point, the first FTEs are already absorbed by the agents. ROI is measured, not promised.
Full documentation, team training, handover. Option: Fractional Chief AI Officer (10K EUR/month) to keep the system alive and ship 1-3 new agents per month.
10 questions. 5 minutes. A personalized recommendation sent by email.
Not all agents are equal. Depending on the problem to solve, you build a triggered agent, an autonomous agent, or a multi-agent system. Each level opens new possibilities.
A specialized agent a human launches on demand to produce a calibrated deliverable.
A 'SEO report' agent triggered before every executive committee. It queries Search Console, cross-references operational data from the wiki, and delivers a complete report in 3 minutes.
One-off, high-value task that requires a human trigger.
An agent that runs on its own, on a recurring cycle, with no human intervention.
A 'CRM Audit' agent that scans HubSpot every morning, detects anomalies (deals with no next step, duplicate contacts, stale opportunities), and posts a prioritized report on Slack at 8am.
Repetitive task, need for consistency, high volume.
A team of specialized agents that collaborate to handle a full perimeter.
Google Ads SEA management: 'Keywords' agent (research and integration), 'Reporting' agent (weekly performance), 'Budget' agent (arbitration). They read the same memory, exchange info, alert a human on critical decisions.
Full perimeter to delegate, complexity beyond a single agent's capacity.
The tools I use day-to-day to build and operate an AI-First architecture.
The architecture works as a stack. Each layer builds on the next. Everything is custom-built.
Operational knowledge captured in .md files: processes, offers, personas, rules, history.
Architects, analysts, executors. Each one has a mission. All read the memory. All document.
Slack · Gmail · Drive · HubSpot · Figma · Google Ads · Search Console. All your tools.
Agents act on their own. They escalate to a human for trade-offs or report sign-off.
Every agent writes its lessons after each run and reads them at the next one. The system learns. This is the layer that makes agents better over time.
The 5 layers are universal. Their content is unique to each company. Your operational memory, your agents, your stack, your governance: nothing is copy-pasted from another client. Anyone selling "off-the-shelf" AI workflows hasn't understood the problem.
Internal proof from my own group. First external missions are underway — client cases will land here as they sign off.
Sales, growth, ops, project
“In 6 weeks, I avoided 4 hires. 9 people now do the work of 50. Less than 50,000 EUR invested. This is the live lab of the method.”
MICE pipeline · brief → PPTX proposal
“Second company in the group, same method. Autonomous sales pipeline shipped in 2 weeks. Proof of reproducibility: memory is unique to each company, the architecture is universal.”
B2B services · distribution · logistics
“This third row isn't a client testimonial — it's the kind of mission I run today. Outcomes depend on context; the method is identical to what's deployed at Albus and Boomrang.”