In-company training for teams working with agentic AI

Train your company
to operate agentic AI.

We train teams to use agentic AI in their actual work. The judgment, the habits, and the mental model that survive the next model release.

In-company training on your workflows, your documents, your edge cases, and the judgment your team needs as the environment keeps changing.

Mental model How to read a tool's claims against what it actually does.

Live context Train on the work, not on generic demos.

Adaptation Habits that survive the next model release.

"Intelligence is the ability to adapt to change." — Stephen Hawking

Three short questions. The first reply is already useful.

Fausto Albers delivering the context engineering keynote at Prosus AI House Amsterdam, April 2026
date2026.04.21 venueProsus AI House · Amsterdam format3-hour keynote + hackathon scale75 attendees · 25 teams toolsClaude Code · MCP servers

// Proof

Date Applied at Capability Proof State
2026 — 2030 AINed AIndustry Lab Industrial AI consortium Step into Liquid participates as AI Tech R&D partner in the AINed/AIC4NL AIndustry Lab consortium for applied AI in manufacturing. Running
2026 Automatic Signal Security RAG Scoping a service-search assistant across Nedap, Genetec, Bosch, and other security documentation. Running
2026 EMT B.V. ML & ERP Assessment for cycle-time prediction and planning support using Ridder iQ ERP data. Running
2026 Eneco Enterprise AI Keynote and demo for the Eneco Asset Management team, routed through the StudyTube enterprise channel. Delivered
2026 IJskoud / FreezerData RAG & maintenance Newton feasibility for refrigeration service engineers, building on the existing Virtuele Servicemonteur. · Newton → Running
2026 HGG Group Steel & CNC R&D pilot reconstructing DSTV/NC1 CNC programs from 2D fabrication drawings. Running
2026 Multimetaal Konstruktie CAD & BOM automation Research engagement on turning isometric piping drawings into BOM and saw-list output, tied to the existing CAD and ERP stack. Running
2026 Prosus AI House × MLOps Community Agentic AI training Three-hour keynote and hackathon on context engineering, delivered at Prosus AI House with MLOps Community. · View the public deck → Delivered
2025 CTBTO Conversational AI Built a conversational AI prototype navigating Nuclear Treaty topics for the Comprehensive Nuclear-Test-Ban Treaty Organization. Delivered
2024 Dell × NVIDIA Masterclass Scaling AI Co-led the Dell × NVIDIA Masterclass on the practical challenges of scaling AI inside organisations, with Jurjen Helmus (HvA). Delivered
2024 European Chatbot & Conversational AI Summit Conversational AI Speaker at the European Chatbot & Conversational AI | Generative AI Summit at Dynamic Earth, Edinburgh. Delivered
2024 ING & DataStax AI event Document AI Speaker at the ING & DataStax AI event in Amsterdam on document processing and the role of validation in LLM systems. Delivered
2024 MLOps Compound AI event LLM systems Speaker at the MLOps Compound AI event in Amsterdam on making LLMs backwards compatible with existing software systems. Delivered
2024 RAAIT — Responsible AI in Haarlem Responsible AI Speaker at the RAAIT event on responsible AI in Haarlem — practical guardrails, not policy theatre. Delivered
2023 — Present AI Builders Club Community & talks Co-founded and host the AI Builders community — monthly meetups and hackathons across six cities, anchored on what is actually shipping. Delivered
2020 — 2023 Waitler.com GenAI product Co-founded and shipped a GenAI hospitality assistant — early production work on LLMs and RAG before the playbook existed. Delivered
2015 — 2024 The Louisiana Lobstershack Operations & UX Founded, ran, and sold a landmark Haarlem restaurant — the operating ground that informs how the training treats live work. Delivered

11 delivered · 6 running

Applied across: agent workflows · RAG · CAD parsing · ERP prediction · Trained: context engineering · industrial AI

Public artifact: Prosus training deck

// the training is what we sell. the delivery work is why.

// What we mean by context

Three families. Every token in a forward pass is one of them. The training is what teaches a team to recognise which — and to design with that distinction in front of them.

deterministic

// path fixed

CLAUDE.md · rules · hooks · auto-memory · loops

probabilistic

// output contingent

subagents · retrieval · MCPs · skills · observers

human

// shaped by you

chat turns · slash commands · reframes

"Knowledge is not what you know — it is knowing where to find it." — Prosus AI House keynote, April 2026

// Who runs this

Twenty years running operations under load. Three years shipping production AI. The training is what the second taught the first.

Fausto Albers — Lead GenAI R&D at the Industrial Digital Twins Lab (HvA) under prof. Jurjen Helmus, founder of WonderWhy.ai / Step into Liquid B.V., and co-founder of the AI Builders Club.

Sociology at the University of Amsterdam (2001-2004), then twenty years of building and running hospitality operations: Barplanet (2007-2015) — the first cocktail and flair-bartender event company in the Netherlands, scaled to a million-euro run rate before selling shares to a co-founder; New York Deli in Haarlem alongside it; and the Louisiana Lobstershack (2015-2024), a landmark venue with €900K of revenue in its opening year, a team of eighteen, and a research line on tasting panels and the semantic shaping of flavour perception.

Co-founded the AI Builders Club in spring 2023 — now over a thousand engineers and founders running monthly meetups across Amsterdam, Berlin, Rotterdam, London, Lisbon, and Miami, with guests including Harrison Chase (LangChain), Bob van Luijt (Weaviate), Mathias Christensen (Netlify), and Nicholas Patry (Hugging Face). Step into Liquid B.V. partners with the HvA Industrial Digital Twins Lab — the practice line that pays the theory line and back.

The training is built on delivery work: Newton for service teams, RAG inside ISO 27001 environments, vision-language pilots in steel and piping, ERP-grade prediction, the conversational AI we built for the CTBTO at the Hofburg in Vienna. Long reads at wonderwhy.ai. The teams we train face the same questions we faced building it.

// Twenty years operating. Three years shipping AI. The two reinforce.

// How the engagement works

  1. 01

    Map the current reality

    We start with how your team already works, decides, and hands off.

    Better judgment in the work that already runs the business.

  2. 02

    Train on live workflows

    Use your documents, tasks, edge cases, and constraints.

    Documents you already share. Tasks you already track. Decisions you already make.

  3. 03

    Build operating habits

    Prompting is not the product. Review, judgment, and adaptation are.

    Prompting is a skill. Review and judgment are the operating model.

  4. 04

    Leave the team stronger

    Internal capability, not dependence on us.

    When we leave, the team runs better without us.

// Writing

Working notes from the operator's seat. Written while building, not after.

2026-04-26

Every Chatbot Is Becoming a Coding Agent

Why that matters even if you never write a line of code

The most powerful AI systems are not the ones that answer questions. They are the ones that execute processes — and the substrate for execution is code.

Read →

Tell us what your team does today, where AI is already showing up, and where the friction is.

We'll come back with a concrete read on whether training fits — and what it would change.