What is an AI harness, and how do you build one?
A harness turns a generic AI model into a colleague who knows your work. How I explain it, with the junior metaphor, a sales example and real workshops.
Written by the two of us who run the workshops. What we build, what breaks, and what still runs a month later.
A harness turns a generic AI model into a colleague who knows your work. How I explain it, with the junior metaphor, a sales example and real workshops.
A use case breakdown of an AI system built on a manufacturer's ERP data: how we scoped it, made the data safe for AI, trained sales and IT, and what we got wrong.
What AI work succeeds in European companies, what you can buy, and an honest look at how AI partners in north-west Europe, like Repetive, SPAIK and ML6, work.
AI workflow automation explained for knowledge teams: how it differs from RPA and agents, real examples, where people stay in charge and how to build one.
How to integrate AI into your business: start with one workflow, buy before you build, fix what your data means, name owners and cover the EU AI Act.
How Dutch SMEs can develop an internal AI team, learn through real work, measure progress and use outside expertise while keeping ownership inside the business.
Explore 10 practical Jev AI use cases for European SMEs, with guidance on GDPR, the EU AI Act, US data hosting and a safer pilot for Dutch businesses.
How to get employees to use AI at work: start with recurring work they want to improve, practise on it, and turn what works into shared team workflows.
AI can save time and still create more work downstream. The reason is usually not the model, but a workflow where production gets faster than review, coordination and decision-making.
How to measure AI productivity at work: track the workflow, not prompts or hours saved. A practical scorecard for cycle time, effort, quality and capacity.
AI workflow redesign means changing the whole process, not speeding up one step. What it looks like, why engineering got there first and how to pick a workflow.
AI vendors profit from confusion. See the adoption order that protects your team from dark patterns, runaway agent bills, and hype-first training.
GPT-5.6 has been generally available since July 9, 2026. Knowledge cutoff: Feb 16, 2026. Sol $4/$20, Terra $2/$12, Luna $0.20/$1.20 per 1M tokens.
How to write AI skills that do your real job: 8 mistakes I made building 50+ skills for Claude and Codex, plus the easy way to start without code.
The US Commerce Department suspended Claude Fable 5 on June 12, 2026. Access returned July 1. Here is what Fable 5.1 changed and which model to use now.
Claude Code's OTEL_EXPORTER_OTLP_ENDPOINT can export prompts, responses, tool calls and file paths, not just costs. What that means under GDPR and Dutch law.
Chat and Cowork are now one Claude. Here is when knowledge workers should use Claude and when to use Claude Code, updated for October 2026.
AI bills aren't the problem — leaky workflows are. Learn how to measure AI productivity with an effectiveness ratio and stop optimizing for cheap.
Your company's security team isn't blocking AI to be difficult, they have real concerns. This post gives you the concrete, plan to address them: what they actually worry about, where your data goes, the secure setup options through your existing cloud provider, and a checklist you can hand to IT on Monday morning.
How I set up Claude for consulting: one Project per client, shared Skills, CLAUDE.md, connectors, client data rules and where Claude gets consulting work wrong.
AI gives generic answers when it can't see your CRM, Slack or docs. Here is how to connect it to your business tools with MCP connectors, step by step.
I replaced 7 SaaS tools with Notion + Claude Code. Here's how to automate repetitive tasks and consolidate your AI workflow — plus a prompt to audit your own stack.
Claude Dispatch runs desktop AI tasks from your phone. Since October 2026 it is closed to new users, and cloud Cowork covers most of the same work.
Your AI gives inconsistent results because your instructions are a mess. We fixed it with 5 layers: contract, contextual rules, skills, agents, and commands. Here's how to build the same system for your team.
OpenAI took a reputation hit with their Pentagon deal while Claude beat ChatGPT in downloads for the first time. Microsoft made Copilot model-agnostic. Anthropic launched Claude Marketplace for outcome-based procurement. The single-model winner narrative just died. Here's what it means for teams.
A complete walkthrough of my multi-agent Claude Code setup — how I split AI workflow automation across Opus, Codex, and Sonnet to ship faster.
Step-by-step AI training case study: how we guided a non-technical professional from idea to a working web app using AI tools and prompt engineering - perfect for AI workshops and hands-on AI training for beginners.
We tested Opus 4.5, GPT 5.1 Codex Max, and Gemini 3 Pro on a real-world coding challenge. Complete AI model comparison: which AI coding tool handled complex logic best - and what that means for your development team.
How two days of focused AI workshops shifted a team from cautious experimentation to concrete automation plans. A 5-step AI adoption framework for teams ready to move beyond the hype.
How we took a room of professionals from AI prompt paralysis to confident practitioners in one hands-on workshop session. A practical prompt engineering training framework for teams and business professionals in the Netherlands.
Companies invest in AI infrastructure while employees learn from YouTube. Why hands-on AI training for teams closes the skills gap faster than any online course.
Tell us about your team and the work you want to change. We'll help you pick the right next step.