Djtal Blog
Regular articles by Laurent Cuénoud, founder of Djtal, on what our deployments teach us and what that means for your business.
Djtal explains how to split an AI agent's work across several models: the orchestrator and subagents pattern, what Claude 5.5 changes and three key settings.
A website translation launched overnight, subagents on Fable and the week's Claude Code quota used up at four in the morning.
Djtal explains the AI audit for Swiss companies: what it contains, deliverables to expect, typical fees (CHF 800 to 8,000) and questions to ask before signing.
OpenAI launched Dots, AI agents that run around the clock, on 29 September 2026. In Switzerland, access is via company accounts. Djtal sets out three questions.
Between AI agents, the telephone game disappears. Everything travels as a written artefact, byte for byte, without exception.
A telling slip of the tongue, and a portrait of the AI agent who became management's right hand, from decisions to orchestration.
Acemoglu (Nobel 2024) puts AI's gain below 0.66% over ten years and a Goldman Sachs report airs the doubts. Djtal turns their sums into a way to invest.
The day before a client meeting, an AI agent studies the requirements and prepares the proposal, overnight and then on the train.
GPT-6 Astra, the first model rated ‘Critical’ for cyber risk, ships restricted, off by default and interruptible. Your AI agents need the same rule.
An AI agent plans a house move, compares the wood floors, briefs the tradespeople and drafts the change-of-address letters.
Claude Fable 5.1 launched on 1 September 2026. Djtal migrated its 14 AI agent repositories that night in five steps. We cover the effort setting and the prices.
Three incidents in 141,006 test runs led Anthropic to freeze most new features. Djtal sets out five minimum governance rules for AI agents in production.
The illusion of an AI that “figures it out” on its own. Code checked, edge cases tested: the developers' expertise stays central.
Twenty agents now handle the admin and supply the expertise. The leader keeps what matters: deciding, and learning all the time.
Many Swiss executives think the EU AI Act is not their concern. Djtal shows why neither reason holds, what applies to Swiss companies and how to prepare.
From 7 to 20 AI agents in a few weeks. The latest was born in fifteen minutes: Sofia, an agent herself, creates the agents.
Orchestrating seven AI agents keeps the business leader at the helm, busier than ever. No arms, no legs: the beach can wait.
Djtal tracks its own visibility in ChatGPT and Perplexity with a free 30-minute monthly protocol. The full method and our measured scores, weak ones included.
A relationship with an AI agent is unprecedented, persistent and disconcerting. You end up needing to come back down to Earth.
AWS commits USD 1 billion, Microsoft USD 2.5 billion and OpenAI a deployment company. Djtal explains what this engineering wave means for Swiss companies.
Consultants recommend, integrators install, forward-deployed engineers build inside your systems. Djtal explains which one to choose for your AI project.
Stuffed with knowledge, my seven AI agents put on weight and squabble over files, until an overhaul gets them agreeing again.
A real case at Djtal: how an AI agent prepares a meeting, works alongside you during it, files everything afterwards and comes back sharper next time.
Days before my daughter's exam, an AI demonstrator summarises a whole course, its run time cut from half an hour to three minutes.
Once my AI agents are up and running, the ideas pour in. One sheet of A4 per agent, and a whole business model reinventing itself.
Setting up an AI agent, seen from the inside: how Djtal's finance agent Émile went from a written role to running our Zoho Books accounting, supervised.
Claude, ChatGPT, Gemini or Zia? Djtal runs its own finance agent and shows why, for a Swiss company, the task, data and sovereignty matter more than the model.
Why do agentic AI projects fail? Djtal, which runs nine AI agents in its own company, sets out the three most common traps and the remedy for each.
Yes, an AI agent can work inside the CRM you already run: Zoho, Salesforce, HubSpot. Djtal explains how it works, where built-in AI stops and where to begin.
Claude Fable 5 launched on 9 June. Djtal ran its seven AI agents on it from 11 to 13 June. What the model changes for a Swiss SME, and what it leaves unchanged.
My AI agents never sleep. Approvals at all hours, nights cut short: how tiring it is to be the human working alongside them.
What is agentic AI? Djtal sets out the difference between a chatbot, a single agent and a team of agents, drawing on the seven agents at work in its own company.
Seven AI agents turn up with neither name nor face. Hours of hard work giving them an identity, and the day Adil became Adila.
MIT's GenAI Divide study found that 95% of enterprise AI pilots never reach a measurable ROI. Method decides the outcome, and method can be learned.
Seven AI agents turn up at my door one Monday morning. The first rule I set them for working together: no flattery, no bullshit.
Should you say please and thank you to your AI? Djtal's founder weighs three circles of impact and a 2025 study in which rude prompts were more accurate.
Hundreds of emails a day and an inbox that stays in order: Djtal runs its own mailbox with a Claude agent that sorts, summarises, processes and logs tasks.
A customer returns her empty bottles and gets her discount voucher by email minutes later, with no manual entry: an AI agent designed by Djtal runs the process.
The Djtal blog regularly publishes articles on applied AI for Swiss companies: lessons from the field, methods, analysis and early signals worth watching.