Blog
Pragmatic deep-dives on LLMs, autonomous agents, hardware and emerging standards. What actually matters when you're shipping AI products today.

A concrete 90-day plan to get AI into production in a small or mid-sized company: what to do in month one, two and three, what each phase costs, and the mistakes that kill these projects before Christmas.

Does your customer data leave the company when you use an AI model? It depends on design choices. What changes between ChatGPT, paid APIs and local models, what GDPR and the AI Act require, and the three setups I use in real projects.

Twelve questions that tell you whether an AI project will work or burn budget: data, processes, people, integrations and measurement. With scoring and what to do at each level.

A ChatGPT subscription costs 20 € a month; a custom agent starts at 2,500 €. When DIY is enough, when it wastes more than it saves, and the numeric rule for deciding without hype.

Real numbers on AI customer service capacity: what share of requests it closes on its own, what it escalates, what it costs per contact, and how to measure whether it is working.

Late payments are a cash problem, not an accounting one. How to build an automated flow that checks overdue invoices, writes the right reminder at the right time and alerts sales before the customer goes quiet.

Most sales are lost in the silence between the first enquiry and the quote. How to qualify leads automatically, prioritise the right ones and never leave a request unanswered.

Anthropic has released two AI agents for e-commerce for free and open source: one helps customers find the right product, the other analyses sales, catalogue and inventory. What they do, what it takes to run them on Shopify or WooCommerce, what they really cost and when they pay off.

OpenAI has launched GPT-6 Astra: a model that operates a computer on its own, keeps working in the background with async tool calling and corrects itself mid-task. What it really means for a business: what you can automate now, what to wait for, what it costs and which risks to handle first.

Google has released TimesFM-3, an open source multivariate forecasting model that beats its main rivals in internal benchmarks. What it can forecast for an SMB — sales, inventory, cash flow, staffing — how it integrates with existing systems, what it costs, and when you don't need it.

Rigid flows are giving way to dynamic workflows built as graphs: the AI decides the path case by case instead of following a fixed sequence. What graph engineering is, how to design it, when it genuinely pays off for an SMB and what it costs — with concrete examples and honest limits.

What n8n is, how it actually works, how it differs from Zapier and Make, when self-hosting is worth it, and five automations an SMB can put into production in a few weeks. With indicative costs and honest limits.