Blog category
Workflow strategy
How to plan automation that stays maintainable and useful in daily work.
Why 72 percent of AI projects in SMEs fail and how to avoid the pilot trap
72 percent of AI projects in SMEs fail, says the latest survey. We explain why most pilot projects do not make the leap into production, which concrete patterns we see in our projects, and how the pilot trap can be avoided in four weeks.
Read articleGitHub Copilot Skills vs. MCP Servers: What to Use in July 2026, and What Not
Microsoft rolled out Copilot Skills in June 2026. Three use cases where Skills clearly beat MCP servers, and three where MCP servers clearly beat Skills. Plus a heuristic and a 4-week recommendation for the inventory.
Read articleMulti-agent architectures 2026: Why German SMEs should wait out the hype
Multi-agent architectures are on everyone's mind in 2026. We explain why vendor promises rarely fit German SMEs, where the technology actually makes sense today, and where SMEs should better wait it out.
Read articleOpen-source AI in 2026: Why the license now matters more than the benchmark comparison
The open-source AI landscape in 2026 has matured: Llama 4, Qwen 3.5, Mistral, and DeepSeek ship production-ready models. We map out what has really changed for SMEs, where open-source makes sense, and why the license is often more important than the next benchmark.
Read articleAI agents 2026: three autonomy levels and why most SMEs stop too early at level one
Anyone running AI agents in production quickly runs into the autonomy question. Three levels help allocate risk and benefit cleanly. Few SMEs scale past level one, even though that is where the largest lever sits.
Read articleAI automation 2026: What SMEs should actually expect in the next twelve months
Where AI automation really lands in 2026: away from copilot hype, toward concrete workflows, affordable agents, and built-in governance. A sober outlook on the next twelve months.
Read articleOpen-source AI frameworks 2026: why community matters more than the benchmark winner
LangGraph, CrewAI, AutoGen and Dify are running a head-to-head race in 2026 for dominance in the open-source agent stack. The framework teams choose is increasingly decided by community signals, not benchmark tables.
Read articleAgent Loops and Loop Engineering: Old wine in new bottles?
In June 2026, "Agent Loops" is trending across the AI community. Peter Steinberger, Boris Cherny, and Addy Osmani frame it as the next step. We investigated what is genuinely new and what is just a relabel of concepts that have been in every good agent document since 2023: ReAct, Reflexion, iterative tool-use loops, stop hooks. A critical comparison for SMEs and decision-makers.
Read article