PMEBudgetAgentsHITLROI
AI agent for SMEs: setup budget, run cost and traps
What an SME AI agent costs: setup ranges, run (tokens, hosting, support), HITL and what to scope before you sign. No vanity metrics.
12 min read
Read article →
Blog
Field notes on production AI
Long-form guides from kengdev field practice: agents, HITL, automation, compliance. Without the hype.
GuidesAI news
LLMOpsEvalObservabilitéProductionCoûts
LLMOps: run LLMs in production without chaos
LLMOps for SMEs: evals, prompt versioning, observability, token cost, guardrails and deploy. Different from classic MLOps and pipeline automation.
15 min read
Read article →
Fine-tuningRAGDécisionCoûtsPME
RAG vs fine-tuning: how SMEs should choose
RAG vs fine-tuning: changing facts, stable style, cost, data, risk. SME decision grid + classic traps. Links to RAG pipeline and LLMOps.
14 min read
Read article →
RAGLLMPipelineDonnéesProduction
RAG LLM: ground a model on your data without fine-tuning
Production RAG LLM: corpus, chunking, retrieval, citations, eval, HITL and cost. SME method for grounded answers, not a generic embedding tutorial.
16 min read
Read article →
Claude CodeContexteMCPSkillsGarde-fous
Claude Code: context, compact, clear and session hygiene
Claude Code in depth: dump zone, compact vs clear, permissions, skills, plan mode, MCP (build/design), autonomy plugins, subagents and tokens. Team method to keep quality.
22 min read
Read article →
Context engineeringProductionAgentsTokens
Context engineering: making AI more deterministic in production
Context engineering for business agents: Write, Select, Compress, Isolate. What enters the model, context rot, MCP and tokens, metrics. Clear boundary with Claude Code.
18 min read
Read article →
Code reviewCIQualitéSécurité
AI code review: speed up without automatic merge
AI code review: first pass, consistency, business limits, comments only, noise calibration. No auto-merge. 2-sprint checklist.
15 min read
Read article →
Vibe codingSpec-drivenQualitéRPI
Vibe coding: when to explore and when to specify
Vibe coding (term popularized early 2025): prototype fast without reviewing code. Useful in spikes; risky in prod (security, maintainability). Switch to specs and RPI when a customer or sensitive data enters.
17 min read
Read article →
AI-nativeÉquipesDeliveryProcess
AI native: process, lucidity and team delivery
AI-native team: conventions before tools, lead time/incident/review KPIs (not commits), enablement, shadow IT, lucidity. Delivery culture under AI acceleration.
16 min read
Read article →
IA localeOllamaSouverainetéSécurité
Local AI: Ollama, self-host and real selection criteria
Local AI vs cloud: privacy, TCO, quality, latency, SLA. Ollama for dev/prototype; multi-user prod is another stack. Decision grid and security checklist for SMEs.
16 min read
Read article →
Multi-agentsArchitecturePatternsOrchestration
Multi-agent: patterns when one agent is not enough
Multi-agent systems: router, specialists, critic, orchestrator, subagents (Isolate). Only if single-agent is unreadable. Token and hop governance.
15 min read
Read article →
MCPIntégrationAgentsSécurité
Model Context Protocol: wire an agent to your tools without chaos
MCP (open standard, Anthropic late 2024): host, client, server, JSON-RPC tools. Schema token cost, security. Computer use as a section. SME guide.
17 min read
Read article →
AutomatisationPipelineCI/CDOps
AI automation: from isolated chat to pipeline
Enterprise AI automation: stages sidebar → scripts → tooled agents → CI/CD. ROI, governance, guardrails. Different from the first business agent (support, follow-ups).
16 min read
Read article →
HITLGouvernanceQualitéProduction
Human in the loop: designing supervision for AI agents
Human in the loop (HITL): thresholds, validation queues, capture, review fatigue. Not agent build: production supervision design.
16 min read
Read article →
AgentsPMEAutomatisationHITL
How to create an AI agent in an SME (without picking the wrong target)
How to create a useful AI agent: process first, LLM + tools + control, first flow that pays, allowlist, KPI, HITL. Multi-agent later. 40-minute scoping method.
18 min read
Read article →