Notes on shipping AI systems.
Working notes on the systems we build: the architecture calls, the tradeoffs behind them, and the production numbers that came out the other side.
The best MCP servers for business, sales and marketing are the four you keep
Which MCP servers for business, sales and marketing are worth connecting: official vs community, read vs write scope, auth, and the cost per extra server.
Vector databases for production RAG: Pinecone, Weaviate, Qdrant, Chroma and pgvector at 500K records
A production RAG comparison of Pinecone, Weaviate, Qdrant, Chroma and pgvector: filtered search, hybrid retrieval, RAM cost and the reindex nobody plans.
Enterprise AI agent deployment case studies - what six known rollouts actually show
Six enterprise AI agent deployments read through one grid: Klarna, Air Canada, drive-thru voice, CommBank, Wells Fargo and one of ours. Sources included.
AI customer support ROI: what a resolution costs and what comes back
What AI customer support costs per resolution, what vendors actually count as one, and a worked ROI calculation you can run on your own ticket volume.
Business process automation: what to automate, what it costs, what quietly fails
What business process automation is, how it differs from RPA and digitisation, the arithmetic that picks the right process, and where builds quietly fail.
ICP scoring for B2B lead qualification: the criteria that predict fit
ICP scoring criteria for B2B lead qualification: which ones predict fit, how to weight them, a worked example, and when a model beats a lookup table.
AI agents vs workflows: one of our seven systems had to be an agent
AI agents vs workflows comes down to one thing: who owns the order of steps. Our production record, the research, and a rule you can apply to your own project.
EU AI Act compliance for automation you already run
EU AI Act compliance for systems already in production: what binds you from 2 August 2026, what moved to December 2027, and what to change in the code.
Connect AI agents to your data without handing them a copy
How to connect AI agents to your data through a tool layer instead of a synced copy: sizing tools, permissions, stale context, latency and the write gate.
RAG for enterprise knowledge base - what actually breaks at 500K records
RAG for enterprise knowledge base search degrades quietly at scale. What breaks on 500K records, in the order it breaks, and what each fix costs to add.
AI automation for small business starts with a process, not a tool
AI automation for small business, decided process-first: four tests you can run this week, evidence from seven production systems, and a worked example.
Why AI agents fail in production - a post-mortem with real numbers
Why AI agents fail in production is rarely a crash. It is a fluent, confident, wrong answer that passed the demo. A post-mortem of five failure modes.
AI grants in Poland, decoded: what a grant of each size actually buys
AI grants in Poland (PARP, Dig.IT, FENG) sell the maximum amount, not the scope. The frame: what a grant actually buys, and the one phase it never funds.
AI agent development cost, itemized: a frame to price any quote
Most AI agent development cost guides give a $10K-$400K shrug. Here's the frame: the four phases, the run cost they skip, and questions to price any quote.
Poland's KSeF e-invoicing is live. Here is the invoice pipeline your Polish entity needs.
KSeF is live in Poland. A practitioner's guide to the e-invoicing pipeline: FA(3) XML, validation, GL coding, ERP, and where AI helps vs where plain code wins.
Build vs buy for AI: the thresholds that actually decide it
Vendors say buy. Platforms say build. A studio that does both walks through the real thresholds - with cost tables, worked examples and a trial checklist.
See what this would look like in your operations.
30 minutes · we map your 3 best automation opportunities · no obligation