Measured costs, real benchmarks and what we learn building AI workflows for small teams.
Blog
Local LLM vs API cost: a 4B model on a 2019 Core i9
We ran Qwen3-4B on a 2019 Core i9 with no graphics card to sort 20 support emails: 5.8 seconds each, 17 of 20 categories right, a few cents per 1,000.
Why AI automations break after launch: model retirements
Model retirements and removed API parameters break AI workflows. A dated retirement table, what breaks first and a migration runbook you can copy.
Is it safe to put confidential documents into ChatGPT?
What ChatGPT, Claude and Copilot keep, train on and for how long on each plan, linked to each vendor's policy, plus the admin settings to change.
AI acceptable use policy: a free one-page template
A one-page AI acceptable use policy for small teams, free to copy with no email, covering approved tools, data rules, personal accounts and review.
How much does AI automation cost? A payback worksheet
The build, run and upkeep costs of an AI automation, priced from published pages, plus a two-week time study to measure what the work costs you today.
Claude Code for teams: cost per developer and first settings
What Claude Code costs a team of 2 to 20 developers on a Team plan or API billing, from Anthropic's own pages, and the settings to set in week one.
Core ML vs Core AI: which Apple framework to use in 2026
What Apple's new Core AI framework does, what stays in Core ML and where Foundation Models fits, taken from Apple's own docs and WWDC26 sessions.
Which processes to automate with AI first: a scoring sheet
A free scoring sheet for picking the first workflow to automate with AI: volume, rules, error cost, data access and exceptions. No email needed.
Claude Team vs ChatGPT Business for a 10 to 50 person team
Prices, seat minimums, admin controls and data settings for Claude Team and ChatGPT Business from each vendor's pages, plus where Copilot Business fits.
n8n vs Zapier for AI agents: one workflow priced both ways
One AI lead triage workflow priced in n8n and in Zapier: run cost at 1,000 and 10,000 runs, error handling, version control and who owns it.
AI agent for invoice processing: when it must ask a human
Approval rules for an AI invoice agent: amount limits, new suppliers, changed bank details, failed checks and duplicates, plus an Odoo and Slack test setup.
AI agent cost per month: tokens, hosting and review time
What one AI agent costs per month at 1,000 and 10,000 runs: API tokens on three models, prompt caching, hosting, monitoring, review time and upkeep.
Open WebUI vs LibreChat for business: permissions and SSO
How Open WebUI and LibreChat control which staff see which documents, plus SSO, audit logs, licences, setup and a two-role leak test to run before rollout.
On-premise LLM hardware: how many staff can one GPU serve?
How to size an on-premise LLM server: GPU memory for the model and the KV cache, active vs total users, and a vLLM load test you can run on one rented GPU.
Foundation Models framework in a business app: what works
What Apple's on-device Foundation Models framework can do in a business app: summaries, field extraction, sorting, device limits and the 5.1.2(i) consent rule.
How to get your business cited by ChatGPT and Perplexity
What OpenAI, Perplexity and Google say about appearing in AI answers, which AI crawlers to allow in robots.txt, and how to track AI visits and leads in GA4.
Not sure which service fits?