AI Governance for Small Businesses: Start in 8 Hours

0:00No governance team? One day is enough

Welcome! No governance team? One day is enough to start. AI Governance Chain's one-day starter builds a minimum viable AI governance programme in 8 hours: 4 steps of about 2 hours each, from scope to evidence. It works today, with commands shipped since v0.1 on 4 November 2026.

0:41A client asks about the EU AI Act

Picture the owner of a small agency that resells an AI chatbot under its own brand. On Monday, a client asks what the EU AI Act means for the product. There is no governance team and no lawyer on staff. By evening, she needs a map of what applies, a file that records her controls, and evidence she can show.

1:03The Obligation Atlas behind the path

The 4 steps follow the chain from OBLIGATION to CONTROL, GATE and EVIDENCE. They draw on the Obligation Atlas: 198 obligations across 17 regimes, dated 4 November 2026.

1:16Step 1, morning: aigovchain scope

Step 1, morning: run aigovchain scope with jurisdiction, role and use. For an EU chatbot, with the roles provider and deployer, it lists 3 duties that apply today: no banned practices, Article 5; AI literacy for staff, Article 4; and telling people they are talking to an AI, Article 50(1).

1:36Step 2, late morning: AI_GOVERNANCE.md

Step 2, late morning: run aigovchain init. It writes AI_GOVERNANCE.md, one plain file with the system, its owner, the duties in scope and a control for each. Replace every to-do with a real status.

1:49Step 3, afternoon: Shield GitHub Action

Step 3, afternoon: add the Shield GitHub Action. Every pull request runs 17 checks on the file. A chatbot is not an agent, so the 5 agent checks are skipped, and any failure names the missing control.

2:03Step 4, late afternoon: the Evidence Pack

Step 4, late afternoon: run aigovchain evidence. The Evidence Pack holds a readable summary, a JSON record and a fingerprint of the file: what was declared, what was checked, and when.

2:14Who uses the one-day starter?

For an agency owner, it turns a client question into a dated answer.

For a product lead, it is the first record of the AI system and its controls.

For a developer, the checks and the evidence sit beside the code, in the repository.

2:31Limits of an 8-hour programme

This is a first programme, not a full assessment. The Action checks the declared file, not the model's behaviour, and the Evidence Pack is not an audit. Confirm the rules and sources for your system. This is governance guidance, not legal advice in any jurisdiction.

2:48Block a day and start

Block one day this week and start with aigovchain scope; the link is in the description.

One day gets you started. What keeps you going for a month?

Thank you for watching!