---
title: "Minimum Lovable Governance"
date: 2025-11-30
description: A principle borrowed from product thinking: build the smallest AI governance people genuinely value, then grow it deliberately.
author: Mario Thomas
canonical: https://mariothomas.com/remake/library/minimum-lovable-governance/
---

## What it is {#what-it-is}

The name borrows deliberately from product thinking — Eric Ries's progression from Minimum Viable Product to Minimum Lovable Product, documented in *The Startup Way*. Applied to governance: heavyweight frameworks get complied with reluctantly or routed around entirely; governance that is lovable gets used voluntarily. It marks the shift from episodic compliance scrambles to continuous, embedded oversight.

## The rule {#the-rule}

Build the smallest governance system that achieves the necessary guardrails **and** that people actually want to engage with. The test is behavioural, not documentary: if people are routing around your governance, it isn't governance — it's documentation.

## Why it exists {#why-it-exists}

Traditional AI governance fails in practice even when it succeeds on paper: comprehensive policies nobody reads, elaborate approval processes for low-risk experiments, and minimal oversight of the high-stakes autonomous systems that actually warrant it. The evidence is stark — when more than 80% of employees use unapproved AI tools at work, the gap between compliance and adoption is the strategic fact. Governance that is heavy where it should be light drives the shadow AI it exists to prevent.

## What it permits and forbids {#permits-and-forbids}

It permits more than governance-minimalists expect: sanctioned low-friction experimentation, proportionate paths for low-risk work, and judgement exercised at the point of decision rather than escalated by default. It forbids the two comfortable failure modes — governance as theatre (weeks of preparation compressed before an audit, then nothing) and governance as blanket weight, where the same process burdens a marketing pilot and an autonomous production system. Proportionality to risk is not a concession; it is the rule's edge.

## Applying it {#applying-it}

Four characteristics mark governance that passes the test: **embedded** rather than separate — controls living inside the tools and workflows where work happens; **continuous** rather than episodic — always-on oversight, not audit-season sprints; **proportionate to risk** — friction scaled to consequence; and **clarity at the point of decision** — the person facing the choice can answer "am I allowed, and who decides?" without leaving the room. In the boardroom, the principle is the counter-question to every governance proposal: what is the smallest version of this that people will actually love enough to use?

Within [Remake](/remake/), Minimum Lovable Governance is declared a **Principle** — a governing rule that guides judgement. It is not read at one stage; it runs through the whole ADAPT motion, shaping how every stage's governance is sized: enough structure to demonstrate good faith, whilst preserving agility.

## Questions {#questions}

No — proportionality cuts both ways. MLG is often *heavier* than the status quo where it matters: autonomous, high-stakes systems get continuous oversight that episodic compliance never gave them. What it removes is unloved weight in the wrong places — the approval queue in front of a harmless experiment.

Watch behaviour, not documents. Shadow AI proliferating, policies unread, approvals bypassed, audit-season scrambles: each is the routing-around signal, and each marks a place where the governance is documentation. Adoption is the metric — governance that is used is governance that exists.