---
title: "Action Research"
id: "1739"
type: "page"
slug: "action-research"
published_at: "2026-09-20T20:56:09+00:00"
modified_at: "2026-09-21T01:36:28+00:00"
url: "https://xedant.com/agents/research/docs/action-research"
markdown_url: "https://xedant.com/agents/research/docs/action-research.md"
excerpt: "Action Research (AR) is a methodology that turns a business — or any activity —…"
---

# Action Research

[https://xedant.com/agents/research/docs/action-research.md](https://xedant.com/agents/research/docs/action-research.md)

Action Research (AR) is a methodology that turns a business — or any activity — into hierarchical plan → act → observe → reflect cycles. The Research Agent keeps the cycles as library documents: the reflection of one cycle becomes the plan of the next, so the work winds into a spiral instead of running in place.

## The cycle

A cycle is plan → act → observe → reflect, and it never really “finishes”: continuous iteration until the desired result — and beyond. The phases are not isolated — act and observe intertwine, and reflection goes on through the whole cycle, not only at the end.

Agile in software development is one implementation of Action Research: the same spiral of short iterations with observation and adjustment. AR applies it to any business.

## Three levels

- **Strategic** (the hour hand) — the evolution of the business as a whole: the business plan, KPIs → `research/strategic-cycles/`;
- **Tactical** (the minute hand) — the evolution of individual directions: the offer, the marketing task queue → `research/tactical-cycles/`;
- **Applied** (the second hand) — concrete tasks as mini-cycles → `research/applied-cycles/`.

The clock rule: the seconds must move the minutes. Short cycles are obliged to genuinely move the long ones — the speed of your business’s evolution equals the speed of cycle iteration at every level.

## Double-loop learning

Single-loop learning changes tactics while keeping the same assumptions; double-loop learning changes the assumptions themselves. The strategic cycle explicitly decides what it is doing: optimizing the model (first-order) or changing the assumptions (second-order). Above the levels runs a meta-cycle — a periodic review of the cycles themselves: which are alive, which have run out of steam, where iterations are missing.

## Rules for running cycles

- the hypothesis is public and testable, with a verification metric — not “let’s try and see”;
- execution records only the facts of what was done; plans and checklists live in the planning section;
- new stages of a cycle are not written in advance — observe and reflect appear as data accumulates;
- reflection ends with a concrete decision: continue, change, or close the cycle;
- brevity, and no interaction during the process: the work runs on the best available assumptions, which are listed at the end.

## Live examples

The product ships with our live cycles: strategic C1 (the business plan), C2 (the Action Research methodology), and C3 (KPIs); tactical — the offer and the marketing task queue. See [Built-in Research](/agents/research/docs/built-in-researches)
.

[← Back to documentation](/agents/research/docs)