LeanX AX Insights

Practical guides to AI at work

Short, practical guides on AI agents, workflow automation, and running your first AI pilot. Written for busy managers.

Book a free call

English guides

12 guides

AI automation

How to find AI automation candidates by shadowing real work

Before you pick an AI tool, follow the person doing the job. Ten interview questions help you find what's safe to automate and where a human should stay.

5 min read

AI agents

AI workflow or agent: how to pick the right one for a task

Not every task needs an autonomous agent. For tasks with fixed inputs and outputs, a workflow may be enough. Use this decision table to choose based on predictability, exceptions, and permissions.

4 min read

AX pilots

The 2-week AX pilot checklist for rolling out an AI agent

AI agent rollouts fail without a task contract, staged permissions, human review, and logs. Run this 2-week pilot checklist before you scale up.

8 min read

AI operations

The 4-step AI delegation ladder: read, organize, draft, execute

Hand off work to AI in four steps: read, organize, draft, then limited execution. Move to the next step only when you can explain how the current one is checked, stopped, and undone.

5 min read

AI agents

AI agent permissions: split read, draft, and execute

Give an AI agent one stage at a time. Let it read the files it needs and draft into a review space. Execute only after a specific approval. Here's how to split permissions by stage.

4 min read

AI operations

How to turn team know-how into an AI agent skill

Your best performer may already know review criteria your AI is missing. Package that judgment into an Agent Skill others can call up, with a ready-to-edit SKILL.md template.

4 min read

AI agents

Context engineering: 7 reasons your AI agent keeps failing

Rewriting the prompt again won't fix an agent that read a stale policy or lost track of its own state. Here are seven context failures to check before you touch the model.

5 min read

AI agents

The AI agent harness: what to define before a longer prompt

Before you write a longer prompt, set up the agent's harness. Define what it can read, which tools it can use, and what it carries forward. Then set how far it can act and when it stops for a human.

4 min read

AI agents

How to build an AI agent evaluation set before you launch

Before you ship an AI agent, build a 20-case test set covering normal inputs, common exceptions, and forbidden actions, not just one good demo.

4 min read

AI operations

Measure AI adoption by completed work, not usage count

Usage counts don't show if AI actually finished the job. Track one task's start, review, and completion for a week to see real value.

4 min read

AI agents

How to calculate AI agent ROI: cost, approval, and lead time

AI agent ROI is time saved and faster processing minus every real cost — model, tool, review, and failure. Here's the formula and a sample P&L.

4 min read

AI automation

Why outreach automation needs a state machine before an AI agent

One builder reported spending 44+ hours on a PR outreach build planned for 10. The lesson: build lead status, evidence fields, and approval before AI drafting.

4 min read