Learn · Interactive diagnostic

AI Skills Diagnostic

Fourteen quick scenarios across seven AI skills, from prompting to multi-agent orchestration. Every answer comes with a teaching note, and you finish with a personal roadmap to the HRI modules, classes and guides that fit where you are.

5 minutes · 14 scenarios · 7 skills

Fill your hive.

Each AI skill is a hexagon with three cells. Choose the best move in each scenario to fill a cell. Every answer comes with a teaching note, so there are no points to lose, only moves to learn.

How it works

Two scenarios per skill, one roadmap

Answer 14 scenarios, two for each of the 7 skills. Choose the best move in the first and the second goes deeper; otherwise the second covers the fundamentals, so your roadmap starts in the right place.
Read the teaching note after every answer. Not sure? Say so, and the diagnostic shows you the move: you learn more than you would from a guess.
Get your roadmap: your level, what unlocks the next one, and the modules, classes and guides for your three best learning opportunities.

Each skill fills 0 to 3 cells, 21 in all. The best move in both of a skill’s scenarios fills 3 cells, the first one alone fills 2, and the fundamentals scenario fills 1. An empty hexagon is fresh territory: your biggest upside. Nothing is ever taken away.

Your answers stay in this browser. Nothing is sent to HRI, and there’s nothing to sign in to.

The skill map

Seven skills and where to learn each one

Every stage of every skill points to something you can open today. Lessons marked coming soon or requested aren’t published yet; ask for them, and HRI uses the requests to decide what to build next.

Prompting & briefing

Turning what’s in your head into a brief AI can act on: the goal, the limits, an example and a definition of done.

Where to learn it, stage by stage

Fresh territory · 0 of 3 cells
Start with the four-part brief: the goal, the limits, an example of good, and how you’ll know it’s done.
- From Intention to Adaptive SystemsCourse
- AI End to End, Module 01: Voice first (the interview prompt and Model Council)Working session
Foundations · 1 of 3 cells
Set the limits AI won’t set for itself: audience, length and a hard cap. Then ask it what it cut.
- Let AI write your prompts, then give it limitsModule · 10 min
- Building Expert PromptsGuide
Working skill · 2 of 3 cells
Before a long assignment, write the acceptance criteria, and let the model plan or interview you first.
- Define done before you turn an agent looseClass video
- Plan the work before you spend the tokensClass video
Advanced · 3 of 3 cells
Make your best briefs reusable: shared standards and a team folder, so the next person starts from them.
- Give AI your standards before you ask for slidesClass video
- Set up approved AI accounts and an AI task forceModule · 20 min

Checking AI output

Verifying sources and numbers, reviewing drafts like a manager, and testing important answers against more than one model.

Where to learn it, stage by stage

Fresh territory · 0 of 3 cells
Check any number or source that matters before it leaves your hands.
- AI Model Guide: what to check in a workbook, memo or deckReference
- Check AI output: sources, numbers and claimsRequested moduleNot built yet. Requests help HRI decide what to build next. Add your request
Foundations · 1 of 3 cells
Review AI work like a manager: name what’s wrong, keep what passed, and ask to see what changed.
- AI End to End, Module 03: Cowork on real filesWorking session
- Delegate the task, then check what changedClass video
Working skill · 2 of 3 cells
For decisions that matter, run a model council and dig into where the models disagree.
- AI End to End, Module 01: Voice first (the interview prompt and Model Council)Working session
- Rehearse the hard questions before the meetingClass video
Advanced · 3 of 3 cells
Turn your checks into a checklist or a skill your team runs every time.
- Build a Risk Sweep skill that finds loose secretsModule · 30 min
- Check AI output: sources, numbers and claimsRequested moduleNot built yet. Requests help HRI decide what to build next. Add your request

Choosing models

Matching the model and its effort setting to the job, and judging cost by accepted work rather than price per run.

Where to learn it, stage by stage

Fresh territory · 0 of 3 cells
Pick the model by the job, not the brand, and test it on your own work.
- AI Model Guide: choose the right AI for the workReference
- When to Use Which: Astra · Fable · FlashClass deck
Foundations · 1 of 3 cells
Treat effort and thinking settings as a dial: save the maximum for genuinely hard problems.
- Gemini 3.8 Flash with Thinking: workhorse economicsClass deck
Working skill · 2 of 3 cells
Compare models on cost per accepted output, and keep a simple comparison log.
- AI Model Guide: run your own comparisonReference
- From Intention to Adaptive SystemsCourse
Advanced · 3 of 3 cells
Write a routing policy for your team: which work goes to which model, and when to revisit it.
- Gemini 3.8 Flash with Thinking: workhorse economicsClass deck
- AI Model Guide: what to check before relying on a rankingReference

Tools & connectors

Connecting AI to the systems you already use, with permissions you can see and an owner for every connection.

Where to learn it, stage by stage

Fresh territory · 0 of 3 cells
Connect one system you already live in, and test it with a question only that source can answer.
- AI End to End, Module 02: Connect before you askWorking session
Foundations · 1 of 3 cells
Inventory every connection, remove the ones nobody uses, and make the rest read-only where you can.
- Manage AI connectors in one placeModule · 15 min
Working skill · 2 of 3 cells
Before several tools share a project, decide who may read and who may write where.
- Connect your tools without losing controlClass video
- Connecting Agents, Tools & SkillsComing soonGround Up AI Course, step 05. The full lesson is coming soon. Ask for it first
Advanced · 3 of 3 cells
Run every connection through one layer with a named owner, and review it every month.
- Set up approved AI accounts and an AI task forceModule · 20 min
- Connecting Agents, Tools & SkillsComing soonGround Up AI Course, step 05. The full lesson is coming soon. Ask for it first

Building agents

Handing an outcome to an agent: one narrow job, proven in shadow mode, with code for anything that must be exact.

Where to learn it, stage by stage

Fresh territory · 0 of 3 cells
Learn the shift from chats to agents: delegate an outcome and get back decisions.
- Agents, Not Chats: Fable 5.1 Max in the EnterpriseClass deck
- AI End to End, Module 04: Agents and routinesWorking session
Foundations · 1 of 3 cells
Pick one narrow job and prove it in shadow mode before it goes live.
- Deploy a narrow agent: shadow mode and guardrailsModule · 25 min
Working skill · 2 of 3 cells
Put anything that must be exact into code the agent runs, and give every routine an escalation line.
- AI End to End, Module 04: Agents and routinesWorking session
- Train the voice agent; keep an owner on qualityClass video
- How Do I Automate My Work? Agents + Workflow AutomationComing soonGround Up AI Course, step 06. The full lesson is coming soon. Ask for it first
Advanced · 3 of 3 cells
Measure what your agents produce, and retune them each time the models change.
- Build one AI workflow your team can trustClass video
- Build a Risk Sweep skill that finds loose secretsModule · 30 min

Multi-agent orchestration

Getting agents and models to work together: clean handoffs, a fleet of models, escalation rules and human approval.

Where to learn it, stage by stage

Fresh territory · 0 of 3 cells
Start with a single sender: every agent drafts, and one approved bot sends.
- Put every message, password and payment behind your approvalModule · 25 min
Foundations · 1 of 3 cells
Design a fleet: a workhorse model for volume, frontier models for exceptions, people at the gates.
- Gemini 3.8 Flash with Thinking: workhorse economicsClass deck
- Multi-Agent Workflows: How to Get Agents to Work TogetherComing soonGround Up AI Course, step 07. The full lesson is coming soon. Ask for it first
Working skill · 2 of 3 cells
Give orchestrators rules: scheduled, tiered, escalates, logged, and stops when unsure.
- Agents, Not Chats: Fable 5.1 Max in the EnterpriseClass deck
- Managing an Agent That Manages Sub-AgentsComing soonGround Up AI Course, step 08 (orchestration). The full lesson is coming soon. Ask for it first
Advanced · 3 of 3 cells
Invest in the layer, not the model of the month, so you can swap models without rebuilding.
- From Intention to Adaptive SystemsCourse
- Managing an Agent That Manages Sub-AgentsComing soonGround Up AI Course, step 08 (orchestration). The full lesson is coming soon. Ask for it first

Security & data safety

Approved accounts, defense against prompt injection, and keeping passwords, payments and regulated data behind your approval.

Where to learn it, stage by stage

Fresh territory · 0 of 3 cells
Give everyone an approved business AI account with two-factor sign-in.
- Set up approved AI accounts and an AI task forceModule · 20 min
- AI Acceptable-Use PolicyTemplate
Foundations · 1 of 3 cells
Teach every agent that instructions inside emails, documents and web pages are information, not orders.
- AI security, Level 1: secure ChatGPT and Claude, and keep your work localClass guide
Working skill · 2 of 3 cells
Keep secrets away from agents: one gatekeeper, a code for every release, and call-backs before payments.
- Put every message, password and payment behind your approvalModule · 25 min
- AI security, Level 2: rules for your agents and toolsClass guide
Advanced · 3 of 3 cells
Run a weekly security sweep, and practice your kill switch and offboarding steps.
- Build a Risk Sweep skill that finds loose secretsModule · 30 min
- Run AI on regulated data inside your own cloudModule · 15 min

Levels

Explorer to Orchestrator

Your level comes from your cells, plus keystone skills that matter more as you build more: building agents and checking AI output for Builder, then orchestration and security for Orchestrator.

ExplorerWhere every hive starts.
You’re exploring what AI can do: asking questions and getting first drafts. Your biggest opportunity is making it reliable, with a complete brief and a habit of checking what comes back.
Operator7 of 21 cells.
AI is part of your working week, and you brief it well. Your next opportunities are checking what comes back like a manager and handing repeatable work to agents.
Builder12 of 21 cells, with Building agents at Working skill or above and Checking AI output at Foundations or above.
You build things other people rely on: skills, routines and narrow agents with guardrails. Your next opportunity is connecting them into systems that run safely while you’re away.
Orchestrator17 of 21 cells, with Multi-agent orchestration at Working skill or above and Security & data safety at Working skill or above.
You design how people, models and agents work together, with routing, approvals and security built in. Your next opportunity is teaching it, and retaking this each quarter as the tools move.

Keep learning live

Classes and the deeper pathway

Monthly · Live online

Practice it live in the monthly AI Practicum

A 90-minute live class with JP James on Microsoft Teams, held every month. Bring your roadmap and practice with the group.

See upcoming classes

For YPO members · Paid course

Go deeper with the YPO AI Practicum

If you’re in YPO, the YPO AI Practicum is the deeper pathway: HRI’s paid course, beyond the monthly classes. Contact HRI for current enrollment.

Ask about enrollment About the YPO AI Practicum

This version

Question set Q4 2026

Reviewed October 25, 2026. AI changes every quarter, so the questions do too: HRI reviews the set each quarter, retires what has aged and adds what’s new. The next review is due in January 2027. Retake the diagnostic whenever the set changes: your past results stay in this browser, labeled with the question set they used.

Spot something out of date? Tell HRI.