Building AI Intuition Through Hands-On Examples
60-Minute Intro Workshop
Help PMs build AI intuition through hands-on examples, then apply that intuition to your product decisions.
| Section | Time | What We Cover |
|---|---|---|
| Mental Models | 10 min | Two modes: "Thinking Partner" + "Task Automator" |
| Live Demo | 30 min | End-to-end PM workflow: idea → research → PRD → user stories → Jira |
| Building AI Products | 10 min | Interface patterns, configuration, context vs concealment |
| Closing + Q&A | 10 min | Toolkit summary, next steps |
The Two Modes of AI
Brainstorms, researches,
structures ideas, challenges assumptions
Interview synthesis, competitive analysis, feature ideation
Executes across tools:
reads, writes, updates, queries
Pull from SharePoint, update Jira, create ADO work items
The shift: AI isn't just a conversation partner — it's an over eager junior team member with very little common sense and crazy memory that can read, write, and act on your behalf.
Context — Background information the AI needs
Role — Who should the AI be? (analyst, writer, critic)
Ask — The specific request
Format — How you want the output structured
Tone — Voice and style guidelines
Context: I'm a PM at a retail POS company. We're evaluating whether to add AI-powered inventory recommendations.
Role: Act as a senior product strategist who has shipped AI features at enterprise SaaS companies.
Ask: Identify the top 3 risks of this feature and how to mitigate each.
Format: Bullet points with risk, likelihood, impact, and mitigation for each.
Tone: Direct and practical, no fluff.
GDPVal is OpenAI's benchmark for evaluating how well AI models can answer real-world GDP and economics questions — demonstrating the gap between "can chat about it" vs "can actually do the task."
Click these links during the presentation to show real examples.
Let's go through 8 common PM workflows or user journeys, see how we transform them with AI and build some intuition.
| # | User Story | With AI | Tool |
|---|---|---|---|
| 1 | Research strategy for a new feature — Explore market trends, competitive landscape, best practices | AI runs async research across dozens of sources, delivers structured report with citations | ChatGPT Deep Research |
| 2 | Synthesize customer interviews — Extract themes, contradictions, key quotes | Upload transcripts, AI identifies patterns, surfaces contradictions, pulls quotes | Claude.ai Projects |
| 3 | Generate and refine a PRD — Draft requirements that translate needs into specs | AI generates structured PRD, critiques its own output, identifies gaps and risks | Claude Desktop |
| 4 | Design prototypes from requirements — Turn specs into visual artifacts | Describe in natural language, AI generates working UI prototype in minutes | Lovable |
| 5 | Publish artifacts to team systems — Get PRDs into Confluence, Slack, etc. | AI publishes to Confluence, posts to Slack channels — one command | Claude Desktop + MCPs |
| 6 | Convert PRD into user stories — Break down into work items with acceptance criteria | AI extracts stories, writes acceptance criteria, maintains links to source | Claude Desktop |
| 7 | Triage stakeholder feedback — Process comments, prioritize, formulate responses | AI reads comments, categorizes by urgency/theme, helps formulate and post replies | Claude Desktop + MCPs |
| 8 | Create work items in Jira/ADO — Push approved stories to project management tools | AI creates stories with proper fields, epic links, acceptance criteria — directly | Claude Desktop + MCPs |
A Skill is a way to teach AI how to do something specific and reusable — packaging your expertise into a format the AI can apply automatically.
Open Standard: Skills are emerging as a portable format across LLMs and agent systems (Claude, GPT, Gemini, open-source agents). Write once, use anywhere — your encoded expertise isn't locked to one vendor.
| Without Skills | With Skills |
|---|---|
| Re-explain your format each session | Format encoded once, applied always |
| Quality varies by who's prompting | Consistent output for everyone |
| New team members start from scratch | Expertise transfers instantly |
Idea to Jira (and prototype if you want!) in One Session
Kick off deep research for strategy & market fit
ChatGPT Deep ResearchLoad transcripts, find themes & pain points
Claude.ai ProjectsDraft structured requirements from insights
Claude DesktopGenerate visual prototype from the PRD
LovableGet a second opinion, identify gaps
Claude DesktopPush to Confluence, notify via Slack
MCPsExtract user stories with acceptance criteria
Claude DesktopPush stories for stakeholder review
MCPsRead comments, prioritize, post replies
MCPsPush approved stories as work items
Jira MCPGray cards are optional steps
This doesn't have to stop here. If you work collaboratively with your dev team:
🔗 The thread from "why are we building this?" to "how is it built?" stays intact.
Click a slide to see notes...