AI Agents and Prompt Chaining for PMs
So far in this course, you've used AI for individual tasks — one prompt, one output. But the most powerful PM workflows chain multiple AI steps together, where the output of one prompt feeds into the next. This is prompt chaining, and it's the foundation of AI agents that can handle complex, multi-step product workflows.
What You'll Learn
- What prompt chaining is and why it matters for product managers
- How to build multi-step AI workflows for complex PM tasks
- Introduction to AI agents and no-code automation tools
- Four ready-to-use prompt chains for product management
What Is Prompt Chaining?
Prompt chaining means breaking a complex task into sequential steps where each prompt builds on the output of the previous one. Instead of asking AI to do everything at once (which produces shallow results), you guide it through a structured process.
Single prompt (shallow):
"Create a complete product strategy for our mobile app."
Prompt chain (deep):
- Analyze our current product metrics and identify the top 3 underperforming areas
- For the #1 underperforming area, research what competitors are doing
- Based on the competitive analysis, generate 5 feature hypotheses
- Score each hypothesis using RICE framework
- For the top-scored hypothesis, draft a PRD
Each step produces better output because it has more focused context from the previous step.
Prompt Chain #1: From User Feedback to Feature Spec
This chain turns raw user feedback into a ready-to-build feature specification:
Step 1: Categorize and Theme
Categorize this user feedback into themes. For each theme,
count occurrences and rate intensity (1-5):
[paste 20-50 pieces of feedback]
Step 2: Prioritize Themes
Based on these themes from our user feedback analysis:
[paste Step 1 output]
Rank them by:
- Frequency (how many users mentioned this)
- Intensity (how strongly they feel)
- Business impact (connection to our metrics: [list metrics])
Recommend the top theme we should address first and explain why.
Step 3: Define the Problem
Based on this analysis, the top user problem is:
[paste Step 2 recommendation]
Write a problem statement that:
- Quantifies the problem with data from the feedback
- Describes the user impact in specific terms
- Explains why solving this matters now
- Includes 3 representative user quotes
Step 4: Generate Solutions
Given this problem statement:
[paste Step 3 output]
Generate 5 solution approaches ranging from:
- Quick fix (< 1 sprint)
- Moderate investment (1-2 sprints)
- Strategic solution (3+ sprints)
For each, estimate: effort, impact, risk, and time to value.
Step 5: Draft the Spec
We're going with solution [X] from the analysis above:
[paste chosen solution]
Draft a complete PRD following our template:
[paste template sections]
Include acceptance criteria in Given/When/Then format for
every requirement.
Prompt Chain #2: Competitive Response Planning
When a competitor makes a major move, use this chain to formulate your response:
Step 1: Analyze the Move
[Competitor X] just announced [what they announced].
Analyze:
- What specific capabilities does this add?
- Who does this target? (their users, our users, new market?)
- What's the strategic intent behind this move?
Step 2: Assess Impact on Us
Based on this competitive analysis:
[paste Step 1 output]
How does this affect our product?
- Which of our user segments are most at risk?
- Does this change our competitive positioning?
- What questions will our sales team get?
- What questions will our customers ask?
Step 3: Formulate Response Options
Given the competitive impact analysis:
[paste Step 2 output]
Generate 4 response options:
1. Do nothing (and why this might be okay)
2. Communicate existing strengths (messaging response)
3. Accelerate a planned feature (which one and why)
4. Build something new (what and rough scope)
For each option, assess: timeline, effort, risk, and impact.
Step 4: Create Action Plan
We're going with option [X]:
[paste chosen option]
Create an action plan with:
- Immediate actions (this week)
- Short-term actions (this month)
- Medium-term actions (this quarter)
- Owner for each action
- Success metrics
Prompt Chain #3: Quarterly Planning
Step 1: Review Last Quarter
Here are our Q[X] results against OKRs:
[paste OKR scorecard]
And our key product metrics:
[paste metrics]
Analyze: What worked, what didn't, and what should we
carry forward vs. abandon?
Step 2: Identify Opportunities
Based on last quarter's review:
[paste Step 1 output]
Plus these inputs:
- User research themes: [list]
- Competitive landscape changes: [list]
- Business priorities from leadership: [list]
Identify the top 5 opportunities for next quarter.
For each: describe it, estimate impact, and note risks.
Step 3: Draft OKRs
Based on these opportunities:
[paste Step 2 output]
Draft 3-4 OKRs for Q[X+1]:
- Each Objective should be inspiring and outcome-focused
- Each Key Result should be measurable with a specific target
- Include 2-3 Key Results per Objective
- Ensure OKRs are ambitious but achievable (70% target)
Step 4: Create the Roadmap
Based on these OKRs:
[paste Step 3 output]
Create a quarterly roadmap with:
- Monthly breakdown
- Feature/initiative mapping to each OKR
- Dependencies and risks
- Resource allocation
Introduction to AI Agents
AI agents take prompt chaining further by automating the flow between steps. Instead of manually copying output from one prompt to the next, agents handle the orchestration.
No-Code Agent Tools for PMs
Make (formerly Integromat) and Zapier let you build automated workflows that connect AI to your tools:
- Trigger: New feedback in Intercom/Zendesk
- Step 1: AI categorizes the feedback
- Step 2: AI assesses severity and product area
- Step 3: Auto-create a tagged ticket in your backlog
- Step 4: If severity is high, send a Slack alert to the PM
ChatGPT Custom GPTs with Actions can connect to external APIs:
- Pull data from your analytics tool
- Create tickets in Jira or Linear
- Send Slack messages
- Update spreadsheets
When to Automate vs. When to Stay Manual
Automate when:
- The task is repetitive and follows the same pattern
- The stakes are low (internal drafts, categorization, summaries)
- Volume justifies the setup time
Stay manual when:
- The task requires product judgment at each step
- The output goes directly to stakeholders or customers
- The context changes significantly each time
Key Takeaways
- Prompt chaining produces dramatically better results than single prompts by breaking complex tasks into focused steps
- Build chains for your most common multi-step workflows: feedback-to-spec, competitive response, and quarterly planning
- No-code tools like Make and Zapier can automate prompt chains by connecting AI to your product tools
- Automate repetitive, low-stakes tasks; keep manual control over strategic decisions and stakeholder-facing output
- Start with manual prompt chains before investing in automation — prove the workflow first, then optimize it