Opportunity Identification
Context
Section titled “Context”Your leadership team has decided: “We’re doing AI now.” The roadmap has room for 2-3 AI initiatives. Engineering suggests an internal coding assistant. Sales wants an AI chatbot. Support wants automated ticket classification. All three sound reasonable.
Your job as a PM: Don’t build what’s demanded loudest — find what creates the most value.
Concept
Section titled “Concept”Where AI creates the most value
Section titled “Where AI creates the most value”McKinsey estimates Generative AI’s annual value creation potential at $2.6-4.4 trillion. 75% of that value concentrates in four areas:
| Area | Value driver | Example |
|---|---|---|
| Customer Operations | Automation, faster resolution | Klarna: 2.3M chats automated |
| Marketing & Sales | Personalization, content generation | Individual campaigns at scale |
| Software Engineering | Code generation, testing, reviews | GitHub Copilot, internal tools |
| R&D | Accelerated research, prototyping | Drug discovery, material science |
Top industries: Banking, High Tech, Life Sciences. But these are averages — your opportunity depends on your context.
Three value levers
Section titled “Three value levers”Every AI opportunity leverages at least one of these:
- Automation: Eliminate repetitive tasks. Klarna automated 2.3 million support chats — work previously done by humans.
- Prediction: Spot patterns before they happen. Churn prediction, demand forecasting, anomaly detection.
- Personalization: Individual experiences at scale. Duolingo adapts learning content to each user in real time.
The biggest value comes from combining all three. A recommendation system that predicts preferences (Prediction), curates automatically (Automation), and adapts individually (Personalization) beats any single dimension alone.
Pain severity: The five critical questions
Section titled “Pain severity: The five critical questions”Before you prioritize an opportunity, quantify the pain:
- How frequently does the problem occur? (Daily > Monthly > Rarely)
- How severe is the pain? (Business-critical > Annoying > Cosmetic)
- How many users are affected?
- Can AI solve it significantly better than the status quo? The bar is 10x, not 10%.
- Is the required data available and of sufficient quality?
Question 4 is the one most often ignored. Marginal improvements don’t justify AI complexity.
Framework
Section titled “Framework”RICE-A: RICE with AI Complexity (based on Dr. Marily Nika)
The classic RICE framework extended with an AI Complexity factor:
Formula: (Reach x Impact x Confidence) / (Effort + AI Complexity x 0.5)
| Factor | Question | Scale |
|---|---|---|
| Reach | How many users/month? | Absolute number |
| Impact | How much does it improve the outcome? | 0.25 / 0.5 / 1 / 2 / 3 |
| Confidence | How reliable are the estimates? | 50% / 80% / 100% |
| Effort | Team effort in person-months? | Person-months |
| AI Complexity | Data quality, model risk, eval effort? | 1 (low) – 5 (high) |
AI Complexity is weighted at 0.5 and added to Effort — because AI-specific risks increase effective effort, but aren’t directly comparable to traditional engineering effort.
Scenario
Section titled “Scenario”You’re a PM at a B2B SaaS with 15,000 paying customers. Three AI opportunities are on the table:
| Opportunity A | Opportunity B | Opportunity C | |
|---|---|---|---|
| What | AI chatbot for support | Churn prediction + alerts | Internal code assistant |
| Reach | 8,000 users/month | 15,000 customers | 4,500 internal requests/month |
| Impact | 1 (medium) | 3 (massive) | 2 (high) |
| Confidence | 80% | 50% | 80% |
| Effort | 3 PM | 4 PM | 2 PM |
| AI Complexity | 2 | 4 | 1 |
| RICE-A Score | 1,600 | 3,750 | 2,880 |
Opportunity A sounds easiest. Opportunity C has a strong score with low effort. Opportunity B scores highest — but with only 50% confidence.
Decide
Section titled “Decide”How would you decide?
The best decision: Start Opportunity C, validate Opportunity B in parallel.
Why:
- Opportunity C (Code assistant): High score, low complexity, 80% confidence. Reach measures internal requests (4,500/month from 45 engineers), not external users — but internal impact is immediately measurable. Follows the Shopify logic: “Before asking for more headcount, show why AI can’t solve it.”
- Opportunity B (Churn prediction): Highest score, but 50% confidence. Start with a 4-week data readiness check: Do your data actually contain churn signals? If yes, this becomes your biggest lever.
- Opportunity A (Support chatbot): Lowest score. Support chatbots are a commodity now — hard to use as differentiation.
The most common mistake: Choosing the most visible opportunity (chatbot) because stakeholders “get it.” Instead: internal tooling and prediction models often deliver the highest ROI.
Reflect
Section titled “Reflect”- 75% of AI value sits in four areas. Customer Operations, Marketing & Sales, Software Engineering, R&D. Start there, not with “cool-sounding” features.
- 10x is the bar, not 10%. Notion AI grew from $67M to $500M+ revenue because it was integrated into the workflow — not bolted on as a marginal feature. Perplexity hit $148M ARR with one clear pain point: search with real sources.
- Don’t overlook internal opportunities. Shopify’s CEO made it policy: before teams ask for more headcount, they must show why AI can’t get it done.
- Data readiness before feature readiness. The best opportunity is worthless if the required data doesn’t exist or is garbage.
Sources: McKinsey “The Economic Potential of Generative AI” (2023), Dr. Marily Nika “RICE-A Framework” (2024), Shopify CEO Memo (April 2025), Klarna “AI Assistant Report” (2024), Notion AI Revenue (2025), Perplexity Growth Metrics (2025)