A structured AI prompting approach for generating distinct solution perspectives across multiple disciplines. Rather than asking for a single answer or receiving minor variations of the same idea, this pattern forces the AI to sample broadly and surface unconventional approaches — including technical, financial, human-centric, strategic, and asymmetric angles.
Related: [1] · [2]
A simple prompt like "how do I fix my stagnating business?" tends to produce the most probable, most generic answer — the same answer every other user is getting. This is what Mark Hope calls "AI slop": output that is technically correct but competitively useless because it's commoditized by default.
The Problem Solving Pattern breaks this by explicitly demanding variety, multi-disciplinary framing, and probability scoring in a single prompt.
I'm facing this problem: [describe the specific situation].
I need five different ways to approach or solve this.
Please generate five distinct solution perspectives,
each from a different angle or discipline.
For each solution:
- Name the angle or discipline
- Describe the core approach
- Assign a probability score (0–1) showing how obvious
or unexpected this solution is
Key constraints to include:
- five distinct — triggers [1], preventing minor variations
- each from a different angle or discipline — forces multi-disciplinary spread
- assign a probability — enables [2] so you can identify and discard obvious answers
Prompt used in the session:
"My automotive stereo sound and safety equipment business is stagnating. I need five different ways to approach or solve this. Please generate five distinct solution perspectives, each from a different angle or discipline. For each solution, assign a probability to show how obvious or unexpected it is."
Results:
| Angle | Core Idea | Probability |
|---|---|---|
| Technical Innovation | Pivot to AI-powered, OTA-updating audio systems — "the Tesla of aftermarket" | Low |
| Financial Engineering | Shift from one-time sales to Equipment-as-a-Service recurring revenue | Moderate |
| Human-Centric Design | Stop selling equipment; sell "automotive sanctuaries" — sound sommelier consultations, stress-reduction acoustic packages | Moderate-Low |
| Strategic Repositioning | Exit saturated consumer market; become exclusive supplier to autonomous vehicle fleets and ridesharing services | Low |
| Unconventional / Asymmetric | Become a Trojan Horse Data Company — installed equipment as IoT sensors gathering anonymous driving and acoustic data | Very Low |
The final option (Trojan Horse Data Company) is the kind of idea that would never surface from a simple prompt. It emerged because the pattern explicitly demanded an "unconventional" discipline.
When you want to ensure genuine variety, you can name the disciplines explicitly or let the AI choose. Common productive axes include:
⚠️ Mark Hope notes: "You don't have to give these examples if you don't want, because they can be leading — it may give you exactly what you say here. Sometimes it's safer to say just 'each from a different angle.'"
After the initial five solutions, you can push further into unconventional territory:
Now generate five more solutions, sampling from the tail of the
distribution — prioritizing responses with probabilities below 0.1.
This is most useful when:
- The client's market is commoditized and conventional differentiation won't work
- You need to stimulate creative thinking, not find an immediately deployable answer
- You're preparing for a client brainstorm and want provocative starting points
See [2] for full tail-sampling guidance.
| Situation | Use This Pattern? |
|---|---|
| Client business is stagnating or commoditized | ✅ Yes |
| Need to present multiple strategic options | ✅ Yes |
| Looking for a single definitive answer | ❌ No — use a direct prompt |
| Generating ad copy or creative assets | ⚠️ Partial — see [3] |
| Predicting future market scenarios | ⚠️ Partial — combine with scenario framing |
Demonstrated by Mark Hope in the internal AI training session "Using AI Part 2: Verbalized Sampling and Probability Control" (2025-11-13). Prompt templates from this session were distributed to the team as a follow-up document.