Profitable Problem Discovery

I’m an “AI guy” so I get it.

But…it’s bad business.

Simply adding "AI" to a business description has become the default move, as if that alone guarantees success.

We’ve been here before in the late 1990s/early 2000s when everyone added .com to their name to be an “internet company”.

Remember Pets.com?

Their sock puppet mascot got a Macy’s parade float, a super bowl advert and was featured in TIME.

All based on hype about the internet, which only had about 100 million users at the time. It all came crashing down 2 years later, along with the entire stock market.

We may be seeing something similar with “AI” being slapped on every business, without anyone knowing what that actually means. 

A few months ago I was chatting with someone who was developing an "AI-powered content management system." When I asked him what specific problem it solved that existing systems didn't, he struggled to articulate an answer. He had fallen in love with the technology, not the problem.

The harsh reality? Just adding AI to something doesn't make it a good business idea. Not even close.

What does then?

Let's get started:

Business = Problem-Solving

Instead of chasing AI for its own sake, we need to return to fundamental business principles. At its core, every successful business solves a problem that someone is willing to pay to have solved. Full stop. Period. End of sentence.

This is true across all industries and price points.

Even luxury items solve problems – the need for status, exclusivity, and social signalling. A £10,000 watch solves the problem of how to demonstrate success and taste to certain people.

There's a direct relationship between your business success and:

1. How painful the problem is

2. How many people have this problem

3. How effectively you solve it

Peter Drucker put it thus: "The size of the problem you solve determines the size of the opportunity." The most successful businesses tackle the most significant problems.

Let’s turn to another heavy hitter. Peter Thiel: "All happy companies are different: each one earns a monopoly by solving a unique problem. All failed companies are the same: they failed to escape competition."

This is crucial for AI entrepreneurs. Your competitive moat won't be the AI itself (which is increasingly commoditised), but the specific, painful problems you solve with it.

Hell, our customers likely won’t care that we are using AI to solve their problem. They just want the solution!

This will be our guiding principle throughout this Playbook.

The Value-Problem Connection

There's a direct relationship between the severity of a problem and how much people will pay to solve it.

Think about it this way:

• A minor annoyance might be worth a few dollars to fix

• A significant time-waster could be worth hundreds

• A business-threatening issue might be worth thousands or more

This is why starting with problems rather than technologies is so crucial. When you solve a genuine, painful problem, pricing becomes much easier - the value is self-evident to your customers.

Because I’m on a roll (and believe in the Rule of Threes!) Paul Graham once said, "The best startup ideas tend to have three things in common: they're something the founders themselves want, that they themselves can build, and that few others realise are worth doing."

OK that’s the last older white dude I’ll be quoting. Promise.

That first part is key “they're something the founders themselves want” - solving a problem you personally understand gives you tremendous insight and advantage.

The Proximity Principle

One of the biggest myths in entrepreneurship is that great business ideas come from mysterious "Eureka!" moments. Research actually shows that successful founders typically build businesses in areas they're deeply familiar with.

Why? Because proximity to problems gives you:

• A visceral understanding of the pain points

• Knowledge of existing solutions and their shortcomings

• Credibility with potential customers

• A built-in network for early feedback and sales

When I teach my AI Audience Accelerator I tell people to focus on educating their industry, their niche, their area of expertise. Ditto for when I license my workshop material for people to go and teach in businesses. Which businesses should they focus on? Those in their industry.

It’s a massive comparative advantage and shouldn’t be wasted. This is especially true when it comes to AI solutions. Generic AI applications rarely gain traction.

If it’s all things for all people no-one will want it. Hell they’ll just use ChatGPT most likely as it is truly all things for all people! Want to compete with OpenAI? No…me neither!

Instead if we build deeply contextualised solutions - the ones that scratch a specific itch - that’s where we can deliver real value. And make real money.

Mining Your Personal Experience

Enough fluff. Enough quotes. I had enough of that during my MBA.

Let's get practical. Your own experience is the richest, most accessible source of problem insights.

There are a tonne of methods for extracting these from you. You could just be mindful and note things down as they come to you.

But honest self-reflection can be challenging – we're often blind to our own inefficiencies or accept frustrations as "just the way things are." To overcome this, I've created several prompts that can help you extract business problems through guided introspection.

These prompts focus specifically on business activities, not personal ones. For even better results, use them with colleagues or others in your industry who might spot problems you've grown accustomed to.

Daily Friction Interview Prompt:

I want you to interview me to identify business problems in my daily work that could be solved with AI. Ask me questions one at a time about:

1. Repetitive tasks I perform regularly
2. Processes that take longer than they should
3. Information I frequently need to search for or organize
4. Decisions I make repeatedly using similar criteria
5. Points of frustration in my typical workday
6. Software or tools I use that have limitations
7. Manual work I do that seems like it could be automated

For each answer I provide, ask 1-2 follow-up questions to dig deeper into the problem before moving to the next topic. After the interview, summarise the potential business problems identified, ranking them by apparent pain level and frequency.

Time Audit Analysis Prompt:

I'm going to share my activities for [period of time]. For each activity, I'll include:
- What I was doing
- How long it took
- How frequently I do this task

Based on this information, please:
1. Identify tasks that appear to consume disproportionate time
2. Flag repetitive activities that might be candidates for automation
3. Detect patterns in how I spend my time
4. Suggest 3-5 potential business problems worth solving based on this time audit
5. For each problem, briefly describe how AI might help address it

Cost Center Analysis Prompt:

I'll share my recent business expenses. For each expense, please:
1. Identify what problem I'm paying to solve
2. Evaluate if this solution seems efficient or optimal
3. Consider if AI could provide a better/cheaper solution
4. Rate how painful/important this problem seems (based on the amount spent)

After analysing all expenses, suggest 3-5 business problems that appear most significant based on my spending patterns, and briefly outline how AI might address them more effectively.

Decision Fatigue Identification Prompt:

Help me identify decision-making processes in my work that cause friction or fatigue. Ask me questions about:
1. Decisions I make repeatedly throughout my workday
2. Choices that require gathering the same types of information
3. Decisions that follow predictable patterns or rules
4. Areas where I find myself procrastinating due to decision complexity
5. Decisions where I've made costly errors in the past

For each area I describe, probe deeper with 2-3 follow-up questions to understand the full context, information requirements, and pain points of the decision process. Then summarise the key decision-related problems that could potentially be solved with AI assistance.

These are just to get you started. Ultimately you know best what problems people in your industry face. These prompts are just ways to kick start the process of getting them out on to the page.

For now set up a Problem Bank - a note on your phone or wherever you can easily access and edit. Make this as frictionless as possible - don’t start setting up complex systems. That’s procrastination creeping in.

Also, don't worry about evaluation or filtering yet - we'll cover that in later parts. Right now this is brain dump.

What's Next?

In this Playbook I’m taking you through a comprehensive framework for finding and validating business problems worth solving with AI:

Part 1: The Problem-First Mindset - We've covered why starting with problems rather than technologies is crucial, and explored techniques for mining your personal experience (this Part).

Part 2: Problem Discovery Beyond Personal Experience - We'll expand our problem discovery to market observation and data analysis, adding even more opportunities to your Problem Bank.

Part 3: Problem Categorisation & Segmentation - You'll learn how to organise and prioritise problems to identify patterns and high-value clusters using AI-assisted analysis.

Part 4: Problem Validation & Verification - We'll explore methods to verify that problems are real, painful, and worth solving before investing in building solutions.

Part 5: Building Problem-Solving AI Businesses - Finally, we'll look at how to turn validated problems into cohesive business offerings matched to the right technical approaches.

In short we’re going to first explore and long list as many ideas as possible. Then we’re going to start the filtering and validation process to come up with the specific problems we can build a business around.

Keep Prompting,

Kyle

Or learned a new word and then heard it three times that week?

That's the Baader-Meinhof phenomenon—once something enters your awareness, you start noticing it everywhere.

The same thing happens with business problems once you train yourself to spot them. Last week, I was having coffee with a fellow entrepreneur who mentioned—almost as an afterthought—how his team wastes nearly half a day every week manually compiling client usage reports. "It's just part of the job," he shrugged.

But is it? Within ten minutes, we'd sketched out an AI automation that could save his team 20+ hours monthly. Yes…I’m just that cool, I know I know!

What he had accepted as an inevitable business friction was actually a solvable problem hiding in plain sight.

Here's the thing: Once you become attuned to identifying business problems, you'll spot them everywhere—mentioned in passing during meetings, buried in customer feedback, or hiding in the "minor annoyances" people have simply accepted.

Let's get started:

Beyond Personal Perspective

In Part 1, we explored how to dig into your personal experience for business problems. This is a fantastic starting point, but it has obvious limitations – you only experience a tiny fraction of the problems that exist in the market.

For better or worse you are just you! It’s a great starting point (and anchor to the industry at large) but to build a more comprehensive Problem Bank, we need to systematically explore beyond our own experience.

Sounds like I’m about to hand you some strange looking mushrooms. Don’t worry. Nothing that exciting!

I’m going to run through a bunch of techniques. And the name of the game is to go and collect as much as possible. Unstructured, messy data. That’s fine.

We’re going to take all of that noise and use a prompt to extract what we need. Let’s get to it.

1. Community Listening

Online communities are goldmines of problem statements. People gather specifically to discuss challenges, seek solutions, and vent frustrations.

Focus on:

• Industry-specific subreddits or forums

• Facebook or LinkedIn groups for professionals in your target market

• Discord/Slack communities

• Specialized industry forums

For each community, search for terms like:

• "frustrated with"

• "problem with"

• "alternative to"

• "help with"

• "how do I"

• "anyone know how to"

Reddit is my favourite for this personally. When I find a related discussion I will wholesale copy paste it all (length dependant obviously) and stick it in a Notion document. Or alternatively save the whole discussion to Raindrop.io.

Why get it all? Because I want to see the discussion and the language used. Not just what is being discussed but how.

2. Review Mining

Product and service reviews are essentially structured problem statements. When people complain about a product or service, they're highlighting an unmet need or inadequate solution.

Analyse reviews on:

• G2, Capterra, Cluch, or TrustPilot (for B2B software and services)

• App stores (for software solutions)

• Professional service review sites (for consulting or agency services)

• Amazon (for books and products in your industry)

For example, looking at reviews for Xero accounting software might reveal frustrations with its reporting capabilities or integration limitations—each representing a potential opportunity for an AI solution.

Look specifically for:

• Common complaints across multiple reviews

• Features users wish existed

• Workarounds people have developed

• Areas where even positive reviews mention limitations

Again though copy and paste it all. We’ll use AI to filter later.

3. Complaint Analysis

Beyond product reviews, people share complaints and frustrations across various platforms. This is a bit “heavier” - ie. BIG problems rather than minor quibbles.

These are direct windows into problems worth solving. Reviews tend to be more emotive and off the cuff. Complaints will generally be more structured. They are harder to find but highly valuable once in hand.

Check out:

• Better Business Bureau complaints (or equivalent in your locale)

• Twitter/X searches for "[company/product/service name] + issue/problem/frustrated"

• Industry watchdog sites

• Consumer advocacy forums

Pay special attention to complaints that:

• Appear repeatedly over time (indicating persistent issues)

• Generate significant engagement from others

• Describe workarounds people have developed

• Mention willingness to switch providers or pay for solutions

Again AI can do a lot of the filtering for you so focus on gathering up first and foremost.

4. Social Listening

Social media platforms can provide real-time insights into emerging problems and pain points. Social listening is a fancy social media marketing agency term for paying attention to what people are chatting about.

Monitor:

• Industry hashtags

• Trending topics in your target market

• LinkedIn posts from thought leaders in your space

• Comments sections on relevant content

Tools like Hootsuite, Mention, or even just Twitter/X's advanced search can help you systematically monitor these conversations.

I've found that the comments sections on "how-to" content are particularly revealing. People often share their specific struggles and ask questions that highlight unmet needs.

5. Search Data

Google is the world’s largest Question and Answer engine. People have problems - they go to Google for the answer.

We can access all of Google's data via Adwords (free) or tools like SEMRush (paid).

As a good mid ground starting point that’s easy to use check out Answer the Public. It will specifically give you the questions that people have. Very very useful resource and you can use it a handful of times a day for free.

Using AI to Process Unstructured Feedback

One of the challenges with market observation is the sheer volume of unstructured data you'll collect. This is where AI can be incredibly powerful—helping you analyse large amounts of feedback to identify patterns and insights that would be difficult to spot manually.

Here's a comprehensive prompt to help you analyse customer feedback, reviews, forum posts, or any other unstructured text data you've collected:

You are an expert business analyst specialising in identifying business problems and opportunities from customer feedback. I'll provide you with a collection of unstructured feedback (reviews, forum posts, comments, etc.). Please analyse this data to help me identify potential business problems worth solving.

The feedback is related to [briefly describe the industry, product, or service].

Please provide a comprehensive analysis including:

1. PROBLEM IDENTIFICATION:
   - List the top 10 distinct problems or pain points mentioned
   - For each problem, provide 2-3 representative quotes from the feedback
   - Rate each problem's severity (1-5 scale) based on language intensity and frequency of mention
   - Estimate how widespread each problem appears to be (percentage of feedback mentioning it)

2. PATTERN ANALYSIS:
   - Identify any correlations between problems (e.g., problems that frequently appear together)
   - Note any patterns related to user types, contexts, or scenarios
   - Highlight any trends in how people are currently trying to solve these problems

3. VISUALIZATION:
   - Create a table ranking the problems by frequency and severity
   - Generate a visual representation of the problem landscape (e.g., concept map showing relationships between problems)
   - If possible, create a simple chart showing the distribution of problem mentions

4. OPPORTUNITY ASSESSMENT:
   - For each top problem, briefly describe how AI might help solve it
   - Rate each problem's suitability for AI solutions (1-5 scale)
   - Identify any gaps or unmet needs that aren't being addressed by current solutions

5. RECOMMENDATIONS:
   - Suggest the 3-5 most promising problems to focus on based on severity, frequency, and suitability for AI solutions
   - For each recommended problem, outline what additional information would be valuable to gather
   - Suggest specific follow-up questions to validate these problems further

Please be objective in your analysis and avoid confirmation bias. Focus on identifying genuine problems rather than validating preconceived ideas.

This prompt turns raw, unstructured feedback into a structured analysis that can directly inform your Problem Bank. It's especially powerful when you've collected dozens or hundreds of pieces of feedback that would be overwhelming to analyse manually.

For best results use a Reasoning model that can take longer to “think through” the data.

If you end up with too much (good job) then first “pre-process” the data by feeding individual methods in and asking for summaries. Then feeding the summaries into this prompt.

What's Next?

The prompt will spit out a handful of potential problems for you to focus on. Exciting!

But hold your horses - first we’ll do some more filtering, sorting and fine detail work to find the best problem. The more pre-production work you do now the better!

In Part 3, we'll tackle the crucial process of categorising and segmenting the problems you've discovered.

Keep Prompting,

Kyle

Well, there's a flip side to that superpower—it's like opening Pandora's box.

Last month, one of my students texted me in a near-panic. "Kyle, I've created a monster," he said. "I followed your methods and now I've got this firehose of problems—hundreds of them—and I have no idea which ones are worth pursuing. I've made my life harder, not easier!"

I couldn't help but laugh. This is one of those “good problems”!

BUT…it’s still a problem!

When you get really good at spotting business issues that could be solved with AI, you can end up drowning in opportunities. Because, hey, it’s pretty broad!

What she needed—what everyone needs once they've built a substantial Problem Bank—was a system for making sense of it all. A way to categorise, segment, and prioritise these problems to reveal which ones are truly worth pursuing.

And guess what? AI can help us here too.

Let's get started:

From Chaos to Clarity: A Simple Filtering Approach

If you've followed Parts 1 and 2, you've likely got dozens (maybe even hundreds) of business problems in your Problem Bank. That's excellent—but it can also be overwhelming.

What you need now is a straightforward filtering approach that helps you identify which problems are worth pursuing. Here's the simple 3-step process we'll use:

1. Basic Categorisation: Organise your problems into meaningful groups

2. Problem Prioritisation: Score problems based on key value dimensions

3. Connection Finding: Identify natural problem clusters or portfolios

This approach is designed to quickly separate the high-potential problems from the rest, without getting bogged down in overly complex analysis.

All sounds very fancy: like I’m dusting off my MBA! But I’ve made a handful of prompts for you to work with - you’re just going to plug in your problems.

Step 1: Basic Categorisation

First, let's organise your Problem Bank into some meaningful categories. The simplest and most useful categorisation is by business function—this naturally aligns with how businesses think about their operations.

Here's a prompt to help you categorise your problems:

You are an expert in business problem analysis. I've been collecting business problems that could potentially be solved with AI. Please help me categorise these problems by business function.

For each problem, assign it to one of these business functions:
- Marketing & Sales
- Operations & Logistics
- Finance & Accounting
- Customer Service & Support
- HR & Talent
- Product & Development
- Legal & Compliance
- Other (please specify)

Also, for each problem, identify whether it is primarily:
- An efficiency problem (saving time/effort)
- A quality problem (improving results/reducing errors)
- A revenue problem (increasing sales/opportunities)
- A cost problem (reducing expenses)
- A risk problem (reducing risk/ensuring compliance)

Here's my list of problems:
[Insert your problem list here]

Present your analysis in a simple table with three columns: Problem Description, Business Function, and Problem Type.

After running this prompt with your Problem Bank, you'll have a simple table that organises your problems by function and type. This gives you a basic structure to work with and might already reveal interesting patterns—like a concentration of problems in a particular business function.

This is just the first pass though - we’re going to start cutting it down now.

Step 2: Problem Prioritisation

Now that your problems are organised, it's time to identify which ones have the highest potential value. Not all problems are equally worth solving—some represent much bigger opportunities than others.

The simplest way to prioritise is using a scoring system based on three key dimensions:

1. Pain Level: How painful or costly is this problem?

2. Market Size: How many people or companies have this problem?

3. Solution Fit: How well could AI solve this problem?

Here's a prompt to help you prioritise:

You are a business opportunity analyst. I need help prioritising business problems to determine which ones represent the best opportunities for AI solutions.

For each problem below, please rate it on a scale of 1-5 (5 being highest) on three dimensions:

1. Pain Level: How painful or costly is this problem for those experiencing it?
2. Market Size: How widespread is this problem across potential customers?
3. Solution Fit: How well-suited is this problem for an AI-based solution?

Then calculate a total score (sum of the three ratings) for each problem.

Problems to evaluate:
[Insert 10-15 problems from your categorised list, focusing on the business functions that had the most problems]

For each problem, provide:
- The three individual scores
- The total score
- A brief explanation of your reasoning
- Any specific aspects that make this problem particularly valuable or challenging

Then rank the problems by total score from highest to lowest.

Now, we are using AI at this stage which is not as good as actually asking customers. BUT we can use this to at least start our winnowing down of problems. I’d recommend using a Research model with this prompt for as much empirical evidence as possible.

Step 3: Finding Connections

The final step is to identify natural connections between your high-scoring problems. We want to start clustering up our problems by similarity. So instead of 10+ problems we might have 2-3 “clusters”.

Once you've identified your top 10-15 problems from the prioritisation step, use this prompt to find connections:

You are a strategic business advisor. I've identified the following high-potential business problems and want to understand how they might naturally connect or relate to each other.

Here are my highest-priority problems:
[Insert your top 10-15 problems from Step 2]

Please help me identify:

1. Natural groupings or clusters of 2-4 problems that are closely related
2. For each cluster, explain why these problems go well together
3. Suggest a simple name or theme for each cluster
4. Recommend which problem in each cluster would be the best one to solve first

Focus on finding meaningful connections where solving one problem would make it easier or more valuable to solve the related problems.

This analysis will reveal natural problem clusters or "portfolios" that could form the foundation of your business offerings. They will be closely associated and complementary. Instead of trying to solve individual problems in isolation, you can develop more comprehensive solutions that address multiple related issues.

We can also (as a bonus) package these up into different product/service levels. Smart!

What's Next?

In Part 4, we'll focus on problem validation. Having identified your most promising problems and clusters, you need to verify they're real, significant, and worth solving before investing in solutions.

We’ve been doing a lot of heavy lifting with AI but the only way to know for sure if these are genuine problems is by taking our solutions to market.

Thankfully we don’t need to fully develop our product or service to do this.

We'll cover initial validation methods, deep AI research techniques, and lean testing approaches to confirm that the problems you've identified are genuinely worth pursuing.

Keep Prompting,

Kyle

Looking back at this entrepreneurial rollercoaster, there's one clear pattern: The businesses that failed weren't necessarily bad ideas—they were unvalidated ideas that I fell in love with too quickly.

I became emotionally attached to my vision and refused to "kill my darlings" when early warning signs appeared. By the time I finally admitted there wasn't a real market need, I'd already invested months of work and thousands of pounds.

The businesses that succeeded? They all started with rigorous problem validation. I wasn't building products because I thought they were cool—I was solving problems that had been thoroughly verified to exist, be painful, and worth paying to solve.

This is the entrepreneurs' dilemma: We need passion to fuel our journey, but that same passion can blind us to reality. The solution is a systematic validation process that forces us to confront the truth about our ideas—preferably before we've invested significant resources.

Let's get started:

The Validation Imperative

In Parts 1-3 of this series, you've learned how to discover, categorise, and prioritise potential business problems. By now, you should have identified 2-3 promising problem clusters or portfolios that seem worth solving with AI.

But here's a sobering reality: Your judgment about which problems are worth solving is likely wrong.

Not completely wrong, but wrong enough that blindly following your intuition is risky.

The reason is simple: We all suffer from confirmation bias. Once we think we've found a great problem to solve, we unconsciously look for evidence that confirms our belief while ignoring evidence that contradicts it.

We want this to work so will see evidence accordingly.

So we’re going to get systematic.

There are three phases to effective problem validation:

1. Initial Validation: Quick research to confirm basic viability

2. Deep Validation: Thorough investigation with AI assistance

3. Real-World Validation: Simple tests with actual potential customers

Let's explore each phase with practical methods you can implement immediately.

Phase 1: Initial Validation

The first phase is about quickly confirming that your problem has basic merit. You're looking for existing evidence that:

1. The problem actually exists

2. People are actively trying to solve it

3. There's a market willing to pay for solutions

If we can’t get this far then there’s no point going forward.

Here are three effective methods for initial validation:

Method 1: Competitive Analysis

Contrary to common belief, the existence of competitors is often a good sign—it means there's a market willing to pay for solutions to this problem. Absence of competition might indicate the problem isn't worth solving (or very rarely, that you've found a blue ocean opportunity - it’s unlikely!!).

Competition risk we can deal with. Market risk we cannot.

For each problem, research:

• Direct competitors solving exactly this problem

• Adjacent solutions that partially address the problem

• DIY methods people currently use as workarounds

What you're looking for:

• Evidence of established businesses in this space

• Pricing models and approximate market size

• Customer reviews highlighting limitations of current solutions

Use an AI Research model for this ideally.

Method 2: Keyword Volume Analysis

Search volumes can provide quantitative evidence of problem significance. If many people are searching for solutions, the problem is likely real.

It's important to note that this is one area where AI isn't particularly helpful on its own. At least not yet!

AI models don't have access to current search volume data from Google, so you'll need to manually use keyword research tools and then feed the data to AI for analysis, or simply do this research yourself (which doesn't take long and gives you valuable market insights).

Use tools like Google Keyword Planner (free), Ubersuggest, or AnswerThePublic to look for:

• Monthly search volumes for problem-related terms

• Trend direction (is interest growing or declining?)

• Cost-per-click (higher CPCs often indicate commercial intent)

Method 3: Asking your audience

Post about the problem (not your solution) to your audience or relevant communities to gauge response.

This is where having your own audience becomes incredibly valuable. If you've built a social media following, email list, or community in your target industry, you have a ready-made validation group at your fingertips.

Some effective approaches:

• "Does anyone else struggle with [problem]?"

• "How do you currently handle [specific task]?"

• "What's your biggest frustration with [process]?"

Pay attention to engagement levels, specific pain points mentioned, and enthusiasm about potential better solutions.

If your problem passes initial validation (evidence of competitors, search volume, and forum interest), move to the next phase. If not, reconsider whether this problem is really worth solving.

Phase 2: Deep Validation

Once a problem passes initial validation, it's time for deeper investigation. This phase is about gathering detailed information and actively challenging your assumptions.

Method 1: AI-Powered Research

AI can help you analyse vast amounts of information about your problem space. However, we need safeguards against hallucination and confirmation bias!

If we ask it if something is a good idea AI tends to be agreeable and say “yessir it’s great!”. That’s not helpful!

Design prompts that actively challenge your assumptions rather than just seeking confirmation:

You are a skeptical business analyst evaluating the following business problem:

[Insert problem description]

First, argue AGAINST this being a significant problem worth solving with AI.
Then, argue FOR this being a significant problem worth solving.

Based on both perspectives, provide your balanced assessment:
- Is this likely a real, significant problem?
- What additional information would help validate or invalidate it?
- What specific aspects of the problem seem most promising?

This "red team/blue team" approach helps overcome confirmation bias and generates more reliable insights.

Method 2: Expert Interviews

Nothing beats talking directly to experts in the field. These could be professionals who experience the problem firsthand, consultants who work in the industry, or vendors of adjacent solutions.

Aim to conduct 3-5 expert interviews for each problem cluster you're validating, focusing on:

• How the problem manifests in their daily work

• Current solutions and their limitations

• The value of solving this problem more effectively

• Decision-making processes for adopting new solutions

• What are you missing?

Keep these interviews conversational and be genuinely curious. You'll often uncover aspects of the problem you hadn't considered, which can dramatically improve your eventual solution.

Phase 3: Real-World Validation

The final and most reliable phase is to test your problem assumptions in the real world without building a full solution.

Ultimately we have to move away from asking AI. Sorry!

Now we need to legitimately out our ideas in front of potential customers.

Method 1: Landing Page Tests

Create a simple landing page that:

• Describes the problem you've identified

• Outlines your proposed solution concept

• Includes a call-to-action that requires commitment

Generally the commitment will be joining a waitlist. I’m currently doing this with the AI Automation Accelerator - we’re at 1,200 people on the waitlist which is solid verification of market demand, especially because the verification target was 500.

Drive targeted traffic to this page through social media posts, small ad campaigns ($100-200 budget), or direct outreach to potential customers. Again, having an audience makes this much easier!

A well-constructed landing page test can validate both problem and solution fit before you build anything.

Method 2: Solution Simulation

One of the most powerful validation techniques is to manually simulate your AI solution before building anything—the "Wizard of Oz" technique.

Many now massive companies started this way. For example, Grubhub began with just a website and founders who manually called restaurants to place orders on behalf of customers.

Orders would come in on the Grubhub website and they’d call the restaurants and place the order, pocketing the difference… mad huh? There was no sophisticated logistics system—just people pretending to be the technology until they validated the concept and could build the real thing.

This approach:

• Validates willingness to use (and potentially pay for) your solution

• Gives you deep insights into exactly what customers need

• Helps you understand the nuances of the problem

• Builds a base of reference customers before you build anything

Importantly, you don't need to tell people you're doing things manually behind the scenes. The goal is to test the value proposition, not the technology.

Method 3: Paid Pilot Program

If your problem has passed the previous validation stages, consider setting up a paid pilot program:

• Approach 5-10 potential customers who would benefit most

• Offer a discounted pilot program (still paid, but lower than your eventual pricing)

• Deliver the solution manually or with minimal automation during the pilot

This approach:

• Confirms actual willingness to pay

• Provides revenue while you're still validating

• Gives you deep customer insights

• Creates case studies for your full launch

The key is to set clear expectations about the pilot nature while still delivering genuine value to participants. Be super open about the fact that they are “beta testers” and price accordingly. If anything they’ll get much better white glove service. And at a lower price! Great for them!

What's Next?

In our final part tomorrow, we'll explore how to turn your validated problems into profitable AI businesses. We'll cover:

• Building a portfolio of complementary problems

• Matching problems to the right AI technology approach

• Creating service packages and pricing strategies

• Initial marketing and customer acquisition approaches

Between now and then, select your most promising problem cluster and run it through at least one validation method from each phase. Your goal is to have at least one thoroughly validated problem ready for solution development.

Kyle

Tbf that’s much (much) better than most of his peers. They are all doing dropshipping or crypto or some-such.

"Great," I said. "So what are you building to solve it?"

He looked at me blankly. "I... don't know. I thought I'd just build an AI app."

Basically: “AI”. Followed by some hand waving.

This is where many aspiring AI entrepreneurs get stuck. They've identified a real problem, but they haven't thought strategically about the right way to solve it.

Saying “I’ll build with AI” is like saying “I’ll build with the internet” or “I’ll build with a computer”. Meaningless.

Not all AI solutions are created equal, and matching the right technological approach to your validated problem is crucial for success.

Let's get started:

Matching Problems to Technology Approaches

Different business problems require different technical approaches, but here's an important insight: most problems can be solved using a sequential approach, starting with the simplest solution and only advancing to more complex ones when necessary.

Think of it as a ladder of sophistication:

AI Assistants: The Entry Point

AI Assistants are often the best starting point for any problem. They're ideal for content creation, information synthesis, expert guidance, and customer service. Almost any business problem can be initially addressed with a well-crafted prompt and a simple interface.

Implementation is straightforward—you can start with a custom GPT or Claude configuration and a basic web form. I personally use Launch Lemonade as it also allows for paywalls and restricted access.

Many successful AI businesses began with nothing more than this and still generated significant revenue while validating their market.

For example, a legal document review problem could start as an assistant that analyses contracts when users paste them into a form. This might take just hours to set up but deliver immediate value.

AI Automations: The Next Step Up

Once you've validated your solution with an assistant approach, you might find users want more automation. Now's the time to move up to AI Automations, which excel at repetitive tasks, data extraction, and defined workflows.

Automations excel when we need to connect multiple tools and pieces of software together and pass data through AI.

Building on our legal document example, you might create an automation that not only reviews contracts but also extracts key terms, generates summaries, and then sends email alerts for problematic clauses—all without human intervention.

This step requires more technical setup but delivers greater value to users who've already proven they're willing to pay for the simpler version.

AI Agents: When Complexity Demands It

If your solution requires orchestrating multiple systems or handling complex multi-step processes, AI Agents become appropriate. But crucially, you should only advance to this level after proving market demand with simpler approaches.

Too many businesses rush to agents when it’s really not required. And by doing so they just create trouble for themselves by adding complexity.

Continuing our example, an agent might not just analyse and extract information from contracts but also compare them to previous agreements, update CRM systems, draft response emails, and schedule follow-up meetings, all whilst learning and improving itself via a centralised repository.

Full AI Applications: The Premium Solution

The most sophisticated approach is a full application with custom interfaces and tight system integration. This represents a significant investment but can command premium pricing—after you've proven the market wants it.

Our legal document solution might evolve into a comprehensive contract management system with dashboards, workflow management, and enterprise integration. But building this from day one without validating with simpler approaches would be extremely risky.

The key insight here is progression. You don't need to jump immediately to the most complex solution. Start simple, validate, then advance only as far as necessary to solve the problem effectively.

Building Your Solution

With this sequential approach in mind, here's how to build your solution:

1. Start Simple, Always

Begin by creating the simplest version of your solution that delivers value:

• For most problems, start with an AI Assistant approach

• Create a basic interface (even just a form input is fine)

• Focus on nailing the core problem-solving aspect, not fancy features

• Get it in front of customers as quickly as possible (days, not weeks)

Importantly: If this basic version isn’t valuable then simply adding MORE is unlikely to help much. Nail the core value.

This approach allows you to validate your solution with minimal investment. I've seen businesses generate six figures in revenue with nothing more than a well-engineered prompt and a simple form.

2. Observe and Learn

Once your simple solution is in the hands of users:

• Watch how they actually use it (not how you expected them to use it)

• Identify pain points and limitations in the current approach

• Ask users, "What would make this 10x more valuable to you?"

• Look for patterns in feature requests and complaints

This real-world feedback is gold—it tells you exactly where to focus your development efforts next.

3. Advance Only When Necessary

Based on user feedback and business metrics, decide if and when to move up the sophistication ladder:

• If users want more automation and less manual intervention → Advance to AI Automation

• If they need integration with multiple tools or systems, all automatic → Consider AI Agents

• If they require custom interfaces and interconnectivity with their existing systems → Explore full applications

The key is to let market demand pull you toward greater complexity, rather than pushing toward it because it seems more impressive or technically interesting. Don’t just chase “that would be cool”.

4. Pricing Should Reflect Value, Not Complexity

Super important.

Just because it took you longer to make doesn’t mean it’s more valuable to the end user.

Customers pay for what they get, not for your pain and suffering. They don’t care.

A simple solution that solves a painful problem is worth far more than a complex one that doesn't. Price based on value delivered, not technical sophistication:

• A basic AI Assistant that saves 10 hours per week is worth roughly $2,000/month in value

• Pricing at 10-30% of delivered value ($200-600/month) is reasonable

• As you move up the sophistication ladder, increase prices in proportion to additional value

Flipside is this: Don't fall into the trap of underpricing simple solutions just because they were easy to build.

Creating Service Packages and Pricing

Once you've selected your technical approach, it's time to create service packages and pricing. The key principle: price based on the value your solution provides, not your costs.

Consider what your solution does for customers:

• Time saved

• Revenue increased or costs reduced

• Risk or liability decreased

• Competitive advantage gained

For example, if your solution saves a business 10 hours of work per week at $50/hour, that's roughly $2,000 monthly in value. Pricing at 10-30% of this value ($200-600/month) would be reasonable.

The pricing model you choose—subscription, usage-based, outcome-based, or hybrid—should align with how your solution delivers value to customers. Do not base it on your effort! Remember: they don’t care!

Wrapping Up The Series

Over these five parts, we've covered a comprehensive framework for finding, validating, and solving real business problems with AI:

Part 1: The Problem-First Mindset - We established why starting with problems rather than technology is crucial for AI business success, and explored techniques for mining problems from your personal experience.

Part 2: Problem Discovery Beyond Personal Experience - We expanded our problem discovery to market observation and data analysis, building a robust Problem Bank of potential opportunities.

Part 3: Problem Categorization & Segmentation - We learned how to organize and prioritize problems to identify patterns and high-value clusters using a simple filtering process.

Part 4: Problem Validation & Verification - We explored methods to verify that problems are real, painful, and worth solving before investing in building solutions.

Part 5: Building Problem-Solving AI Businesses - We completed the framework with strategies for turning validated problems into cohesive business offerings matched to the right technical approaches.

The entrepreneurs who build the most successful AI businesses in 2025 and beyond will not be those with the most advanced technical knowledge. Quite the contrary.

It’ll be who become masters at finding and solving real, painful, validated problems. And remember, the bigger the problem the bigger the business.

Keep Prompting,

Kyle