Use AI research tools to discover and shortlist real problems people pay to solve
This week shifts the focus from exciting ideas to costly, frequent problems. You will use AI to speed up research while keeping evidence from real people at the centre of your decision.
A promising opportunity begins with a person who is struggling to make progress. Look for workarounds, repeated complaints, wasted time, lost money, missed deadlines, or risky manual steps. Avoid deciding on a solution before you understand the situation.
Reviews, forums, professional communities, help pages, and public discussions reveal the language people use. Capture the exact problem, who reported it, the situation, the current workaround, and any consequence. A repeated complaint is a signal to investigate, not proof of a market.
Describe what the person is trying to accomplish rather than the tool they say they want. A useful statement is: When I am in this situation, I want to make this progress, so I can achieve this outcome. This wording leaves room for different solutions.
Compare ideas using the same criteria: severity, frequency, willingness to pay, ease of reaching the customer, and your ability to help. Use a one to five scale and add a short evidence note beside every score. The note matters more than the number.
Approximately 1.6 million views
Why it fits: Explains how good startup ideas emerge from real problems rather than brainstorming random app concepts.
Identify three recurring problems in your own work or daily life, then score each for frequency, urgency and willingness to pay.
The video supports the lesson above — it doesn't replace it.
Act as a research assistant. I will paste public comments about a customer group. Extract repeated problems in a table with columns for situation, desired progress, current workaround, consequence, frequency in the sample, and direct evidence. Do not invent facts. Label weak or ambiguous evidence and suggest questions for customer interviews.
Week 2 · Case study
Source note: Documented customer story from Replit. The learning improvement is reported by Replit, not independently verified here. The student-sized version below is an illustrative exercise, not part of the reported case.
Kelly Vanderbeek was homeschooling her 13-year-old son, Silas, when ordinary spelling applications failed to meet his needs. The products felt too young for him and did not provide the feedback he needed. Kelly used Replit to build Spell Strike, a spelling game designed with her son rather than for an imaginary customer.
Silas influenced important product decisions, including the game format and a feedback tile showing what he got right and wrong. Replit’s case study reports that he progressed by two grade levels in three months.
Do not start with “I want to build an AI app.” Start with one person experiencing one repeated problem. Observe that person, build the smallest useful interaction, and let them test it.
Which feature came from direct user evidence rather than the builder’s assumption?
Respond in your submission →Produce an AI assisted research report on ten problems and select your top three. For each problem, show the customer, evidence, current workaround, likely consequence, and score. End with a short explanation of why your top three deserve interviews.