| QA work | Inefficient Meetings | User Needs Failure | |
| description | Large amount of QA (quality assurance) work when using AI | Lots of meetings, lack of progress | Failing to understand real user needs / problems & reaching users |
| ICP | PMs at YC startups on a constrained time scale building user facing software who don’t have enough resources to quality check AI-generated work | PMs and CEOs at companies with more than 100 employees who spend more than 50% of their time in meetings and don’t have time to get meaningful individual work done | PMs at pre-PMF companies that have tried administering user surveys and customer interviews but feel no closer to PMF |
| SAM | PMs and engineers @ startups | PMs/CEO @ larger companies | PMs @ pre product-market fit startups |
| SOM | YC startups | PMs/CEOs with a lot of meetings | YC startup |
| Competitive landscape | Red ocean. AI code reviews like Graphite, Macroscope, and even Github products all help make sure that code is production grade when it gets merged. New IDEs like Claude Code and Codex can also write scripts to test functionality. However, these are all reliant on AI and still miss a lot of more human gaps in products (like is this actually nice to use) | Red ocean. Lots of products like Granola etc provide meeting and notetaking services to get the most out of your meetings. However, less clear if there are good solutions for making the actual meetings better. | Red ocean. Products like Typeform and Qualtrics and Google Forms all aim to take the pulse of your customer, while products like Gong and Granola allow you to record meetings and get insights. |
| Evidence strength | Brought up in 3 meetings | Brought up in 5 meetings | Brought up in 2 meetings |
| Current workarounds | Using AI to write and run QA tests and asking AI to make a guide for manual testing | Use meeting summarizing tools including Granola, Loom, and Fellow. Interviewees mention they just suffer through it and spend time in meetings and work on product ideation at end of day | Spend lots of time conducting user surveys and interviews. Use tools such as Typeform, Qualtrics, and Google Forms to collect user survey information and Gong and Granola to get user insights after interviews. |
| Competitive gap | AI doesn’t replicate how humans use it.
Some accessibility tooling tools that only do accessibility. |
Meeting summarizing tools do not provide real-time time management tips or encourage decisions (such as decision polls or automated messages to move on to the next topic) | This process is still pretty manual and there are not many tools that help automate the process and find users that fit the ICP to interview. |
| Buildability | Maybe | Yes | Maybe |
| AI fit | Weak, problem arises because of AI use, high risks & no human review | Strong, recurring issues | Strong, recurring/routine. Weak, flexible.
Overall: medium |
As a team, we’re leaning towards the meetings problem for two reasons:
- We think AI is a great fit to solve this problem.
- We think it is reasonably scoped for this class.
While we are excited about brainstorming solutions to solve this problem, we do acknowledge that it might be harder to differentiate our product from more established products such as Granola that could add in real-time features to keep meetings on track. Although we were also excited about the QA problem, it’s less clear how good of a fit AI is to solve a problem whose solution is human review of AI work. We also had 5 different interviews mentioning a struggle of too many meetings and not enough work getting done, so it felt like the most evidence-backed problem of the bunch. When brainstorming our ICPs for each problem, we followed the suggestions from the “How to Identify Your Ideal Customer Profile” reading. For each problem, we brainstormed the 9 characteristics that were presented and narrowed these down to the three that we thought were the most important.
