How much market research should AI be doing?
The speaker advocates for a heavy reliance on AI in market research, starting at 100% AI during initial avatar validation and shifting to 80/20 AI-to-human ratio after validation. They argue AI can match or exceed human performance in knowledge work, and suggest building automated tools to replace manual human research tasks like scanning forums and social media.
Summary
The speaker outlines a phased approach to AI usage in market research based on where a business is in its validation process. At the outset, before any confirmation that a target avatar or market is worth pursuing, the speaker recommends using AI entirely — a 100% AI approach — since there is no reason to invest significant human effort before validation is established.
Once a market or avatar shows promise and has been validated, the speaker suggests shifting to an 80% AI and 20% human research split. This reflects a belief that AI is capable of performing knowledge work at least as well as humans, if not better, and that the limiting factor is simply how well the tools are designed to accomplish specific goals.
To illustrate this point, the speaker challenges the notion of manual human research — such as browsing forums or scrolling through recent social media posts — by proposing an automated alternative. They suggest building a simple application that scans social platforms for specific keywords and delivers results each morning, effectively replacing repetitive human research tasks with an AI-powered workflow. The speaker frames this as not only possible but as the logical evolution of how research should be conducted.
Key Insights
- The speaker argues that initial market research should be 100% AI-driven because there is no validation yet that the target avatar or market is worth pursuing, making heavy human investment premature.
- The speaker claims that once a market or avatar shows promise, the appropriate research ratio shifts to approximately 80% AI and 20% human.
- The speaker holds the belief that AI can perform any knowledge work task at least as well as a human, asserting that the quality gap between AI and human output depends entirely on how well the tools are built.
- The speaker challenges manual human research practices — like browsing forums or recent social media posts — by arguing they can and should be replaced with automated applications.
- The speaker proposes building a simple application that scans all social platforms for specific keywords and delivers daily research summaries, positioning this as a direct and achievable replacement for human social media research.
Topics
Transcript
[0:00] All the initial research is done with AI, because you don't have any validation that that avatar is even worth pursuing. So initially, it's literally like 100% AI. After you have validation that there is some promise [music] in this particular avatar this market, then I think the ratio would probably split to I would say 80% AI and 20% [music] human. I just have this belief that anything a human can do that is knowledge work, AI can do just as well if not better. It just all comes down to how you build the tools to accomplish said goal. So in this case, [0:31] like if the human research is going into, you know, forums or looking…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Mark Builds Brands
the ad styles that actually print
The speaker discusses which ad formats perform well across markets, highlighting native statics, AI animation, and UGC as generally effective options. However, success ultimately depends on testing and understanding your specific target demographic's native content consumption patterns.
how to make ads that actually print
The speaker identifies two critical mistakes in native static ad campaigns: treating ads like content and overestimating creative importance. For native statics, copy is the primary persuasion tool, and most advertisers fail by not testing enough variations and using poor funnel architecture.
How to structure your post purchase upsell
Effective post-purchase upsell (OTO) pages require connecting the main product to complementary products by addressing customer needs strategically. Two proven frameworks are offering more of the same product at a discount, or creating an OTO 1 that solves a new problem created by the front-end product.
how to turn 1 winning ad into 30 variations
The key to scaling ad creative isn't finding one winning ad, but rather identifying the winning sales message behind it and then transmuting that message across multiple formats and variations. A single winning VSL can generate 30+ testable ad variations by extracting core variables and adapting them across different creative formats.
ABO vs CBO for creative testing, which is better?
The speaker discusses CBO (Campaign Budget Optimization) challenges in creative testing, recommending a minimum allocation of $50 per day per ad to prevent budget spreading. They explain that budget must scale proportionally with new ad sets, and suggest moving to ABO (Ad Set Budget Optimization) if managing too many ads becomes problematic.