Almost every conversation with start-up founders includes a question of Gen-AI applications in their organisations. They typically end up providing two answers –

1. It enables my workforce to be more efficient –
   • Code generation is faster (we need less engineers)
   • Documentation is easier; product mock-ups can be brough to life much faster
2. x% of customer service is fully automated using AI

Most founders talk about Gen-AI as a cost lever. I think that’s the least interesting use of it. The real opportunity is not cheaper operations, but radically better user experience – especially in search and discovery.

The last decade of “consumer-tech” innovation has seen a number of marketplace models rise up and delight consumers by offering a large number of options for any desicison. If I want to order a shirt from the internet, I need to browse through a million choices. How would I go about it today on Amazon?

1. Type Shirt
2. Open Filters
   • Select a price range
   • Select Average Rating > xx
   • Select brands
   • Reluctantly select size (Varies by brand)
3. Sort by Popularity or Sort by Price! (Why can’t I do both? – Also what does popularity really mean?)
4. Browse through all the options and try to spot any patterns in customer review

As you can tell, this is an extremely time consuming experience. Compare this to “Show me medium sized shirts which have ratings > 4.5 and customers believe is a good smart causal shirt.”

Imagine, just saying this out loud and the the you are presented with select options. The value of such prompts increase with the complexity of the product or the service. For example, selecting a restaurant which optimises for the location between you and fellow diners along with cuisine, rating, popular items etc. Or figuring out a hotel which optimises for location, noise levels, amenities, service constraints, budget….

Life is so hard! The user is doing the work that the marketplace should be doing. I especially feel the following online marketplaces must innovate their search and discovery –

1. Travel and Hotels
2. Ecommerce marketplaces
3. Dining out & Food delivery
4. Job marketplaces

This begs the question why have marketplaces still not fully integrated AI which can potentially enhance the experience of users? Firstly, I think there is an incentive mismatch. AI-led discovery threatens sponsored listings and ad revenue.

Secondly, there is a almost a kind of algorithm sunk cost. Years have been spent by these companies in fine tuning filters, ranking and SEO-friendly mechanics.

I also think hallucinations are a relevant concern. Their is skepticism in shifting to a solution which is not 100% accurate yet, especially with knowledge recall and more complex prompts.

Shout out to District for introducing solutions for Dining Out. Early experiments are hinting at whats possible – Its doing a great job so far of recommending restaurants and items to order based on the prompts.

The next generation of consumer marketplaces might not win by offering more choice, they’ll win by understanding intent better than users can articulate themselves.