Is AI the straw that stirs the discovery drink?

eCommerce and Retail Talk
1

Over the last few weeks - the AI hype has been showcased over the Internet. What is AI, what is generative AI, etc? These are all questions I have been pondering. Also, this is not a contrarian post in comparison to what Marc Andreessen recently wrote, “Why AI will save the world”

Let’s first get the elephant in the room out - AI that will replace humans is years away.

Let’s talk about generative artificial intelligence, which in simple terms, is the use of algorithms to create new content such as images, videos, code, and text.

Content creation is expensive - whether it’s product descriptions, short videos, or product images. Brands and platforms increasingly offer the minimum effort to minimize costs and get items in front of customers on websites. In the future, relevance will become harder as the signal-to-noise ratio is about to get considerably more noisy.

Discovery in e-commerce in the West is a missing element to help customers browse and find products they might not be aware of. Tagging of images is a laborious process and, in some cases, has not been done. Are we about to find a new generation of e-commerce experiences based on generative images and short videos?

The most obvious first example of this is Contextlogic, the company behind Wish Shopping which created tagged imagery based on advertising which morphed into cross-border shopping. Wish is a browse first and be haggled by retargeting ads next experience.

The manipulation of images is about to become considerably easier, and discovery could likely become more widely adopted by platforms as customers are having to search heaps of ai generated text, making the text-based search less effective. This will not happen overnight but rather become more prevalent as more companies use AI as part of their operations,

I might be having a case of confirmation bias, having spent a large chunk of time researching AI use cases for commerce and the relevant startups that can offer these solutions. Or this could genuinely be on the horizon.

3 Likes
2

I see a few benefits to retailers

  • Idea-starters for titles, descriptions and attributes of all-types.

  • Brand voice needs to be a big component. A lot of generators don’t have this component yet, but it should be coming everywhere as I have heard Jasper.ai advertising it recently.

  • Long-tail of retail catalogs. The long-tail of retail catalogs gets the least amount of attention from merchandisers, so it’s likely that this will be the place where a lot of people start.

  • Imagery is another huge usecase. Only the top luxury products will use real models in the future.

3 Likes
3

don’t forget AI includes ML and retail is far from exhausting that capability:

  • Pricing based on price elasticity (maybe in real time)–some products are elastic, some are not (and will sell based on other criteria, e.g. availability, perception of need/fit). Trick is to cluster products based on similarity of attribution so you can “gross up” the thin data signals you have. Also models to help determine initial price from attributes (which may make it easier to not progress products that are CR a P --cannot realise a profit).
  • customer clustering, location clustering (into marketing, assortment, allocation)
  • ML for next best action for customer (what to recommend next)

a lot of ML is about “what”, generative will help automate the “how” (content).

What is needed is increasing democratisation of ML (possibly cross retailer/marketplace, although care is needed for anti-trust e.g. no price collusion).

2 Likes
4
1 Like