AI is now a shopping companion – People are increasingly adopting AI in their shopping journey. According to a recent McKinsey study1 53% of consumers globally shop with an AI with the number being highest in China (80%) and India (76%) and lowest in Japan (37%). Unsurprisingly, the study found that GenZ and Millennials are more likely to be using an AI to shop than older age groups. The pathways and levels of engagement vary with 17% using it for product recommendations while 10% have engaged an AI powered shopping assistant2.
AI
is making inroads even in South-East Asia - Shopping journeys in South-East
Asia are known to be different from the rest of the world due to the prevalence
of live streaming and social commerce. In fact, 71% of TikTok Shop’s global GMV
comes from South-East Asia3.
So,
to unpack the evolving landscape of South-East Asia and the impact of AI on
shopping journeys I worked with PercepSense which is PercepTech.AI’s moderation
platform that covers what shoppers do (BehaviorSim) and what they say (PercepSense).
Using BehaviourSim, 29 online shopping missions across Thailand, Singapore,
Indonesia and Philippines were passively observed without disrupting real life
behaviour. The shopping missions spanned FMCG (face wash and shampoo), durables
(earbuds and powerbank), apparel, walking shoes and in home dining. This was
supplemented with 101 in-depth qualitative interviews conducted by an AI
moderator using PercepSense.
We
found that 17 out of the 29 journeys (i.e 59%) used an AI. More importantly we
found that AI is reshaping shopping journeys in South-East Asia in a manner
that can be described as an evolution rather than a revolution.
So what is different with AI in the shopping journey.
1.
Search queries are different – A BCG study4 done in the US in association
with Yale found that when people use an AI to search, they type three times
more words than a traditional Google search. For instance, a traditional search
for a holiday might be “Europe beach destination” while the agentic search
equivalent would be “My husband and I (Both 30) want to take our honeymoon in
Europe. We prioritise good local cuisine, rocky cliffside beaches, and villages
with culture. Not too touristy”.
In South-East Asia, when people shop using an AI assistant,
the queries are both detailed and nuanced. For instance, one face wash shopper
in Thailand prompted ““Help me find a face wash. My problem is dry
facial skin that lacks moisture. Please find a product that fits, with a budget
of no more than 500 baht.” – Face wash shopper, Thailand
The
equivalent without an AI assistant would likely have been “face wash for dry
skin”.
There are some prompts where it is difficult to imagine an
equivalent without an AI assistant. For instance, a dinner shopper in
Philippines “Hello I just arrived from work, it has been a busy day tbh I
dont wanna think too much I just want go have delivery for dinner after a take
a nap”. Another dinner shopper in
the Philippines prompted “can you recommend a restaurant that is present in
grab, quite cheap 200ish budget range + have short delivery time”
Shoppers have learnt that with AI searches the prompts can be specific e.g., restaurant present on Grab but can be expansive e.g., dry facial skin.
2.
Outputs for search queries are different – Searches on online shopping
platforms bring up all the options related to the keyword, ranked basis
relevance. For instance, when I search for Anti-dandruff shampoo on Shopee I
get a list of 10 brands (not variants) above the fold on the first page. “So many earbuds come up when you
search on Shopee itself… ChatGPT lists the top few straight away” – Earbud shopper, Philippines.
AI
searches on the other hand, produce specific and narrow recommendations. For
instance, when I asked Claude “I have dry scalp. Can you recommend an anti-dandruff
shampoo”, it gave me three options – Head & Shoulder’s Dry Scalp Care,
Vichy Dercos Anti-Dandruff for Dry Scalp and Nizoral. This, in a manner of
speaking, shrinks the erstwhile limitless digital shelf down to three brands. “A
specific prompt really matters to me because it narrows the options”
walking shoes shopper, Indonesia.
Results
from AI searches return clear facts against what shoppers have asked about. For
instance, one shampoo shopper in Thailand explained their interaction with
ChatGPT as “ChatGPT asked about my scalp condition and budget first, then
explained why it suits people with dry, itchy scalps”. This makes the
recommendations more credible for the shopper.
So what has not changed with AI in the shopping journey
1. Importance of brands – When the AI search throws up a brand that shoppers are aware of and have a positive impression about, they are likely to buy it because the LLM has confirmed their positive predisposition. For instance, a Powerbank shopper in Thailand said “I’ve bought an Anker data cable before, so I had no doubts choosing Anker”.
When AI searches throw up a brand that is unknown, shoppers do further research. For instance, one dinner shopper said, “going straight to (my app) without ever having eaten there is a total blind buy”. For new brands entering the shopper’s repertoire, it is important that the product perform in order to elicit repeats. “When I receive a product that is different from what I expected, it’s disappointing. And that’s a big factor in me not checking out that website or that seller again.” – Shopper interview
2. AI does not make the shopping quicker – Using AI in the journey does not result in time saving but instead the time is spent differently. With AI, more time is spent on prompting and less time on checking. While they spend less time checking while using an AI, people still check reviews, photos and creators indicating that trust in AI search results is still in its formative era. “It needs to go through maybe 10 to 20 times for me to actually trust that this system actually works” – Shoper in Singapore.
3.
Shoppers are not yet delegating their credit card to agents – While
AI is used across multiple stages of the shopper journey including for product
recommendations and price comparisons, it is still not used for autonomous
purchases. The McKinsey study1 found that globally only 9% of
shoppers used an AI tool to complete a purchase. In our study in South-East
Asia, we found that none of the 17 AI powered journeys used the AI to complete
the purchase. Trust in the AI recommendations and in the payment, systems will
need to improve significantly before we start to see truly agentic shopping. “If
it’s below like $5 then sure” Singapore shopper
Machine
discoverability – The new imperative to win preference – To win in online
shopping without AI, brands need to have physical availability and mental
availability. In an online shopping environment without AI, a brand will appear
in search lists as long as it contains the keywords that the shopper is
searching for. Hence, physical availability was about getting listed, getting
ranked and ensuring being in stock. Mental availability was about shoppers
having a memory structure associated with the brand.
In
shopping journeys where AI search is involved, brands need to have physical
availability, mental availability and machine discoverability.
In
the study we did we found that in 14 out of the 17 journeys (82%) shoppers
selected a brand that was recommended by the AI. Hence, it is imperative for
brands to be discovered by machines. This is not akin to SEO which could be gamed
by techniques like keyword stuffing. To be discoverable by machines, brands need
to be discovered by the LLM against the needs expressed in the prompt in the
places that the LLM looks including social media posts, ratings and reviews
etc.
Hence,
brands must now be readable by machines and lovable by humans.
Conclusions
– AI is reshaping shopping journeys in South-East Asia. The fundamentals of
physical availability and mental availability to win preference remain intact.
However, brands need to consider an additional element to win preference in AI
powered shopper journeys: machine discoverability. Hence, brands must now be
readable to machines and lovable by humans.
My
sincere thanks to Anshul, Siva and Rajeev at PercepTech.AI for their
partnership on this piece of work.
3.
https://digitalinasia.com/tiktok-shop-southeast-asia-growth-10x/

