{"id":220701,"date":"2024-07-26T13:56:17","date_gmt":"2024-07-26T05:56:17","guid":{"rendered":"https:\/\/www.grab.com\/sg\/?post_type=editorial&p=220701"},"modified":"2024-09-09T14:49:59","modified_gmt":"2024-09-09T06:49:59","slug":"how-genai-accelerates-product-development-cycles-at-grab","status":"publish","type":"editorial","link":"https:\/\/www.grab.com\/sg\/inside-grab\/stories\/how-genai-accelerates-product-development-cycles-at-grab\/","title":{"rendered":"How GenAI accelerates product development cycles at Grab"},"content":{"rendered":"\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t

Generative AI (GenAI) is a buzzword in the tech industry, sparking both excitement and misconceptions about its capabilities and limitations.<\/p>

On the one hand GenAI can be seen as a progression from previous applications of AI, like machine learning. Indeed, Grab\u2019s use of AI is not a recent development. We already deploy over 1,000 different AI models across multiple touchpoints\u2014in areas such as personalisation, estimating wait time, trust and safety.<\/p>

However, it’s crucial to understand GenAI\u2019s unique potential. At Grab, we do view it as a game-changer, because it helps us accelerate product innovation in many important ways.<\/p>

That\u2019s because GenAI excels at creating new content by learning patterns from existing data. These traits make it a versatile tool in product development. We can use GenAI to augment data sets, generate test cases and simulate user behaviour, create personalised text and images at scale, and so on.<\/p>

How GenAI\u2019s specific strengths help us speed up product development is best explained with a simple example. Let\u2019s look at translation technology.<\/p>

Fewer training sets<\/h5>

We’ve been using translation technology on the Grab app for a while, but GenAI has allowed us to do much more in less time.<\/p>

In the past, it would have required a significant investment of effort to build training sets for the specific languages we target. Training sets, in this case, are large collections of text data used to train the AI translation model.<\/p>

Now, with the help of LLMs, we can do this without creating specific training sets for each language. This has greatly reduced the development effort and time required to deliver visitors from overseas a great user experience on Grab\u2014no matter where they are from.\u00a0<\/p>

(Also read: Grab\u2019s Help Centre articles are now available in Chinese, thanks to machine translation<\/a>)<\/strong><\/p>

We can use LLMs to translate menu items and reviews, for example, from Thai into Korean. We can make thousands of help centre articles available in various languages in a short period of time.<\/p>

Language expert<\/h5>

GenAI\u2019s strong capabilities when it comes to interpreting text and language has also let us make significant progress with important features such as Grab\u2019s voice mode. This will make our services accessible to more people, for example those with a vision impairment.<\/p>

The technology is also present in our ‘last-mile knowledge mining’. One of the most complex tasks for our drivers is finding and meeting the consumer at the end of the trip. Consumers often provide complicated instructions, especially in countries like Indonesia where addresses are less structured.<\/p>

Previously, this valuable last-mile information would have only helped the individual delivery driver they were sent to at that moment. Now, we use LLMs to parse the many notes customers may have left about a particular location over the years and turn those into additional navigation points. This has significantly reduced stress for our drivers and has been well-received.<\/p>

Vision language<\/h5>

LLMs reasoning capability has helped us enhance our mapmaking efforts.<\/p>

We’ve invested significantly in creating the most efficient maps in Southeast Asia for our delivery and mobility drivers. To achieve this, we’ve deployed our map making cameras\u2014called KartaCam\u2014across our fleet to capture extensive imagery. This data is then used to continually update our maps.<\/p>

However, there have always been complex scenarios that posed a challenge. For instance, interpreting a roundabout with seven exits can be difficult even for humans, let alone traditional AI models.<\/p>

Now, with the help of vision language models, we can provide the model with data and context, and ask it specific questions like, “Is it allowed to turn right into this street?” The model can then update the map accordingly.<\/p>

(Also read: How Grab uses AI to generate more precise delivery instructions<\/a>)<\/strong><\/p>

The latest models have significantly sped up our map creation process, resulting in better, more accurate tools that help drivers navigate and manage their tasks throughout the day.<\/p>

Productivity tool<\/h5>

Translation and navigation are two of many areas we\u2019ve seen a step-change in product innovation on our platform thanks to GenAI.<\/p>

The technology has also been instrumental in enhancing productivity across various teams at Grab.<\/p>

For instance, my team, the product team, uses an internal productivity tool to write product specifications and user stories.\u00a0<\/p>

We also have a tool that allows us to query our data in natural language. If I want to know how many users booked rides to Sentosa yesterday, I can simply ask this tool and get an answer within seconds. Previously, I would have had to contact a data analyst and spend time discussing it with them.<\/p>

This has significantly sped up our productivity, which, in turn, allows us to explore and deploy more product updates\u00a0<\/p>

As our teams, our drivers, and merchant-partners become more efficient, we can make our services more affordable, which in turn drives growth.\u00a0<\/p>

Ultimately, our goal is to drive growth by increasing efficiency, and\u00a0 GenAI acts as an accelerator on this path. So far, we are seeing encouraging results.<\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<\/div>\n\t\t","protected":false},"parent":180237,"menu_order":0,"template":"grab21-default","acf":[],"_links":{"self":[{"href":"https:\/\/www.grab.com\/sg\/wp-json\/wp\/v2\/editorial\/220701"}],"collection":[{"href":"https:\/\/www.grab.com\/sg\/wp-json\/wp\/v2\/editorial"}],"about":[{"href":"https:\/\/www.grab.com\/sg\/wp-json\/wp\/v2\/types\/editorial"}],"version-history":[{"count":44,"href":"https:\/\/www.grab.com\/sg\/wp-json\/wp\/v2\/editorial\/220701\/revisions"}],"predecessor-version":[{"id":220897,"href":"https:\/\/www.grab.com\/sg\/wp-json\/wp\/v2\/editorial\/220701\/revisions\/220897"}],"up":[{"embeddable":true,"href":"https:\/\/www.grab.com\/sg\/wp-json\/wp\/v2\/editorial\/180237"}],"wp:attachment":[{"href":"https:\/\/www.grab.com\/sg\/wp-json\/wp\/v2\/media?parent=220701"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}