- Useful Now
- Posts
- Tiny AI nudges, big sales jumps
Tiny AI nudges, big sales jumps
Grab shows how one chatbot can lift merchants and drivers without armies of engineers.
Alex Payne
April 24, 2025
Hi everyone! I am taking a new direction with Useful Now to focus on Asia-first AI wins and provide real life case studies, that you can replicate. At the very least I hope to provide a little inspiration on how you can make better use of some of the AI tools out there!
Last week OpenAI released their new o3 and o4-mini reasoning models. They are very good, I recommend having a play around if you have a plus or pro account!
This week, as I’ve been scrolling through case studies I noticed a pattern: the best AI uses feel less like software, more like a helpful colleague tapping you on the shoulder. Let’s unpack one example that proves the point. No jargon, no armies of engineers, just tiny pushes that move the needle.
Real-World Example: Grab’s AI Merchant Assistant & Driver Companion
Earlier this month, Grab rolled out two agentic assistants at its inaugural GrabX showcase in Singapore
– AI Merchant Assistant sits inside the GrabMerchant app. It reads each stall’s sales, photos and promos, then suggests simple actions - better menu shots, rival-beating deals, or ad boosts. Future updates will surface GrabFin loan offers.
– AI Driver Companion lives in the Driver app. A ride-guidance feature steers cars towards real-time demand hotspots; a voice tool lets drivers log hazards hands-free in multiple Southeast Asian languages.
Results so far
– Over 250 000 drivers use ride-guidance every.
– Drivers file 16 000 spoken road reports a day, feeding GrabMaps and cutting dashboard typing.
– Hazard logs are 5× richer than the old tap-and-type method
How to Copy It – 5-Step Playbook
List pain points – Pick one task you or your staff repeat daily (e.g. weekly promo planning, triaging customer emails, or logging field-staff notes).
Fetch the data – download last 30 days of sales, support tickets, or WhatsApp voice notes.
Pick a public LLM – Start with ChatGPT or Claude or Gemini; no fine-tuning needed.
Wrap in simple rules – “If item sales drop > 15 %, suggest a bundle”; “If email tagged refund, draft a reply”.
Ship a one-week pilot – Test with five users, measure clicks, time saved and revenue shifts.
Level | Quick build |
|---|---|
Basic | Use ChatGPT’s Advanced Data Analytics read your last 30-day sales file and output three promo ideas. |
Medium | Pipe Shopify or Square data into Zapier → Claude → Slack DM with daily suggestions e.g. “Today’s slow-moving items - try a 20 % bundle?” |
Advanced | Add speech-to-text (OpenAI Whisper) on mobile, and trigger actions in your POS via webhook. |
Why It Matters
Southeast Asia hosts 9 million+ micro-entrepreneurs who juggle marketing, finance and logistics alone. Grab’s approach—off-the-shelf LLMs plus first-party data and light rule-sets—shows startups and SMEs can bolt AI onto daily workflows without armies of engineers. As CEO Anthony Tan puts it, “Humans who don’t embrace AI will be replaced by humans who do.”
Asia Pulse
– China’s Zhipu AI opens Singapore office ahead of IPO - Reuters.
– Seoul’s new AI framework Act takes effect in 2026, blending risk tiers with startup grants.
– Oracle wins Singapore’s first military gen-AI cloud deal, building an air-gapped system that spots faces and number-plates in drone video — expect more ‘secure AI’ contracts across ASEAN.
Quick Win
Paste your last month of sales into ChatGPT and ask:
“Act as a retail analyst. Spot one under-performing item, one rising star, and craft a three-line promo for each.”
Forward Fuel
Know someone whose phone never stops buzzing with order alerts? Forward them this briefing and gift them a tiny AI sidekick.
Alex