How Asia’s Marketers Strike Gen AI ROI First

Here is what’s new in the AI world.

53% Don’t Lie: Asia’s Gen AI ROI Exposed in Bombshell Report

A 2025 study by Adobe and Econsultancy reveals that marketers in Asia are ahead of the global curve, with over half (53%) already generating revenue returns from generative AI despite facing data fragmentation and budget constraints.

The report, surveying 3,400 executives globally (predominantly client-side) between November and December 2024, highlights Asia-Pacific (APAC) as the largest respondent region (37%), followed by Europe and North America.

Key industries represented include retail, B2B tech, financial services, healthcare, and media/entertainment. Senior executives in South Korea, Hong Kong, Singapore, and Southeast Asia are identified as regional leaders in applying AI to strategic marketing functions like customer journey orchestration and content creation.

However, significant challenges persist. Only 6% of organizations have implemented measurable, scaled AI solutions, indicating a gap between optimism and operational reality.

Data silos remain the primary barrier to achieving real-time personalization, with just 9% rating their digital customer experience as "exceptional" and fewer than half using analytics to proactively anticipate customer needs.

Privacy and security concerns (cited by 43% of leaders) are the biggest obstacle to cross-departmental data integration. Alarmingly, 28% of respondents reported that their leadership doesn't even view data as a strategic asset, the highest rate in the JAPAC region.

Furthermore, 45% of senior leaders struggle to secure budgets and prove ROI at scale, while internal pressure mounts as 87% anticipate demands for faster, higher-volume content in 2025. Adding to the strain, 69% of practitioners feel overwhelmed by the influx of new technology.

Looking ahead, investment priorities focus on overcoming these hurdles: 61% of executives plan to spend on data integration and real-time insights over the next 1-2 years, followed closely by AI/ML capabilities (57%).

Notably, Asian practitioners report seeing AI-driven ROI across a broader range of use cases (including journey orchestration at 16%, creative production like virtual photoshoots at 14%, and chat/customer support at 14%) compared to other JAPAC markets, where returns are concentrated mainly in chat/service.

To manage the workload and automate routine tasks such as database management and content delivery, many are turning to agentic AI embedded in copilots and assistants. Adobe emphasizes that combining this agentic approach with generative AI offers the promise of scalable personalization and reduced support costs.

Zero Planning Fatigue: Gen AI Crafts Dream Trips While You Scroll

Artificial Intelligence (AI), particularly generative AI (Gen AI) and agentic AI, is revolutionizing how travel companies help travelers confidently discover and book experiences tailored to their needs.

Accenture's 2025 Consumer Pulse Survey reveals that for active Gen AI users (those using it weekly), Gen AI has become the primary channel for travel discovery, surpassing social media and Online Travel Agencies (OTAs).

This technology presents a major opportunity for the industry to fundamentally reimagine the traveler journey. Adoption is widespread, with 80% of travelers across airlines, hotels, and platforms now using Gen AI tools, signaling that these users are the new mainstream, not just early adopters.

Personalization Drives Demand and Loyalty

Travelers overwhelmingly seek control and personalization: 86% want to shape their own experiences, and this is critically important for building a personal connection with the brand for 93% of active Gen AI users. Trust in AI for decision support is high, with 93% of active users having used or willing to use Gen AI to aid purchasing decisions.

Furthermore, 78% are open to using a trusted AI-powered personal shopping assistant that understands their needs, and 57% specifically desire an assistant that works seamlessly across multiple brands to autonomously solve problems.

The payoff for delivering personalized, emotionally engaging experiences is significant: travelers are 1.3 times more engaged and 1.7 times more likely to accept a higher price from providers who achieve this. Reinforcing the need for personal recognition, 79% of travelers want brands that make them feel special by remembering them personally.

Industry Perspective: Solving Overload and Creating Joy

Emily Weiss, Accenture’s Global Travel Lead, emphasizes AI's impact: "In just a year, AI is already tackling the problems of choice overload and complex decision-making, creating a real opportunity to restore enjoyment to planning and booking memorable trips."

She notes that "Gen AI takes this further by enabling highly tailored experiences that feel more human and natural."

Weiss highlights the broader potential: "For travel, the AI opportunity extends beyond securing bookings. Instead of travelers drowning in options and conflicting reviews, Gen AI can act as a personal travel concierge.

It can provide bespoke recommendations based on preferences, budget, location, past travel history, loyalty status, and even real-time data on local events and attractions."

How Alibaba’s Embedding Models Power the Next AI Wave

Alibaba Group has released its Qwen3 Embedding model series, reinforcing its commitment to global open-source AI leadership. Ranked third globally for large language models (LLMs) by Stanford's 2025 AI Index Report, Alibaba aims to solidify its position with this release.

Key Capabilities:

Embedding Model Purpose:

These models convert text into numerical representations, enabling computers to understand semantic meaning beyond simple keywords. This leads to more accurate and relevant results in tasks like search and retrieval.

Technical Development:

The Qwen3 series uses Alibaba's established three-stage training approach:

1. Contrastive Pre-training: Trains on vast raw data to distinguish relevant information.

2. Fine-tuning: Uses curated, high-quality data to refine capabilities.

3. Integration: Combines learnings for enhanced overall performance.

Alibaba states these models will optimize its foundational Qwen model, improving the efficiency of its embedding and search reranking systems (which refine result ordering based on user queries). The company positions this release as a "new starting point" and encourages developers to implement the open-source models across diverse applications.

No More ‘Dumb’ Robots: SmolVLA’s AI Magic Just Dropped on GitHub

Hugging Face has released SmolVLA, an open-source Vision-Language-Action (VLA) model designed to accelerate robotics development. Unlike proprietary VLA models from major tech firms, SmolVLA addresses key industry bottlenecks by being:

1. Open-source: Provides full access to weights, datasets, and training code.

2. Hardware-efficient: Runs locally on a single consumer GPU or newer MacBook (450M parameters).

3. Open-data trained: Uses LeRobot community datasets, enabling broader research collaboration.

Despite AI advancements, robotics progress has lagged due to proprietary models and scarce high-quality data. SmolVLA empowers researchers to reproduce, customize, and build upon state-of-the-art VLA capabilities without restrictive resources.

Technical Innovation:

Capabilities:

SmolVLA can interpret real-world scenes via camera input, understand natural language instructions, and generate precise actions for robotic hardware. Hugging Face claims it outperforms larger models in efficiency benchmarks.

Accessibility:

Available immediately for download, SmolVLA enables:

• Researchers to advance open robotics workflows.

• Enthusiasts with robotic arms to test real-time AI control.

• Democratizing development beyond well-funded labs.

OpenAI Lands in India: Academy Launch to Train 100,000 AI Coders by 2027!

OpenAI has partnered with the Indian government's IndiaAI Mission to establish OpenAI Academy India, its first educational platform outside the United States. This initiative aims to significantly broaden access to AI skills training across India.

Key Objectives:

• Accessibility: Provide AI learning tools to India's developers, startups, innovators, digital platforms, educators, students, civil servants, nonprofits, and small businesses.

• Supporting IndiaAI: Directly contributes to the IndiaAI Mission's "FutureSkills" goal of nationwide AI literacy.

• Building Capacity: Empower individuals and organizations to create faster, scalable AI solutions.

Program Features:

Leadership Statements:

• Ashwini Vaishnaw (Union Minister): Emphasized that the initiative aligns with PM Modi's vision for accessible advanced technology, enabling faster innovation by startups, developers, and researchers.

• Jason Kwon (OpenAI CSO): Highlighted India's rapid AI progress and stated the collaboration aims to equip people with the skills and confidence to use AI meaningfully in their work and communities.

Your opinion matters!

Hope you loved reading our piece of newsletter as much as we had fun writing it. 

Share your experience and feedback with us below, ‘cause we take your critique very seriously. 

Thank you for reading

-Shen & Towards AGI team