• AI KATANA
  • Posts
  • Emerging markets embrace AI trust

Emerging markets embrace AI trust

Also: The top 10 enterprise generative AI applications

Author

AI KATANA
April 28, 2025

Hello!

Today’s newsletter brings a sharp global perspective on how emerging markets are outpacing advanced economies in AI trust and adoption, with countries like Indonesia, India, and Brazil showing far greater optimism toward AI’s future compared to the US, Japan, and Germany. Meanwhile, a massive real-world survey of enterprise generative AI deployments reveals customer service and marketing as the leading applications, dominated by North American tech firms. In the corporate world, Apple is quietly reshuffling its AI teams to regain competitive momentum ahead of a crucial developer event. OpenAI, on the other hand, is pushing ChatGPT into e-commerce by enabling in-chat product research, challenging Google’s dominance in shopping discovery. Alibaba has released its Qwen3 open-source model series, emphasizing hybrid reasoning for more efficient task handling, while Bloomberg reports growing investor anxiety over the staggering capital expenditures fueling the AI boom. Plus, enterprise writing platform Writer has launched a leaner LLM model promising major cost and compliance benefits, and infrastructure player Nscale is pursuing a major $2.7 billion raise to strengthen its competitive position in AI data centers.

Sliced just for you:

  • 🌏 Emerging markets embrace AI trust

  • 🏢 The top 10 enterprise generative AI applications

  • 🍏 Apple reshuffles internal AI team

  • 🛍️ ChatGPT becomes a shopping concierge

  • 🤖 Alibaba Introduces Qwen3, Setting New Benchmark in Open-Source AI with Hybrid Reasoning

  • 💸 Markets question relentless AI capex

A 47-nation study finds optimism about AI far higher in Indonesia, India and Brazil than in the US, Japan or Germany. Two-thirds of respondents now use AI regularly and 83 percent expect it to boost healthcare, education and wages. Yet only 40 percent in advanced economies trust autonomous services versus 61 percent in emerging ones. Researchers warn this “confidence gap” could hard-wire geopolitical blocs: countries seeing AI as a growth engine are racing to adopt, regulate and export it, while richer nations move more cautiously over privacy and job-loss fears. Policymakers are urged to share safety frameworks or risk divergent standards that fragment global markets.

Based on an analysis of 530 real-world deployments, enterprise adoption of generative AI is most prominent in customer support, marketing, IT, operations, and R&D, with customer issue resolution leading as the primary application, comprising 35% of projects. Technology companies and North American organizations dominate the landscape, each representing 56% of implementations. Key use cases include automating customer service inquiries, streamlining post-sale support, accelerating content creation for marketing, assisting software development, optimizing operational processes, enhancing IT support, innovating product design, expediting prototyping, and conducting feasibility studies.

Apple quietly shifted major parts of its AI division away from a high-profile executive recruited from Google in 2018. Sources say core teams building on-device foundation models and EdgeTPU-style inference engines now report directly to software chief Craig Federighi ahead of WWDC. The restructuring follows criticism that Apple’s “Apple Intelligence” suite lags rivals in multimodal reasoning and code generation. Observers expect a Gemini-class model preview and tighter silicon-software co-design disclosures at June’s keynote.

OpenAI added a commerce workflow that lets users research products directly inside ChatGPT. Typing “running shoes under $120 for flat feet” now yields curated lists that cite professional reviews and Reddit threads; checkout still happens on retailers’ sites, but OpenAI plans affiliate experiments. Early testers say the conversational filter feels less ad-driven than Google Shopping, though questions remain about brand influence and disclosure. The move positions ChatGPT as a search alternative where discovery and purchase intent converge.

Alibaba has unveiled Qwen3, its latest open-sourced large language model family, introducing a groundbreaking hybrid reasoning approach that dynamically toggles between thinking and non-thinking modes to optimize task performance and efficiency. The Qwen3 series includes six dense models and two Mixture-of-Experts (MoE) models, now globally available, with parameter sizes ranging from 0.6B to 235B. Trained on 36 trillion tokens, Qwen3 delivers significant advancements in reasoning, instruction following, multilingual support across 119 languages, and agent-based capabilities, while offering cost-effective deployment through models like the Qwen3-235B-A22B.

On Bloomberg Markets, analysts debated whether the “AI super-cycle” can justify Alphabet’s $75 billion capital-expenditure plan and similar splurges by Meta and Microsoft. Investors worry returns will taper as memory and power costs rise, yet bulls argue that proprietary data advantages and emerging agent ecosystems will unlock new revenue layers. The discussion highlighted a growing divide between software valuations—still soaring—and hardware suppliers, where margins are under pressure despite record demand.

🛠️ AI tools updates

Enterprise writing platform Writer released a new generation of its Palmyra-family LLM that slashes inference costs by 40 percent and posts a 15K context window. Benchmarks show parity with GPT-4o on summarization and style-transfer tasks at roughly one-third the token price. Writer says the model’s smaller footprint allows full on-prem deployment for compliance-sensitive clients, and it introduced an optional fact-verification layer that flags uncited claims in drafts.

💵 Venture Capital updates

Fresh filings reveal GPU cloud rival Nscale is courting ByteDance as an anchor customer while pursuing a combined $2.7 billion in debt and equity. Insiders say proceeds would fund two 500-MW data-center campuses optimised for Blackwell-class accelerators plus the design of a proprietary interconnect pitched as cheaper than InfiniBand. If successful, the raise would value Nscale near $12 billion and intensify competition with CoreWeave for AI training workloads.

🫡 Meme of the day

⭐️ Generative AI image of the day