AI March Wrap-up

If you fell asleep and missed all of March, for a variety of reasons, here is a summary of important AI developments and it's like you did not miss anything at all..

AI Industry Developments – Major releases, new tools, disruptions, or advancements in AI technology.

• ​Manus AI: Initial Appearance and Hype – Launched by the Chinese AI startup Monica.lm on March 11, Manus AI was reported as a potentially fully autonomous AI "general agent" for complex business operations.
• Controversy and Open-Source Alternative – Users discovered Manus AI might be a wrapper around Anthropic’s Claude 3.7 Sonnet, leading to questions about its innovation. Despite this, it was still recognized for its capabilities.
• ​Google's Gemini 2.5:
  • Pro Version Released – Debuted on March 26 as the top AI model on the LMArena benchmark, excelling in reasoning and coding.
  • Experimental Model Publicly Accessible – By March 28, the Gemini 2.5 Pro experimental model was made public.
  • Integration with Canvas – Gemini 2.5 Pro was integrated into Google's collaborative workspace, Canvas.
• Gemma 3 – A collection of lightweight, open-source AI models based on Gemini 2.0, optimized for running on a single GPU, making high-performance AI more accessible.
• New Tools from Perplexity AI:
  • Sonar API with Live Web Searches – Now supports real-time web searches for AI applications.
  • Windows Desktop App – Launched on March 12, offering AI-powered searches with voice input and shortcuts.
  • Vision for TikTok – Proposed an advanced AI-powered answer engine for truth-seeking on TikTok.
• OpenAI Developments:
  • Enterprise AI Agent Subscriptions – Priced between $2,000 and $20,000 per month.
  • Responses API & Open-Source Agents SDK – Launched on March 11, integrating web search, file scanning, and computer automation.
  • o1 Pro API – Released on March 20 as OpenAI's most advanced model but at a significantly higher cost.
  • Next-Gen Audio Models – Speech-to-text (STT) and text-to-speech (TTS) models, improving accuracy and customizable voice outputs.
  • Hardware Experimentation – Sam Altman hinted at OpenAI working on a "cute" AI computer device by March 28.

AI in Society – Broader societal impacts, including risks, public perception, labor market effects, and cultural shifts.

• Europol warned that AI is radically modifying organized crime, making it faster, more scalable, and sophisticated, highlighting the potential risks of AI misuse. (Source: The Deep View, 19 March 2025)
• Google's aggressive rollout of AI Overviews in search, now used by over a billion people, is changing how information is accessed and potentially impacting website traffic for content creators, illustrating AI's growing influence on information consumption. (Source: The Deep View, 10 March 2025)
• OpenAI's launch of sophisticated AI agents for enterprise use (Source: AI - March 2025) with high price points and ambitions to represent a significant portion of long-term revenue signals a major shift towards AI as a core business service and a potential disruption to traditional knowledge work.

AI Use Cases & Applications in Insurance and Financial Services – How AI is being used in real-world business scenarios.

• ​NatWest, a U.K. bank, is leveraging OpenAI’s digital assistant capabilities. NatWest already uses several AI chatbots (Cora and AskArchie) for customer service and fraud reporting and has 275 AI projects, with 25 already in use. The specific details of the OpenAI integration and its current real-world impact are not detailed, but it signifies a move towards more advanced AI assistants in banking. (Source: The Deep View, Friday, March 21, 2025)
• Stripe has built custom AI-powered agents using OpenAI's tools for invoicing automation. This demonstrates a real-world application of AI agents streamlining financial processes for businesses. (Source: The Rundown AI, Tuesday, March 11, 2025)
• Box has implemented custom AI-powered agents using OpenAI's tools for searching enterprise documents. While not directly insurance or banking, this highlights the practical use of AI agents in managing and accessing information within large organizations, which could be applicable in these sectors. (Source: The Rundown AI, Tuesday, March 11, 2025)
• Delve is an AI-powered platform used for reducing manual compliance efforts for standards like SOC 2, ISO 27001, and HIPAA. This is a real-world example of AI being used for compliance in regulated industries, including potentially financial services and healthcare (relevant to health insurers). (Source: TLDR AI, Friday, 7 March 2025)
• Salesforce has launched ‘AgentExchange’, an AI-powered marketplace offering pre-built AI tools designed for enterprise automation and digital labor. This indicates the availability and real-world adoption of AI tools that could be used by insurance and financial services companies to automate workflows and enhance customer engagement. (Source: The Rundown AI, Tuesday, 11 Mar 2025)
• ServiceNow acquired Moveworks, an AI-powered conversational automation platform. This acquisition by a major enterprise software leader highlights the real-world value and increasing adoption of AI for internal and external communications within businesses, including potentially financial institutions and insurers. (Source: The Rundown AI, Tuesday, 11 Mar 2025)
• Send Technology provides AI agents for underwriting that streamline decision-making and automation in insurance workflows. This is a specific real-world application of AI directly within the insurance underwriting process. (Source: "Eye on AI | Fortune", Tuesday, 11 Mar 2025)
• Trustpilot uses an AI-driven reviews system (in the UK) to analyze and manage customer feedback. While not specific to insurance or banking, it showcases the real-world implementation of AI for enhancing trust and transparency for businesses interacting with customers, a concept relevant to financial services. (Source: "Eye on AI | Fortune", Tuesday, 11 Mar 2025)
• Addresscloud is an AI-driven risk assessment company that provides insights into 150M properties and 20 natural hazards. Insurance companies are likely using such services for more accurate risk modeling and underwriting. (Source: "Eye on AI | Fortune", Tuesday, 25 Mar 2025)
• JBA Risk Management is integrating AI-driven weather models powered by NVIDIA's Earth-2 for improved flood risk assessments. This is a direct application of AI in enhancing risk modeling for the insurance industry. (Source: "Eye on AI | Fortune", Tuesday, 25 Mar 2025)
• Cowbell Cyber is expanding cyber insurance solutions, leveraging AI and data analytics. This illustrates the real-world use of AI to enhance and innovate within the cyber insurance market.
• Liberate is an AI-driven system used for disaster and large-scale claims management, demonstrated by handling 14X normal claim volumes after back-to-back hurricanes. This showcases the real-world capability of AI to manage surges in claims processing during catastrophic events. (Source: Insurance Journal Research & Trends, Fri, 21 Mar 2025)
• EvolutionIQ uses AI to reduce claim review times in insurance by over 12 hours. This is a quantifiable real-world benefit of AI implementation in claims handling. (Source: Insurance Journal Research & Trends, Fri, 21 Mar 2025)
• Payman AI platform advances AI in Finance & Payments indicating a shift towards autonomous financial transactions. While specific real-world integrations in established financial institutions are not detailed, its existence points to real-world development in AI-driven financial processing. (Source: Unwind AI, Wed, 19 Mar 2025)5)
• Perplexity Enterprise Pro with Softbank: The specific product being launched is Perplexity Enterprise Pro. This suggests a premium version of Perplexity's AI-powered search and information retrieval service tailored for business use.

AI Research & Development – Notable research papers, breakthroughs, or advancements in AI models.

• A study by METR found that AI capabilities are doubling every 7 months in terms of the length of tasks they can autonomously complete, suggesting a much faster pace of AI advancement than traditional Moore's Law. (Source: The Rundown AI, 20 March 2025)
• Alibaba released R1-Omni, a multimodal model capable of understanding visual and audio context to read emotions, representing progress in AI's emotional intelligence capabilities. (Source: The Rundown AI, 13 March 2025)
• Researchers at the University of Oxford introduced Agentic Reasoning, enhancing LLMs by integrating AI agents for external tasks and achieving PhD-level performance on complex science benchmarks, demonstrating significant advancements in AI reasoning. (Source: Unwind AI, 11 March 2025)
• Mistral AI released Small 3.1, an open-source multimodal AI model outperforming Gemma 3 and GPT-4o Mini with a fast inference speed, highlighting the rapid progress in efficient and high-performing open-source AI. (Source: TLDR AI, 18 March 2025)

​

​

​Unsubscribe · Preferences​

​