The Brian SG Intelligence Brief — March 13, 2026: AI Agents Take Center Stage

Today's Overview

The AI world is buzzing with the expansion of "AI agents" – intelligent systems that can take actions, not just answer questions. We're seeing these agents move from specialized developer tools to everyday desktop applications and enterprise platforms, promising major boosts in productivity. This shift is driving new infrastructure investments and fierce competition between proprietary and open-source AI solutions.

Top Stories

Anthropic Launches 'Cowork' Desktop Agent for Non-Technical Users

What happened: Anthropic released Cowork, a new AI agent feature in its Claude macOS desktop application, allowing non-technical users to give Claude access to a specific folder on their computer. Claude can then read, edit, and create files within that folder to complete tasks like organizing documents or generating reports from notes.

Why it matters: This moves AI beyond chat interfaces into direct action on your computer, making advanced AI capabilities accessible to a much broader audience. Businesses can explore new ways to automate repetitive administrative tasks, potentially freeing up employee time for more strategic work.

(via VentureBeat)

Salesforce Rebuilds Slackbot as Advanced AI Agent for the Workplace

What happened: Salesforce launched an entirely redesigned Slackbot, transforming it from a simple notification tool into a powerful AI agent. This new Slackbot can search enterprise data, draft documents, and take actions across various applications, running on Anthropic's Claude large language model (LLM).

Why it matters: This is a major step in bringing "agentic AI" into the daily enterprise workflow. Companies using Slack can expect significant productivity gains as employees offload information retrieval and basic task automation to the new Slackbot, improving efficiency and reducing context switching.

(via VentureBeat)

Railway Secures $100 Million to Build AI-Native Cloud Infrastructure

What happened: Railway, a cloud platform, raised $100 million in funding to build infrastructure specifically designed for artificial intelligence applications. The company claims its platform can deploy applications in under one second, significantly faster than traditional cloud providers, which often take minutes.

Why it matters: As AI coding tools generate code rapidly, traditional cloud infrastructure can become a bottleneck. Railway's investment highlights the urgent need for faster, more cost-effective cloud solutions optimized for AI development and deployment, which could reshape how businesses build and run their AI-powered applications.

(via VentureBeat)

Creator of Claude Code Reveals Multi-Agent Workflow for Exponential Productivity

What happened: Boris Cherny, the head of Claude Code at Anthropic, shared his personal workflow, revealing he runs five Claude AI agents in parallel in his developer environment for coding tasks. He also exclusively uses Anthropic's most capable model, Opus 4.5, despite it being slower, because it requires less human correction. His team also uses a shared file (CLAUDE.md) to log AI mistakes, effectively teaching the models over time.

Why it matters: This provides a concrete blueprint for how to use advanced AI to multiply human output in software development. Business leaders in technology should observe this approach for maximizing developer productivity, emphasizing model quality over raw speed and establishing mechanisms for continuous AI learning within their teams.

(via VentureBeat)

Open-Source AI Coding Model NousCoder-14B Challenges Proprietary Systems

What happened: Nous Research, an open-source AI startup, released NousCoder-14B, a new coding model trained in just four days that reportedly matches or exceeds several larger proprietary systems on competitive programming benchmarks. The company also open-sourced its entire training environment, allowing others to replicate and extend its work.

Why it matters: The rapid advancement and openness of models like NousCoder-14B signal a tightening race between open-source and proprietary AI. Businesses can increasingly consider high-performing, free alternatives for coding assistance and other AI tasks, potentially reducing costs and increasing control over their AI infrastructure.

(via VentureBeat)

In Plain English: AI Agents

You've heard a lot about AI chatbots that can answer questions, summarize text, or generate images. But an "AI agent" is different. Think of a chatbot as a very smart librarian who can find and tell you about books. An AI agent, on the other hand, is like a highly capable personal assistant who can not only tell you about books but can also go to the library, find a specific book, read it, highlight key passages, then draft a report based on its contents, and even email it to your team – all on its own after receiving an initial instruction. Essentially, an AI agent is an AI system that can autonomously plan, execute, and verify a sequence of actions to achieve a goal. It can interact with other software, access files, use tools (like a web browser or a coding environment), and even self-correct if its initial attempts don't succeed. This ability to take action and operate independently is what differentiates an agent from a conversational AI assistant that primarily provides information. For businesses, this means AI can move beyond being a helpful query tool to becoming an active participant in workflows, automating complex multi-step tasks that previously required human oversight or intricate programming. It represents a significant shift from AI as a passive responder to an active doer.

What the Major Players Are Doing

• Anthropic: Launched Cowork, a desktop agent for non-technical users, available for Claude Max subscribers on macOS. Also made its 1 million token context window generally available for its Opus 4.6 and Sonnet 4.6 models, allowing AI to process extremely long documents or codebases (like an entire book or massive software project at once). Claude AI can now also respond with charts, diagrams, and other visuals during conversations. The company is in a legal battle with the Pentagon regarding an AI supply chain risk designation, raising questions about government use of AI for surveillance. (VentureBeat, Claude Blog, The Verge, The Verge)
• Salesforce: Rolled out a completely rebuilt Slackbot, transforming it into an AI agent powered by Anthropic's Claude model. This new Slackbot can now search enterprise data, draft documents, and take actions within Slack. (VentureBeat)
• Google: Announced that Gemini's task automation features are now available in beta on its newest Samsung and Pixel devices, allowing the AI to order food or book rides on a user's behalf through apps. (The Verge)
• Microsoft: Launched Copilot Health, a secure space within Copilot for users to ask questions about medical records, analyze wearable data, and find providers. The company also announced that its Gaming Copilot AI assistant will be coming to current-gen Xbox consoles this year. (The Verge, The Verge)
• OpenAI: Acquired Promptfoo, an AI security platform, to enhance the safety and security of AI systems. The company also introduced new ways to learn math and science in ChatGPT with interactive visual explanations and highlighted how customers like Rakuten and Wayfair are using their models for faster software delivery and improved catalog accuracy. (OpenAI, OpenAI, OpenAI, OpenAI)
• Meta: Facebook Marketplace is adding new AI-powered tools, including AI auto-replies for messages like "Is this still available?" and the ability for Meta AI to use photos to speed up item listings. (The Verge)
• Amazon (AWS): Introduced OpenClaw on Amazon Lightsail, allowing users to run autonomous private AI agents. AWS also reflected on 20 years of S3 storage innovations and released account regional namespaces for S3 general purpose buckets. (AWS, AWS)
• xAI: Faced reports of Elon Musk pushing out more founders as its AI coding effort falters, suggesting ongoing challenges in its development. (Hacker News / Financial Times)

What This Means For Your Business

Embrace AI agents for productivity across your organization. The shift from AI as a conversational assistant to an autonomous agent capable of taking actions is a significant step. Look for opportunities to deploy tools like the new Slackbot or Anthropic's Cowork to automate routine tasks, streamline information retrieval, and empower employees to focus on higher-value activities. Evaluate your current cloud infrastructure for AI workloads. As AI development accelerates, traditional cloud environments designed for a slower era may become bottlenecks. Consider specialized AI-native cloud platforms or assess how existing providers are adapting to the demands of rapid AI model deployment and processing to ensure your infrastructure can keep pace. Leverage AI for deeper customer understanding and market research. Companies like Listen Labs are showing how AI can conduct in-depth customer interviews at scale, providing rich qualitative insights rapidly. This can transform product development, marketing, and customer service by offering an honest, unvarnished view of customer needs and preferences. Explore open-source AI alternatives to control costs and data. The rapid progress in open-source models for tasks like coding (e.g., NousCoder-14B) and agentic operations (e.g., Goose) offers powerful, often free solutions. This can be a strategic move for businesses concerned about proprietary model costs, data privacy, or the ability to customize and run AI on their own infrastructure.

Quick Hits

• Listen Labs raised $69 million after a viral hiring stunt, scaling its AI-powered customer interview platform. (VentureBeat)
• A free, open-source AI agent named Goose offers similar coding functionality to Anthropic's paid Claude Code, running locally on user machines with no subscription fees. (VentureBeat)
• Steven Spielberg stated he has never used AI in his films, emphasizing human creativity over AI replacement in the arts. (TechCrunch)
• Defense officials reveal the US military might use generative AI systems to rank lists of targets and recommend strike priorities, with human oversight. (MIT Technology Review)
• Future AI chips could be built on glass, a material thousands of years old, to make next-generation computing hardware more powerful and efficient. (MIT Technology Review)