Get tomorrow's brief in your inbox
Today: Google and OpenAI both launched AI agents that can actually complete work tasks across your business tools, Chrome got an AI intern that handles tedious web tasks, and Meta started recording employee keystrokes to train AI models on how humans work.
AI agents just went from hype to deployment. Both Google and OpenAI shipped production-ready AI agents this week that run autonomously in the cloud and complete multi-step workflows across your business tools. OpenAI's workspace agents (available in Business, Enterprise, and Edu plans) can gather product feedback from the web and send Slack reports, draft follow-up emails in Gmail, and execute other tasks without you watching them. Google's version, called "auto browse" for Chrome Enterprise, handles web-based work like inputting CRM data from Google Docs, comparing vendor pricing across tabs, and pulling competitor product details. Both require human approval before final actions, but they're designed to run while you're doing other work.
This is different from chatbots that answer questions. These agents execute tasks. If you run a service business, you could set up an agent to monitor client inquiries, draft responses, and queue them for your review. If you manage vendor relationships, an agent could track pricing changes across multiple supplier sites and alert you to better deals. The barrier to entry is low: if you already use Google Workspace or ChatGPT for teams, these features are rolling out now. OpenAI says its GPTs will convert into workspace agents soon, so any custom chatbots you built can become task runners. Start identifying repetitive workflows that span multiple tools (email, spreadsheets, CRM, calendars). Those are your best candidates for agent automation.
Google redesigned Workspace into an AI productivity suite. The update brings AI directly into Sheets, Docs, and Meet. In Sheets, you can now prompt Gemini to build formatted spreadsheets, retrieve data, and auto-fill cells based on context (Google claims 9x faster than manual entry). In Docs, Gemini generates, writes, and refines content using your Drive, Gmail, and Chat history, and can match your writing style. For meetings, Gemini now takes notes for in-person meetings and Zoom/Teams calls, not just Google Meet. You start recording from the Meet app, and it creates a summary with action items in Google Drive. All of this is available to Workspace users now. Read more
OpenAI partners with Infosys to deploy Codex at enterprise scale. Infosys (one of India's largest IT services firms) integrated OpenAI's Codex coding assistant into its Topaz AI platform, bringing AI-powered software development to its global client base. The partnership includes engineers from both companies working directly with clients to implement Codex for legacy modernization, DevOps, and software engineering workflows. OpenAI also launched Codex Labs, a program where its engineers help deploy Codex alongside partners like Accenture, Capgemini, Cognizant, PwC, and TCS. Codex now has 4 million weekly active users. If your business works with any of these consulting firms, expect Codex pitches soon. The play here is moving AI coding from individual developer tools to organization-wide deployment. Read more
Google launched two new AI chips designed for agents. TPU 8i is optimized for inference (the fast responses agents need to feel responsive), while TPU 8t handles training for large models. Google claims 3x faster training, 80% better performance per dollar, and the ability to cluster over 1 million TPUs together. These chips supplement (not replace) Nvidia GPUs in Google Cloud, but they signal where Google thinks AI workloads are headed: autonomous agents that need to respond instantly. If you're building on Google Cloud or considering it, these chips could reduce your inference costs significantly compared to Nvidia-based options. Read more | Google announcement
Qwen released a 27B coding model that rivals 397B models. Qwen3.6-27B delivers flagship-level coding performance in a 16.8GB quantized model you can run locally. It outperforms Qwen's previous 397B model (807GB) across major coding benchmarks. If you're experimenting with local AI coding tools, this is a massive efficiency gain. You can download it from Hugging Face and run it with llama.cpp. The performance is genuinely impressive for local inference (25 tokens/second on standard hardware). Read more
32% of B2B buyers now discover vendors using AI chatbots. That number is only growing, which means if your business expertise isn't indexed by AI answer engines (ChatGPT, Gemini, Perplexity, Claude), you're invisible to a third of potential customers. HubSpot published a detailed guide on Answer Engine Optimization (AEO) for B2B. The tactics overlap with SEO but prioritize structured content that AI can summarize and cite. Key actions: create FAQ-style content that directly answers buyer questions, publish pricing information openly (buyers use AI to compare vendors and can switch in 15 minutes if they can't find your pricing), and structure content with clear headings and definitions. One buyer interviewed for the study used Gemini Deep Research to identify five new SaaS vendors and switch providers based on AI-generated shortlists. If your site doesn't have clear, crawlable pricing and capability descriptions, you're losing deals to competitors who do. Read more
Google showcased startups building on its cloud AI stack. Notable names include Lovable (the vibe-coding startup on a $400M ARR track), Notion (using Gemini for text/image generation), Gamma (the AI PowerPoint alternative valued at $2.1B), and ComfyUI (open source image generation tool). Google also announced a $750M partner fund to help consulting firms and startups sell AI agents to enterprises, covering proof-of-concept costs, cloud credits, and deployment rebates. If you're building an AI-powered product, Google is actively funding go-to-market support through this program. Read more
OpenAI made ChatGPT free for U.S. clinicians. Verified physicians, nurse practitioners, and pharmacists can now access ChatGPT for Clinicians at no cost for clinical care, documentation, and research. This is a significant move to build adoption in healthcare. If you run a medical practice or healthcare business, this is free access to a tool that can draft patient notes, summarize research, and assist with clinical documentation. Read more
SpaceX agreed to acquire Cursor for $60 billion. Microsoft explored buying the AI coding startup before SpaceX (now merged with xAI) swooped in with a $60B deal (or $10B if the deal doesn't close by year-end). Cursor had been raising at a $50B valuation. This is Elon Musk consolidating AI coding tools under the SpaceX/xAI umbrella as the company prepares for a record IPO. For context, Cursor competes directly with GitHub Copilot (Microsoft), OpenAI's Codex, and Anthropic's Claude Code. The coding tool market is now a battleground for the biggest tech players. If you're evaluating AI coding assistants, expect rapid feature development and aggressive pricing as these companies fight for market share. Read more
Meta is tracking employee computer activity to train AI agents. Meta installed a tool called Model Capability Initiative (MCI) on U.S. employees' work computers that records mouse movements, clicks, keystrokes, and screenshots to train AI models on how humans interact with software. The goal: teach AI agents to automate work tasks by learning from real human workflows. Employees cannot opt out on company-provided laptops. This sparked internal backlash, but Meta CTO Andrew Bosworth defended it, saying the vision is "agents primarily do the work and our role is to direct, review and help them improve." The data isn't used for performance reviews (according to Meta), but it signals where AI development is headed: learning from watching humans work, then replacing those tasks with agents. Read more
CISA reportedly doesn't have access to Anthropic's Mythos cybersecurity model. While the NSA and Commerce Department are testing Anthropic's powerful security-focused AI (designed to find vulnerabilities in operating systems and browsers), the Cybersecurity and Infrastructure Security Agency was apparently left out. This is concerning because CISA coordinates cyber defenses for state/local governments and critical infrastructure. The Trump administration has been cutting CISA's budget and staff, and this signals the agency isn't prioritized for access to cutting-edge defensive tools. If you rely on CISA for threat intelligence or vulnerability alerts, this could slow their ability to protect critical systems. Read more
Set up an AI agent for a repetitive workflow this week. Pick one task that spans multiple tools: pulling data from emails into spreadsheets, summarizing Slack threads into status reports, or tracking competitor pricing across websites. If you use Google Workspace, try the new Sheets auto-fill or Chrome auto browse features. If you have ChatGPT Business/Enterprise, create a workspace agent. Start small: one workflow, human review required. Measure time saved. If it works, expand to similar tasks.
Audit your website for AI discoverability. Open ChatGPT, Gemini, or Perplexity and ask: "What does [your company] do? What are their prices? Who are their competitors?" If the AI can't answer accurately, your content isn't structured for answer engines. Add an FAQ page, publish clear pricing, and use structured headings. A third of B2B buyers are using AI to find vendors. Make sure they can find you.
The shift from AI assistants to AI agents is accelerating faster than expected. Google and OpenAI both shipped production-ready agents this week, and Meta is training models by watching employees work. The pattern is clear: 2026 is the year AI moves from answering questions to completing tasks. Businesses that identify repetitive, multi-step workflows and automate them with agents will gain efficiency advantages. Those that don't will face competitors who work faster with smaller teams. Meanwhile, the Cursor acquisition battle shows that AI coding tools are now strategic assets worth tens of billions. Watch for consolidation in this space and aggressive feature competition. If you're evaluating AI coding assistants, lock in pricing now before these platforms raise rates.