SPEAKER_00 0:00

Welcome back to Sumo Growth AI Insights. I'm walking you through this week's AI news the way an operator actually uses it, not as a pile of headlines. Lead story. OpenAI launches dots. Always on agents powered by GPT6 Astra. OpenAI used Dev Day 2026 on September 29th to launch Dots. Always on agents that keep working after you close the chat. Each dot runs on GPT-6 Astra with its own cloud computer and browser, connects to more than 4,000 apps through plugins, and can message you in ChatGPT, Slack, or Teams. OpenAI says Dots ask for approval on sensitive actions and support custom rules for what they may do alone. Your first dot is included on Pro and Business Premium Plans in eligible markets with enterprise access through an admin-enabled beta. Specialist dots for org roles and a Microsoft Agent 365 integration are in early preview. For operators, this is the moment always on agents go mainstream, which means your clients are about to ask why their CRM bot can't do the same thing and what stops it from doing something it shouldn't. Plus story: Trump and AI CEOs sign a morally binding White House Superintelligence Accord. President Trump and House Speaker Mike Johnson hosted a White House lunch on September 29th, where major AI leaders signed the White House Accord on Superintelligence, a voluntary pledge Trump called almost like a constitution and morally binding. Signers included Anthropic CEO Dario Amodet, Google CEO Sundar Pichai, Meta CEO Mark Zuckerberg, OpenAI president Greg Brockman, Nvidia CEO Jensen Huang, and Elon Musk. Zuckerberg said the deal centers on robust internal controls, detection, layered auditing, and board review of auditor reports. Trump floated a roughly 10-person committee to watch the industry and said he sees tremendous self-policing. The same day he signed an executive order directing federal agencies to say superintelligence or SI instead of AI in official non-statutory communications. The operator read, the rules of the road are voluntary, built on internal controls and audits, and that is exactly the language your enterprise clients will start borrowing. Three more headlines. First, Anthropics IPO prospectus shows deep big tech cloud dependence. A Reuters exclusive on Anthropics confidential IPO filing says 47% of last year's sales were routed through Amazon and Google, with about $351 million in distribution fees, roughly 16 cents per dollar of those sales. Two unnamed customers each drove 12% of revenue, and long-term compute commitments exceeded $417 billion, covering 3.5 gigawatts of dedicated capacity. Second, the Trump executive order tells federal agencies to say superintelligence, not AI. The September 29th order directs executive departments to use superintelligence and SI in place of artificial intelligence and AI in official non-statutory communications, with a proposed federal definition due in 60 days. Third, OpenAI shipped GPT 6.1 Sol and Astra ultra fast at Dev Day. Alongside DOTs, OpenAI launched GPT 6.1SOL, a lower cost model aimed at coding and professional work near Astra level performance, plus an ultrafast speed tier for Astra and Sol. Tools of the day.com lets you run LinkedIn ads and account-based marketing from Claude or ChatGPT with Xena, including bids, budgets, reporting, and CRM Sync without living in campaign manager. Lucy Desktop at Lucy.memories.ai gives Claude code, cursor, and codex local screen and meeting memory so agents recall what you saw without endless paste threads. iFixAI at iFixai.ai independently audits AI agents for misalignment and operational assurance, then ships a graded report clients can understand. Business AI tip of the day sell always-on agents with written approval rules. OpenAI just put always-on dots in ChatGPT, Slack, and Teams, and the White House got labs to sign a voluntary superintelligence accord built on internal controls and audits. Clients will not buy autonomous until you can show who approves what. For every sumo AI CRM, Go High level, or client agent you ship this week, write a one-page custom rules card covering what the agent may do alone, what needs human approval, and what is blocked. Start proactive work as read-only, things like CRM lookup, calendar check, and research before you allow send, book, or file write. Name a kill switch owner who can revoke keys or pause the agent in under five minutes. Log every outbound action, who, tool, and target, and review the log before you expand autonomy. Use Zen ABM patterns for LinkedIn ads. AI can draft and optimize, but budget and go live wait for a human confirm. Use iFix AI or a similar audit before you tell a retainer client the agent is production ready and keep a backup model path so a dot or astra outage does not kill a live campaign. Agencies that package approval rules, audit trails, and a named owner will sound enterprise ready. Agencies that only demo cute always-on mascots will sound risky. Pro tip: offer each retainer client a free always on agent safety card this week, listing tools allowed, approval gates, kill switch owner, and last audit date. Bundle a 15 minute stack review if they add one scoped automation this month. That is your Sumo Growth AI Insights briefing. Write the approval rules this week, and I will catch you on the next issue.