SPEAKER_00 0:00

Welcome back to Sumo Growth AI Insights. I'm Darius, and 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 MasterCard opens agent pay for machines so AI agents can pay for services on a budget. On Tuesday at Cloudflare Connect, MasterCard opened Agent Pay for Machines Early Access, giving developers buyer and seller APIs, a toolkit, a sandbox, and documentation to build agent-to-agent commerce. It builds on X402, an open protocol that lets a paid service reply 402 payment required. The agent then answers with a digitally signed payment voucher that stays inside the budget and permissions a business approved. Providers verify the voucher, deliver the service, and can combine vouchers to redeem later. The launch follows pilots with Cloudflare, Coinbase, Base, Ant International, AISA, BVNK, and Hype, and MasterCard says it's still a pilot, with no date for general availability. In the demo, marketing agency Hype gave an agent a defined budget to plan a protein bar launch campaign, and the agent paid AISA for social media research and Cloudflare AI gateway for campaign images, without a separate billing setup for each. For agencies, this points to agents that buy data and tools per task, instead of needing an API key and prepaid credits for every service. That makes spending rules part of the build. Plus Story OpenAI brings its models and codecs to Google Cloud, putting GPT on all three major US clouds. On Wednesday, OpenAI and Google Cloud said they're bringing OpenAI's Frontier Models and Codec'Agentic workflows into Google Cloud's model garden, purchased through the Google Cloud Marketplace. Enterprises with Google Cloud spend commitments will be able to draw them down on OpenAI purchases, and OpenAI models will also come to Google Anti-Gravity and the Gemini agent. General availability is planned in the coming weeks, and OpenAI says Microsoft remains its primary cloud partner. CNBC notes the move puts OpenAI's models on all three of the top US cloud providers, less than six months after Microsoft's exclusive license ended and Amazon started hosting them. For agencies with clients on Google Cloud, this removes a procurement hurdle. GPT models can now ride on the same bill and governance controls those clients already use. Three more headlines. First, the U.S. ordered federal civilian agencies to remove Deep Seek products. DHS ordered civilian executive branch agencies to remove DeepSeq products, affiliated products, and third-party services built using DeepSeq from their IT systems and to stop buying them. The Department of Defense is excluded, and agencies can seek written exceptions or compliance delays within 60 days. Second, Minimax launched M3.1 Flash at 15 cents per million input tokens. Minimax says its new 1 million token context model cuts agent task costs 70% versus its predecessor and runs nearly three times faster, with output priced at 60 cents per million tokens. VentureBeat flags a high hallucination rate, and Minimax says the weights will be released later. Third, otter.ai raised $320 million from General Catalyst's Customer Value Fund. Otter says the non-dilutive financing will go to enterprise growth, go-to-market, international expansion, and product work. The company says 40 million people have used its meeting note taker, up from 25 million last year. Tools of the Day.im is an AI design agent that turns a one-sentence idea into real, editable print files like stickers, gift boxes, and brand kits. Needle at Needle.app is an agent-native outbound engine that finds, enriches, and reaches your leads from one prompt in Claude, ChatGPT, or Grok. The library, from Query at Query.com, organizes your business context from over 100 platforms so every team gets repeatable answers in plain English. Business AI tip of the day. Give every AI agent a budget before you give it a wallet. MasterCard's Agent Pay for Machines shows where client automation is heading. Agents that pay for data, models, and tools on their own, one task at a time. The agencies that win will set the spending rules before a client ever asks. So set up a simple agent spend policy for each client this week. List every paid service each agent touches today, from lead enrichment and search to image generation and model calls. Give each agent a per-task budget and a monthly cap and write both into the client's statement of work. Require human approval for any new vendor and for any single purchase above a set amount. Run agent spending through one card or account per client so it's easy to trace and easy to shut off. And log every agent purchase in Sumo AI CRM or GoHigh Level against the client record and show it in your monthly report. Clients will let agents spend money once they can see the limits. Make the limits part of what you sell. Pro tip Sign up for MasterCard's early access sandbox and run one test workflow with a tiny budget. You'll understand the voucher flow before your competitors do, and you'll have a demo ready for your next client call. That's your Sumo Growth AI Insights briefing. Set those budgets this week, and I'll catch you on the next issue.