Wayfair is joining at limited scale, and Angi launched a homeowner-facing agent, alongside Newegg, Best Buy, Lowe’s, and VistaPrint. Angi’s branded agent surfaces in ChatGPT conversations about home services and routes homeowners into Angi’s service-request flow. “Our AI-first strategy is simple: make it easier for homeowners to connect with and hire the right pro for their project,” said Angi co-founder Angie Hicks. “With OpenAI’s Sponsored Agent pilot, homeowners can go directly from discussing a home project in ChatGPT to connecting with a skilled local pro who can help.”

Alongside the pilot, OpenAI opened an Ads Manager plugin that lets advertisers create, update, and analyze campaigns with natural-language prompts — including turning a website or campaign brief into a campaign, reviewing performance, and receiving recommendations. It can suggest copy and images from a landing page, tailor headlines to the context of a conversation, and translate them into a user’s preferred language. HubSpot is OpenAI’s first CRM partner, and Shopify is its first e-commerce partner; the Shopify app goes international starting September 23.

Why it matters: This is the biggest change to ChatGPT’s interface since it launched: answers now come with a paid salesperson one click away. For users, the labeling matters — sponsored agents are marked and kept separate from independent answers. For small businesses, it’s a brand-new channel: a store can show up inside the conversation where shoppers are already asking questions.

• Sponsored Agents pilot began September 16, 2026, with select U.S. advertisers
• Launch brands include Wayfair, Angi, Newegg, Best Buy, Lowe’s, and VistaPrint
• Agents are clearly labeled and kept separate from ChatGPT’s independent responses
• ChatGPT Ads Manager plugin builds campaigns from natural-language prompts, a website, or a brief
• HubSpot is OpenAI’s first CRM partner; Shopify is its first e-commerce partner, going international September 23

Sources: https://www.reuters.com/business/media-telecom/openai-tests-advertiser-sponsored-agents-expands-ai-tools-chatgpt-ads-2026-09-16/, https://aiweekly.co/alerts/openai-pilots-sponsored-agents-in-chatgpt-with-wayfair-angi, https://www.globenewswire.com/news-release/2026/09/16/3363209/0/en/angi-among-first-brands-to-pilot-sponsored-agents-in-chatgpt.html

The abilities extend to all devices in the Google Home ecosystem, including Nest devices, Matter devices, and anything labeled “Works with Google Home,” according to Android Police. Connected agents can control devices, analyze summaries, build custom dashboards, and even use a Google Home speaker as the agent’s “voice.” Google Home product manager Taylor Lehman called it “just the next step in unlocking agentic smart homes.” Setup requires creating a Google Cloud project and giving the agent the MCP configuration; the agent then asks the user to sign in and grant permissions.

There are catches. Right now, access is available only to customers in the US with Google Home’s $20-per-month Premium Advanced subscription, and Google says it’s “rolling out” over roughly one to two weeks. And Google warns that connecting a third-party agent can produce “unexpected or even undesired behavior” — sensitive actions like unlocking doors are prohibited outright.

Why it matters: This is the first time Google has given rival AI assistants a standardized way into one of the biggest smart-home ecosystems. For Google Home owners who live in ChatGPT or Claude, it means one assistant could finally run the house. The paid gate and the security warnings show Google is moving cautiously — but the door is open, and competing assistants will want to walk through it.

• Google announced the official Google Home MCP server on September 16, 2026
• Compatible agents include ChatGPT, Claude, Hermes, and OpenClaw
• Covers Nest, Matter, and “Works with Google Home” devices
• Agents can control devices, analyze summaries, build dashboards, and use a Google Home speaker as a voice
• Limited to US customers with the $20/month Google Home Premium Advanced subscription; rollout over roughly 1–2 weeks
• Door unlocking and other sensitive actions are prohibited; Google warns of possible unexpected behavior

Sources: https://www.androidpolice.com/google-home-opens-the-door-to-chatgpt-claude-and-other-ai-agents/, https://www.explainx.ai/blog/google-home-mcp-claude-chatgpt-2026

The update also brings Claude Design, the visual tool introduced in April, directly into every conversation, and launches two new products in beta: Claude Docs and Claude Slides. Users can co-create documents — adding sections, asking questions, commenting on completed parts — and build presentations, then export documents to Google Docs or Microsoft Word and presentations to PowerPoint or PDF. Finished work can be shared with a link, and a document started on desktop can be tracked from the mobile app.

The rollout begins with Pro and Max subscribers on web, desktop, and mobile over the coming weeks; Team and Free plans follow later, with nothing for users to turn on. The move follows OpenAI’s July launch of ChatGPT Work, which similarly combined chatbot and coding tools — the latest step in the two IPO-bound companies’ race for enterprise and non-coder customers.

Why it matters: The pattern is clear: AI assistants are absorbing the apps they once pointed you to. Claude now writes the document, builds the deck, and runs the project without you switching tabs. For anyone using AI for real work — reports, slides, drafts — the assistant just became the workspace.

• Announced September 16, 2026: chat and Cowork merge into one Claude interface
• Claude Docs and Claude Slides launch in beta, with exports to Google Docs/Word, PowerPoint, and PDF
• Claude Design now works inside any conversation; finished work can be shared via link
• Rolling out to Pro and Max first (web, desktop, mobile) over coming weeks; Team and Free plans follow
• Follows OpenAI’s July launch of ChatGPT Work, which combined chatbot and coding tools

Sources: https://techcrunch.com/2026/09/16/anthropic-merges-claude-chat-and-cowork-in-one-interface/, https://www.reuters.com/business/media-telecom/anthropic-fold-claude-ai-features-into-one-interface-launches-document-tools-2026-09-16/

The lineup: Casey handles customer help, Paige covers IT and HR employee requests, Carter is a shopper agent for commerce, Marshall runs supply chain, Piper generates inbound pipeline, Fin manages customer experience, and Hunter does outbound sales (currently in pilot, targeting general availability in November 2026). Salesforce says the idea is to start with an agent designed for one job, then teach it to handle more complex work over time — connecting directly to the customer data businesses already run in Salesforce.

The numbers behind the launch are big: Salesforce says 7 billion Agentic Work Units have already been delivered across Agentforce and Slack, including 3.2 billion in Q2 alone. Early customer results include 90% of core shopper journeys handled by Hibbett’s AI within six weeks, and 79% of Anthropic’s own conversations resolved autonomously by Fin. New platform capabilities landing alongside: Multi-Agent Orchestration (available now), plus AI Skills in Agentforce Coworker and an Agent Optimizer (both October 2026).

Why it matters: This is the clearest signal yet that AI agents are moving from DIY projects to off-the-shelf hires. For small and mid-sized businesses without an AI engineering team, “download an agent that already knows the job” is a much shorter path to value than building one from scratch.

• Seven agents: Casey, Paige, Carter, Marshall, Piper, Fin (generally available); Hunter in pilot, GA November 2026
• Unveiled at Dreamforce 2026 in San Francisco
• Multi-Agent Orchestration generally available now; AI Skills in Agentforce Coworker and Agent Optimizer GA October 2026
• 7 billion Agentic Work Units delivered across Agentforce and Slack
• Early results: 90% of Hibbett’s core shopper journeys in six weeks; 79% of Anthropic conversations resolved by Fin

Sources: ITWeb — Salesforce deploys new AI agents amid rising AI adoption, Futurum Group — Agentic AI Agents Close Enterprise Gap at Salesforce

The platform leans on the same infrastructure behind Komodor’s existing AI SRE product. The pitch, in co-founder and CTO Itiel Shwartz’s words: “The hardest part of running agentic operations in production is not building the agents. It is keeping every agent grounded in the right context, giving it memory that persists across incidents, ensuring that accuracy improves rather than degrades, and securing every run — all while keeping model and compute costs under control.”

The timing tracks with where the market is: coding agents are helping teams ship software changes faster, which is increasing incidents, support requests, and maintenance load. Komodor cites a survey finding 60% of senior enterprise leaders are deploying agents in production — while Gartner projects more than 40% of agentic AI initiatives will be decommissioned by 2027 over governance gaps, unclear ROI, or escalating costs.

Why it matters: The agent conversation is shifting from “look what it can do” to “how do we run it safely at scale.” Platforms that solve governance, memory, and cost are the picks-and-shovels layer of the agentic era — and a sign the technology is growing up.

• Announced September 16, 2026 (Tel Aviv and San Francisco)
• Covers AI SRE, AI Software Operations, and Cost Optimization workflows
• Build or import agents under shared governance, context, and memory
• 60% of senior enterprise leaders deploying agents in production (cited survey)
• Gartner projects 40%+ of agentic AI initiatives decommissioned by 2027

Sources: GlobeNewswire — Komodor Launches Agentic Operations Platform, Daily Guardian — Komodor Launches Agentic Operations Platform

The disclosure follows a dramatic weekend: Anthropic CEO Dario Amodei published an essay calling on the industry to slow the pace of frontier AI development and warning of potentially catastrophic outcomes. OpenAI CEO Sam Altman, Google DeepMind’s Demis Hassabis, and xAI’s Elon Musk all publicly backed Amodei’s call. Altman said OpenAI would join Anthropic in embedding third-party evaluators inside the company to monitor safety.

The backdrop is tense. Days earlier, former OpenAI and Anthropic researcher Jacob Coxon resigned, accusing both labs of racing toward self-improving superintelligence without acting responsibly. And the cooperation itself raises legal eyebrows: coordinated slowdowns could look like output restriction under antitrust law. Lehane said OpenAI doesn’t believe the talks need an antitrust waiver — but noted he’s in Washington meeting lawmakers, and bipartisan legislation on catastrophic AI risk is on the table.

Why it matters: When the three fiercest competitors in AI start coordinating, the pace of model releases, safety testing norms, and possibly regulation all shift. For everyone building on these models — including small businesses — this is the conversation that decides what guardrails the next generation of AI ships with.

• OpenAI’s Chris Lehane confirmed weeks of safety talks with Anthropic and Google DeepMind (via Bloomberg, Sept 15)
• Follows Amodei’s weekend essay urging the industry to slow frontier AI development
• Altman, Hassabis, and Musk publicly supported the call; Altman pledged third-party safety evaluators at OpenAI
• Lehane is in Washington meeting lawmakers; says no antitrust waiver needed for the coordination
• Separately, CNN reported the labs have discussed forming an AI industry standards body

Sources: Reuters — OpenAI is working with Anthropic, Google on AI safety, TechCrunch — OpenAI, Anthropic, Google have been in talks on AI safety for weeks, Neowin — OpenAI confirms AI safety talks after extinction warnings, CNN — Top AI companies have discussed creating their own standards body

The pitch is simple: “Look at what we merged this week, draft a release note for customers, and leave it as a draft for me.” Connected once through OAuth — no API keys to manage — the agent can read what the team merged, write a customer-ready note in the team’s voice, file it under the right category, attach images, and schedule it. It can also pull up earlier posts to edit or reference, and report how each one performed. Every draft waits for a person to approve it; nothing publishes until someone presses the button.

Each published entry ships to three places at once: inside the product via a 4.3kb in-app widget, on a public changelog page hosted on the company’s own domain, and as an LLM-ready Markdown feed. That last one closes a neat loop: the AI that helped build the feature announces it, and the AI assistants customers ask about the product all read the same source of truth instead of guessing. “Teams already write release notes with AI. Now the AI can ship them too,” said Felix, CEO of ReleasePad.

Why it matters: MCP has quietly become the standard way to plug AI agents into real software. Tools like this show what that unlocks in practice: agents that don’t just chat, but do the unglamorous operational work — with a human still holding the publish button.

• Launched September 14, 2026 via GlobeNewswire
• MCP server works with Claude, Codex, Grok, and other LLM coding agents
• OAuth connection, no API keys; human approval required before anything publishes
• Distributes to in-app widget, public changelog page, and LLM-ready Markdown feed

Sources: Tech Startups — ReleasePad Launches MCP Server, Pluralist (GlobeNewswire) — ReleasePad Launches MCP Server