AI Insights of the Week: September 28th

AI Insights of the Week: September 28th - October 5th

All Articles

Resources

Oct 5, 2026

The Authentic .AI Staff

OUR THOUGHTS

On September 29, at its annual DevDay in San Francisco, OpenAI introduced dots, always-on agents that each get their own cloud computer, connect to more than 4,000 apps, and keep working after you close the laptop. It also opened a Marketplace that lets enterprise customers spend part of their existing OpenAI commitment on partner software, with Salesforce, HubSpot, and ServiceNow among the first 32 names. The next day, Google launched Gemini 4 Argon, a model it says matches or beats OpenAI's and Anthropic's flagships on key coding and cyber benchmarks, at an introductory $2 per million input tokens. That same day in Las Vegas, Gong announced Gong Activate, a system where agents and reps work "from the same blueprint," built on the institutional knowledge Gong already holds for its customers: buyer personas, competitive positioning, and the plays their best reps use to close. For us, that sequence is the real story. In 48 hours, the always-on agent became a standard feature and a third lab made a credible claim to the frontier, which makes the model the most interchangeable part of the stack. What isn't interchangeable is the context an agent works from, and in our view the platforms made it clear this week that they intend to hold it. The open question for B2B sales and marketing leaders now isn't which model your agents run on. It's where your playbook lives, who owns it, and whether it comes with you when you switch.

This week we've curated three pieces that trace this sequence in the order it broke. Decrypt's rundown of DevDay is the clearest account of what OpenAI shipped on September 29, and the detail worth noticing isn't the agents themselves but the Marketplace, where CRM and marketing automation spend can now flow through an AI lab's contract. For anyone who signs software renewals, that changes who is in the room. Constellation Research's report on Gemini 4 Argon, from September 30, shows why that matters: with Google claiming parity or better at a lower price, the model underneath any agent just became something you can swap. Argon is still limited to cybersecurity partners, so the signal is the pricing pressure and the pace, not an immediate switch. Gong's Celebrate '26 announcement, the same day, shows where the value moves once that happens, toward whoever holds the record of how your team actually wins deals. The customer results Gong shared on stage are vendor-reported, and we'd read them that way, but the strategic claim stands without them. Together, these three tell one story. Agents are becoming a standard feature, models are becoming interchangeable, and the durable asset is the institutional knowledge underneath both. The diligence question for B2B sales and marketing leaders has shifted. It's no longer whether a vendor's AI is good enough. It's whether the knowledge you feed it stays yours, in a form you can move, when the next model or platform comes along. We'll be watching which vendors are willing to say that out loud, and we suspect the answer will sort the field quickly.

ARTICLES OF THE WEEK

OpenAI Gave AI Agents Their Own Computers at DevDay 2026. Here's Everything It Announced – Decrypt

Published September 29, 2026, this piece rounds up the more than 20 launches OpenAI announced at DevDay in San Francisco, headlined by dots, always-on agents that each run on their own cloud computer.

Key takeaways:

Each dot runs on GPT-6 Astra, gets its own cloud computer and browser, connects to more than 4,000 apps, and can be reached in ChatGPT, by voice, or inside Slack and Teams. Built-in rules decide when a dot may act on its own, and sensitive jobs like changing a password stay with the human.
OpenAI paired dots with ChatGPT Space, a shared workspace where people and agents edit the same documents, and GPT-6.1 Sol, which OpenAI says nearly matches Astra on agentic and professional work at one-fifth of Astra's token prices.
For revenue leaders, the sleeper announcement is the OpenAI Marketplace: enterprise customers can apply prepaid OpenAI commitments toward partner software including Salesforce, HubSpot, and ServiceNow, which puts an AI lab in the middle of how the revenue stack itself gets bought.

Google launches Gemini 4 Argon – Constellation Research

Published September 30, 2026, this report covers Google's launch of Gemini 4 Argon, the long-awaited top model of its Gemini 4 generation, released first to cybersecurity partners under its Fairwinds program.

Key takeaways:

Google says Argon is on par with or ahead of Anthropic's and OpenAI's premier models, and CEO Sundar Pichai described it as showing "frontier performance in complex workflows, cyber defense and software engineering."
Argon carries an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens 95 percent off, though Google has not given a timeline for broad public release.
For B2B sales and marketing leaders, three credible frontier labs competing on price means the model under your sales and marketing tools is increasingly a swappable component. The decision that lasts is about the data and playbooks those tools run on, not the model brand on the box.

From Data to Action: Gong Debuts New Revenue AI Innovation at Celebrate '26 – Gong

Published September 30, 2026, this announcement from Gong's annual customer conference introduces Gong Activate, alongside its Mission Callisto releases: Gong Enrich, Agent Builder, and Deep Mode in Gong Assistant.

Key takeaways:

Gong Activate is built around what the company calls a "self-improving revenue loop." The Gong Revenue Graph captures buyer personas, competitive positioning, and top reps' deal strategies, and always-on agents evaluate every account and opportunity to prescribe the next step for reps and agents alike.
Agent Builder lets teams define triggers so custom agents run automatically when conditions are met, and Gong Enrich pulls third-party contact and account data from partners including ZoomInfo, Apollo, and Lusha into the same graph.
Gong shared customer results on stage, including Cisco reporting roughly 32 percent larger deal sizes and 26 percent higher win rates among early-adopting teams. Those numbers are vendor-reported, but the strategic signal for sales and enablement leaders is clear: the platform that stores how your team wins is positioning itself as the thing your agents can't work without.