Daily Pulse · Wednesday, September 16, 2026

Hi, this is Koko from Koko Knows, and there's a lot to get into today, so let's dive right in.

The big story is out of Dreamforce, where Salesforce just struck a much deeper alliance with Google Cloud, and it's a real signal about where enterprise software is headed. Starting in the fourth quarter of this year, Salesforce is going to migrate select US customers onto Google Cloud through its Hyperforce infrastructure, and it's wiring Agentforce directly into Gemini Enterprise using the MCP standard. That's not a small integration tweak, that's Salesforce folding its agent platform straight into Google's data layer. And Salesforce didn't stop there. They also unveiled a CRM-specific reasoning model called Koa, built on Nvidia's Nemotron architecture and trained on synthetic data drawn from twenty seven years of CRM deployments, aimed at sales and service tasks across more than fourteen industries. On top of that, there's a new interface layer called AIforce. Put it all together and you can see the pattern: the biggest SaaS vendors are no longer treating AI integration as something you bolt on later. They're building it straight into the hyperscaler infrastructure from day one.

Now, while all this agent infrastructure is getting stitched together, there's a real warning sign flashing on the security side. A new report from IDC and GuidePoint finds that non-human identities, meaning bots, agents, and service accounts, now outnumber human identities by as much as seventy five to one in some environments. And these agent identities were the initial entry point in nineteen percent of security incidents. Security teams are simply not built to track that kind of sprawl yet, and that gap is widening right as companies are racing to deploy more agents, not fewer.

That security theme connects to a broader trend I'm watching: non-human identity governance is becoming one of the central IT headaches of this AI moment. A separate report from AvePoint backs this up, showing the same pattern of machine identities multiplying faster than anyone's ability to govern them. Meanwhile, there's a second trend worth flagging, which is that data quality, not data volume, is becoming the real bottleneck for AI returns. Channel partners scaling AI ops and security tools are finding this out the hard way, and it echoes warnings from Gartner and MIT Sloan about what they're calling a capability mirage, where AI looks impressive in a demo but falls apart on messy real-world data.

On the product front, a few things shipped worth knowing about. Snowflake rolled out something called Intent-Driven Governance, aimed at automating how data access policies get applied. Celonis teamed up with Microsoft Copilot Studio to launch an HR Process Agent for automating HR workflows. And Snowflake also introduced CoCo, its own AI coding agent. All three are live now, all three fit into that bigger theme of infrastructure providers racing to own the agent layer.

Zooming out for a second, there's an interesting tension playing out around AI's pace. Anthropic's Dario Amodei has been pushing for a more coordinated, careful pace of AI development, and that's drawn some quiet support from rival labs, apparently including some discussions between Anthropic, OpenAI, and Google about a joint safety standards body. But politically, it's a different story. Trump and House Speaker Johnson have both pushed back hard against any regulatory slowdown. So you've got labs privately entertaining more caution while Washington's political leadership is leaning the opposite way.

A couple of things to keep an eye on this week. In Congress, watch whether Senators Hawley and Blumenthal can actually get a floor vote on their federal AI monitoring bill before recess. There's rare bipartisan energy behind it after some agent-swarm hacking incidents, so this could actually move. And in Europe, watch Broadcom. The ECCO group just handed down a "critical" rating on how Broadcom has been handling VMware licensing, and that could be the trigger for the EU Commission to open a formal antitrust case.

That's the rundown for today. Big alliances, bigger security gaps, and a policy tug-of-war that isn't going away anytime soon. Thanks for listening, this has been Koko from Koko Knows, and I'll catch you next time with more of what you need to know.