Salesforce expands cross‑cloud agents

Enterprise AI is starting to look less like one all-knowing copilot and more like a team of specialists. That is the real story behind Salesforce’s latest push with Google Cloud and the parallel expansion at Unisys. The problem has been obvious for a while — most big companies keep customer data, documents, chat, scheduling, and service workflows in different systems. AI sounds great until the agent hits that wall. Salesforce’s answer, announced April 22 with Google Cloud, is to let agents move across both stacks instead of staying trapped in one app. (salesforce.com) ### What actually changed between Salesforce and Google? Salesforce and Google Cloud widened their partnership so agents can pull context and take actions across Salesforce, Slack, Google Workspace, and Google Search. Salesforce also tied Agentforce more closely to Google’s Gemini models, while Google’s side gets d(salesforce.com)t a human stitching the tools together. (salesforce.com) ### Why is “cross-cloud” the important phrase? Because most enterprise work breaks at the handoff. A service rep may have the customer history in Salesforce, the technician schedule somewhere else, and the supporting documents in Google tools. A single assistant inside one product can summarize what it sees, but it(salesforce.com)s in another, then hand the result back into the workflow people already use. (salesforce.com) ### Where does Unisys fit into this? Unisys is the proof point Salesforce wants. On April 22, Unisys said it was expanding Agentforce 360 inside its Digital Workplace Solutions field-services operation. That deployment now runs across more than 120 countries, manages 7.4 million devices, routes more than 1 million (salesforce.com)5%, which is the kind of operational metric buyers actually care about. (unisys.com) ### Why do those numbers matter? Because AI demos are cheap. Service operations are not. If a company can really route over 1 million tickets a year and optimize 80,000 monthly appointments, that means the software is touching dispatch, technician availability, customer records, and on-site execution — not just writing email drafts. Basically, this is where “agentic AI” stops being a slide and starts acting like infrastructure. (investing.com) ### Is Salesforce trying to replace core systems? Not really. Turns out the strategy is almost the opposite. Salesforce is positioning Agentforce as a coordination layer that sits on top of systems of record rather than a replacement for them. The company keeps talking about connecting apps, data, and agents on (investing.com)agents move between them. (unisys.com) ### Why does that matter for the market? Because it is a more believable enterprise story. Big companies rarely rip out core platforms just to chase a new AI model. But they will pay for software that makes the existing stack work faster, with fewer manual handoffs. That is why these announcements matter beyond Salesforc(unisys.com)ed software, not the ones promising one universal bot. (salesforce.com) ### What is the catch? Integration is still the hard part. Cross-cloud agents only work if permissions, data quality, workflow design, and human oversight are solid. A bad process automated across two platforms is still a bad process — just faster and harder to unwind. But Salesforce’s latest moves show where the industry is heading: less chatbot theater, more workflow plumbing. (salesforce.com) ### Bottom line? Salesforce is making a practical bet. Enterprise AI will be won by agents that can navigate the messy reality of existing software estates. The Google partnership gives that bet broader reach, and the Unisys rollout gives it numbers that sound a lot more like operations than hype. (unisys.com)