Last month we promised a closer look at the announcement that mattered most for business leaders at the annual Google I/O Conference. On May 19, 2026, Google delivered it: Gemini Spark, a personal AI agent that runs 24/7 on Google Cloud virtual machines and keeps working even when your phone and laptop are off.[1] Spark is Google’s entry into a category that did not exist six months ago, the “always-on agent.”
What an Always-On Agent Actually Is
Spark is built on Gemini 3.5 and Google’s Antigravity agentic harness (similar to Claude Cowork and OpenAI Codex), and it lives in the cloud rather than on your device.[2] You can email it at a dedicated Gmail address, and it pulls context from Gmail, Calendar, Docs, and Google’s “Personal Intelligence” layer without manual setup.[1] Through the MCP (Model Context Protocol)—the open standard introduced by Anthropic that has been widely adopted across the industry—it reaches beyond Google into 30-plus third-party services such as Salesforce, Adobe, Canva, Zendesk, GitHub, and WhatsApp. Beta access opened the week of May 19, 2026, to U.S. Google AI Ultra subscribers, on a tier Google cut from $250 to $100 per month. This new price is in line with Google, Anthropic, and OpenAI, all of which now anchor a premium tier near $100. “What makes Spark stand out then?” you ask. It’s the only one bundling a persistent cloud agent, with the contextual advantages of the Google workspace ecosystem integration, at that monthly subscription price.[3]
Google I/O: more than Just Spark
The deeper signal from Google I/O was the shift of agents from standalone capabilities to core infrastructure. Google announce that it had made AI agents a standard inclusion in every enterprise Workspace contract,[4] and the company positioned Gemini 3.5 Flash as their strongest agentic model yet, running roughly four times faster than rival frontier models.[5] One observer summarized the conference bluntly: ‘Google is no longer selling AI tools, it is selling AI employees’. For leaders, this means autonomous workflows are about to arrive in your organization, whether or not you procure them deliberately.
What Early Users Are Telling Us
My review of reactions on X and Reddit over the past month, while limited by the narrow beta release, is revealing. Enthusiasts describe Spark’s persistent background execution using words like “magical” for the way in which it strings together multi-step business workflows with minimal supervision: triaging an overflowing inbox, drafting client follow-ups, reconciling expense spreadsheets, or assembling a meeting brief from your Calendar and email threads. Google Workspace enterprise users report the largest gains, since the deep integration tailors output to your actual data rather than generic boilerplate.
The same access that delivers those gains can also unsettle. Some reviewers call the data reach “terrifying,” flag prompt-injection risk, and question why Spark exists separately from ordinary Gemini chat at a $100 entry point. The sharpest concern is autonomy itself: a leaked onboarding screen warned that Spark “may do things like share your info or make purchases without asking” and urged users to supervise it.[6] That single sentence captures the whole category’s promise and peril.
This Month’s Action Item for Leaders
Always-on agents introduce new requirements for onboarding, management, and governance. I’ve suggested below three moves before you scale; if you’ve read my memos the past few months, the steps won’t surprise you. First, pilot in a contained, low-stakes domain. Start with a single recurring workflow such as meeting prep or expense summarization, with a mandatory human checkpoint before any irreversible action. Second, write an agent permissions policy before deployment. Define for your human and agentic teams exactly which data and systems it can access and where the hard stops are. Third, classify your workflows by autonomy tier, specifying supervised, semi-autonomous, or fully autonomous. Do not let anything graduate to a higher tier without an audit trail. The organizations that win in 2027 will be those that learned to delegate to agents effectively and responsibly.
One Other Thing
Watch the “agent org chart” pattern beginning to form. As leaders run a blend of human reports and persistent agents, the management questions converge: who owns the outcome, who reviews the work, and who is accountable when an autonomous worker errs? Regulators are already circling. In the EU and U.K., the availability of Spark is pending AI Act compliance review, with analysts expecting a resolution in Q3 2026.[3]
Next Month at a Glance
Normally, this is where we tell you what we’ll cover next month. In July the pace of AI announcements makes that difficult. I’ll see where the industry lands over the next few weeks and pick up the most important development in our next issue.
Mike Grandinetti is an Executive Fellow at Harvard Business School. He’s a serial tech entrepreneur, board member, AI & innovation consultant, VC EIR, and award-winning professor in the practice. A former Silicon Valley engineer and McKinsey consultant, Mike has been C-Suite leader roles across 8 tech startups, resulting in 2 NASDAQ IPOs and 7 strategic exits.
He’s led senior executive workshops for Berkeley, Brown, Carnegie Mellon, Columbia, Cornell, Harvard P&ED, NYU & Oxford. He’s been a senior advisor and organizing team member for the MIT CIO Symposium for a decade.
https://www.linkedin.com/in/mikegrandinetti
www.mikegrandinetti.com
Sources
[1] TechCrunch, “Google introduces Gemini Spark, a 24/7 agentic assistant with Gmail integration, at I/O 2026,” May 19, 2026. https://techcrunch.com/2026/05/19/google-introduces-gemini-spark-a-24-7-agentic-assistant-with-gmail-integration/
[2] DataCamp, “Gemini Spark: Google’s Always-On AI Agent Explained,” 2026. https://www.datacamp.com/blog/gemini-spark
[3] TechTimes, “Google Cuts AI Ultra to $100, Launches Gemini Spark Agent and Android XR Glasses at I/O 2026,” May 19, 2026. https://www.techtimes.com/articles/316853/20260519/google-cuts-ai-ultra-100-launches-gemini-spark-agent-android-xr-glasses-i-o-2026.htm
[4] Enterprise DNA, “Google I/O 2026: Enterprise AI Agents Are Now Standard,” 2026. https://enterprisedna.co/resources/news/google-io-2026-gemini-3-5-enterprise-agents/
[5] Google, “I/O 2026 developer highlights: Antigravity, Gemini API, AI Studio,” May 2026. https://blog.google/innovation-and-ai/technology/developers-tools/google-io-2026-developer-highlights/
[6] TechTimes, “Gemini Spark: Google’s 24/7 Cloud AI Agent Now Executes Tasks in Third-Party Apps,” May 25, 2026. https://www.techtimes.com/articles/317144/20260525/gemini-spark-googles-24-7-cloud-ai-agent-now-executes-tasks-third-party-apps.htm