On May 30, TechCrunch published a hands-on review of Gemini Spark, the 24/7 agentic assistant Google introduced at I/O 2026 in mid-May. Its job is to help users “navigate your digital life”: the ongoing chores of handling online to-dos, summarizing unread content, and organizing things like expense spreadsheets. It runs on cloud virtual machines, so long-running tasks do not need your device present — Sundar Pichai joked at launch that yes, you can close your laptop.
The verdict is cautiously positive: it is “actually pretty useful.” But the review is equally clear about where the product’s boundaries and identity problems sit.
What Gemini Spark Actually Is
Gemini Spark launched on May 19, 2026 at Google I/O as a 24/7 agentic assistant with deep integrations into Gmail, Calendar, Docs, Sheets, and Slides. The reviewer tested it through early access; the article offers no pricing or tier information.
The architectural difference is where it lives. Spark runs on cloud VMs and keeps executing server-side, in contrast with agent frameworks like OpenClaw that need the local machine kept awake. Close the laptop and the task keeps running. It is a classic Google move: trading infrastructure for experience.
Six Tasks, Mixed Results
The reviewer threw six kinds of work at Spark, with uneven results:
- Drugstore shopping research: it found Walgreens sale items, coupons to clip, and stacking strategies; one promo code turned out invalid, but it surfaced buy-one-get-one deals and rewards
- Day-trip packing list: it checked the weather and event details on its own, suggested chairs, sunscreen, and water, and flagged that dogs were not allowed; but the list could not be exported to Keep — only Docs or email
- Teen summer activities: it produced a solid local list with driving distances, but omitted costs and dates unprompted
- Friday newsletter digest: it summarized top articles with links, though links required an extra click through a Google redirect, and it returned 4 items when asked for 5
- Friday weekend-events roundup: combining web search with Gmail-scanned local newsletters, it surfaced obscure finds like a local “Beaver Queen Pageant”; adding events to the calendar required a reply plus a confirmation click
- Price-drop tracking: it interpreted the request as a check every two weeks, which the reviewer doubts is frequent enough
The pattern: recurring email and newsletter summaries, local event discovery, and contextual list-building are dependable strengths. Cross-product export, link handling, and prompt interpretation are where it stumbles.
The small details matter here. The packing list went beyond the obvious to suggest blankets, a light layer, a shopping bag, and an umbrella, and it pulled the no-dogs rule from the event page unprompted. That is the kind of contextual inference that separates an agent from a search box with extra steps. But the same system handed back four newsletter items when asked for five, which is the kind of miss that erodes trust in a tool meant to run unattended.
Runs While Your Laptop Is Closed: The Cloud Trade-Off
Being cloud-resident is the selling point and the constraint. Tasks can run around the clock, but outside Google’s ecosystem Spark is currently helpless — it cannot book a restaurant through Resy or hunt flight deals. The review notes MCP integrations are planned, and that will be the decisive step from “Google’s in-house butler” to a general-purpose assistant.
The Keep gap is telling. An assistant that generates lists but cannot save them to Google’s own note-taking app suggests Spark’s integrations were prioritized around the Workspace productivity core — mail, calendar, documents, spreadsheets — rather than the full consumer stack.
On the iPhone, Spark is a separate toggle inside the Gemini app and cannot be targeted by the hardware Activity Button. The reviewer also wishes she could simply text it rather than opening the app each time.
One Brand Too Many?
The sharpest criticism in the review targets positioning: Spark should not be a standalone brand at all. With AI product names already proliferating, adding another one increases cognitive load for no benefit. The reviewer’s argument is that these capabilities belong inside Gemini itself — users should not have to mentally separate “questions go to Gemini” from “tasks go to Spark.”
For product teams, the review’s real value is that it prices out the actual exchange rate of an agentic assistant: task completion tracks integration depth and detail polish, not raw model capability. Spark’s next milestones are obvious — MCP support and more export paths. Until then it remains a useful butler fenced inside Google’s garden.
Sources
- I put Google’s 24/7 AI assistant Gemini Spark to work, and it’s actually pretty useful — TechCrunch
- Google introduces Gemini Spark, a 24/7 agentic assistant with Gmail integration — TechCrunch
- Gemini Spark overview — Google
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.