Google Gemini improvements: users demand reliable Workspace integrations and better tool calling

Google Gemini users have made their priorities clear. After a Google executive asked what the AI assistant still gets wrong, over 1,700 replies identified the biggest pain points: unreliable Google Workspace integrations, inconsistent tool calling, and missing research workflow features. For AI builders building agents and automation on top of Gemini, these gaps directly affect how well your cross-app workflows perform.

What happened

Josh Woodward, a VP at Google covering Google Labs, the Gemini App, and Google AI Studio, asked users what Gemini still could not do well and should have been fixed long ago. After collecting more than 1,700 replies, he shared the top 10 requests along with Google's stance on each. The number one complaint was that Google Workspace integrations need to work more reliably, especially across Gmail, Docs, Slides, Keep, and Tasks. The second was more reliable tool calling, which Google said it strongly agrees with and expects to deliver noticeable gains soon.

Users also asked for better project and folder organization for chats, the ability to export Deep Research results to NotebookLM, and the option to switch between Deep Research mode and Flash/Pro models within the same chat. Woodward said he was not aware of those pain points, called them good ideas, and said Google would figure something out. Other requests included expanding MCPs and Custom Skills, message editing, improved voice dictation accuracy, and fixes for mobile app scrolling bugs. The one item that will not change is celebrity likeness guardrails, which Google plans to keep as they are.

Why AI builders should care

If you build agents or automation that use Gemini to interact with Gmail, Docs, or Google Drive, the reliability of Workspace integrations is not a nice-to-have -- it is the foundation of your workflow. Unreliable tool calling means your multi-step agentic tasks may fail silently or produce inconsistent results. The planned improvements to Gemini Spark and tool calling directly affect how you design and test agent pipelines. Additionally, the ability to export Deep Research results to NotebookLM and switch between research and chat modes could unlock more flexible research-to-product pipelines, but those features are still in the idea stage.

Practical implications

Google has acknowledged the top complaints and is actively working on several fronts. Tool calling reliability should see noticeable gains soon, though no specific date was given. Workspace integration reliability is being improved, with some recent upgrades already in Gemini Spark. Voice dictation accuracy and mobile UI scrolling bugs are also being prioritized. For builders, this means you can expect incremental improvements in the near term, but features like Deep Research export and chat organization are further out with no timeline.

Caveats

All of the above is based on a single source summarizing user feedback and Google's stated intentions. Timelines are vague -- "near-term gains" and "figure something out" are not commitments. Some requested features, like celebrity likeness guardrails, are explicitly not changing. The improvements mentioned may also roll out gradually or change in scope. If you are building production workflows on Gemini, test against the current reliability levels and plan for potential instability until these fixes land.

FAQs

What improvements is Google planning for Gemini Spark and its tool-calling reliability?

Google acknowledged that tool calling reliability needs improvement and said users should expect noticeable gains soon, though no exact timeline was provided. The company also noted that some improvements recently went into Gemini Spark.

How can I export Deep Research results to NotebookLM?

Users requested the ability to export Deep Research reports to NotebookLM. Google's VP called it a good idea and said the company would figure something out, but no concrete timeline or implementation details have been shared.

Will Gemini support better cross-app workflows across Gmail, Docs, Sheets, Keep and Tasks?

Yes, that was the top complaint. Google agrees that Workspace integrations across Gmail, Docs, Slides, Keep, and Tasks need to be much more reliable, especially for the Gemini Spark agent. Work is ongoing, but no specific timeline was given.

How can I switch between Deep Research mode and Flash/Pro models in the same chat?

Users asked for the ability to switch between Deep Research mode and Flash/Pro models within the same chat. Google described the idea as a good suggestion and said it would figure something out. No rollout date is available.