Googles new Agentspace enables business users to create productivity-focused AI agents.GoogleGoogle has just launched Agentspace, a no-code-ish environment geared towards personal work productivity. Many of Googles competitors are also playing in this market, so its no surprise that Google is getting into it. At this stage, each of the major cloud vendors now has an AI development platform (in Googles case Vertex AI) and multiple tools geared towards different types of developers. Googles foray into no-code for power users is different than its competitions, and quite interesting.Whats Cool About AgentspaceAgentspace is an extension of Googles viral NotebookLM offering. In fact, as part of the packaging, the enterprise-grade NotebookLM Plus is the base offering. NotebookLM has been in trial mode for a while now. I like it as a means for a user to create spaces around a topic or project and then leverage LLMs to search and create new content from that space. I know the create a podcast feature is what gets the clicks, but the whole concept is cool and useful.The higher-end Agentspace offerings add in new capabilities including third-party integrations (think SharePoint, Jira and Salesforce) and team space sharing. These are all good nods to how enterprise users can work together and share best practices via agents.Agentspace also has some prompt memory capabilities that enable agentic behaviors, such as prompt chaining. Its not really a structured workflow like we are seeing with Salesforces and ServiceNows agent development tooling. However, if you are a typical business user, it is reflective of the less structured processes that people create for themselves while performing daily tasks.What I Am Not Sure About With AgentspaceAbove all, Agentspace is a prompt-driven experience. I know that this approach works for searchand is therefore a natural fit for Googlebut I am not sure that a prompt-only agent will be accepted by most users. My issue is that sometimes I need to see something to act upon it. Is just a context window or a notification enough? Also, your mileage will vary based on how well you are writing promptsand there is a vast range of skill in that area in the average workplace. Ask anyone who does not understand good prompting how their experience with AI is going. The response will probably not be very good.MORE FOR YOUAs with other platform-centric agents, I remain concerned about Agentspaces cross-platform integration and how it will scale. For low-scale scenarios (for instance if I do only a few prompts a day), its likely fine. But when a whole organization is hitting agents that then need to reach back to on-prem IT resources or to another cloud, how will it perform? And what will it cost? I am optimistic that over time we will start to see more of an integration between no-code tools and AI development frameworks such as Vertex AI or AWS Bedrock. In fact, we are starting to see this with Salesforces Agentforce, which leverages Salesforces own Mulesoft technology to provide high-scale integration connectors. I am hopeful that this becomes a trend rather than remaining a one-off.Like its competitors, Google is putting a lot of different tools out there, and some of them overlap. For example, how does one compare Agentspace to Vertex AI Agent Builder? I appreciate that in the brave new world of AI, experimenting and failing fast is a good way to innovate. But I think that all of these AI vendors should be more declarative of their intent for each of these tools. Who was Agentspace designed for, and how will it improve those users lives or their work? I have this issue because I believe AI should elevate the capabilities of a business user to the point where we should see some sort of blurring or blending between no-code and low-code tools. But that may not be realistic yet, considering that Google has also announced a preview of a new tool called Jules, which seems kind of like AWS App Studio.Looking For Disruptive Approaches In The No-Code SpaceGoogle has made some pretty bold user experience decisions when it comes to no-code, which I respect. For example, I like the NotebookLM concept of creating a more freeform workspace rather than the more form-centric approach of other no-code tools. And extending the functionality with collaboration and third-party integrations provides a real opportunity to test whether that model can be a disruptor against SaaS no-code plays. Hopefully, Google has picked the right time to push against the convention with a solution that is pretty innovative.