The Not So Secret Agents: What Every Decision Maker Should Know About Agentic AI
228 announcements over three days in Google Cloud’s Next ‘25 event in Vegas. Glancing through the list, it is hard to absorb the scale of what Google does with cloud and AI. This year, though, one phrase took centre stage: Agentic AI.
A License to Build
One of the conceits of Ian Fleming’s novels was that James Bond would always announce himself, bold as brass, informing the villain exactly who he was and what he was up to. Last week, Google’s agents did the same thing on stage in Las Vegas. In front of a huge audience, we got to learn who they are and their plans for the future.
Today, we will look at those agents and review their plans in a little more detail. Hopefully, we will help you understand why how you get things done is about to change forever. But first of all, what exactly is an AI agent?
According to Google, a good agent displays the following qualities;
- Reasoning: Understands context, analyzes data, makes smart decisions, and offers relevant advice (e.g., when sales are up, the agent reminds you to thank the new sales exec who brought new clients with him).
- Acting: Takes action and gets things done in the real world (e.g., it books the regular meetings with all those new clients).
- Observing: Gathers information through various senses and data (e.g., sees all the new customer data, deeply researches details, and provides insights).
- Planning: Strategizes based on the situation, anticipates problems, and suggests solutions (e.g., when your customer support gets stretched with the new customers, the agent prioritizes tasks and recommends training).
- Collaborating: Works with people and other agents to achieve goals, communicating and coordinating effectively (e.g., multiple agents track new sales’ impact across the entire business and share updates).
- Self-Refining: Learns from experience, adapts, and continuously improves. (e.g., next time a new sales exec arrives with new customers, prepare in advance the support she needs to shine!).
Agents are autonomous; they have a clearly defined role and persona, remember, and autonomously employ tools to work with. At the core of the agents, Large Language Models (LLMs) allow them to process and generate language, whilst other components provide reasoning and action.
Just like Bond against the villain in Casino Royale, Google went ‘all-in’ on Agentic AI with a slew of announcements last week. Here’s what you can expect in the coming days and weeks, and how you can benefit from the latest advances in the field.
“Take a Letter, Miss Moneypenny”
Now that we know what an agent is, let me introduce you to Google Agentspace. Agentspace is a platform for building and deploying AI agents that understand, reason, and act in the real world. Agentspace provides tools and infrastructure to create those intelligent assistants that can automate tasks, provide information, and collaborate with users in a dynamic and contextual way. Think of it as Google’s hub for unleashing the power of autonomous AI.
Agentspace agents act on your behalf, but they remain firmly under your control, and it is a truly liberating experience. What do I mean? Well, each week, I send out meeting requests for daily scrum gatherings, and organise weekly client updates and governance meetings. Before each meeting, I need to create an agenda; after the meeting, I should write up the minutes. All of that is a chore, and it takes time away from me being able to add value elsewhere. Agentspace allows me to offload the work to my digital double and get on with doing things elsewhere that create a greater impact for clients and customers.
The power of Agentspace lies not just in its ability to ‘do’, but also very much in its ability to ‘know’. Like every organisation on the planet, yours also keeps its knowledge in many disparate and almost certainly unconnected systems. You can hook Agentspace into all of them via ready-made connectors. And if a ready-made connector does not exist for your particular legacy system flavour, you (or a friendly code-writer) can build one. The agents in Agentspace can provide you with an in-depth overview of your organisation, one that you can interact with, one that you can ask questions of — and most importantly, one that is a unified, amalgamation of all your data, regardless of silos.
As well as apps, Google also offers Notebook LM as part of Agentspace. Google Notebook LM is an AI-powered research assistant that helps you understand and synthesize information from your documents. It acts like a smart, personalized notebook that can answer questions, summarize key points, extract insights, and even suggest related topics, making research faster and more efficient. Not only does Notebook LM offer you written summaries, but it can also come up with audio content for you, creating a bespoke ‘podcast’ dedicated to the topic you are interested in. Notebook LM is your AI research sidekick for making sense of your notes and documents.
Google has positioned Agentspace and its other agent tools as competitors to Microsoft and its Copilot tooling – and has already established a lead over the competition. Out of the gate, Agentspace has a degree of autonomy that is missing elsewhere. Agentspace doesn’t just react to questions, it pre-empts them. It doesn’t just do things when you ask, it does things when you need them done. The competitors are playing catch-up right now.
With an Agentspace application or Notebook LM working on your behalf, you are better informed, you are in control and your time is freed up for all the other important things you have to do – all with the peace of mind that comes from knowing that Agentspace is backed by enterprise-grade security, privacy, and compliance controls.
“Now, pay attention, 007”
You might expect that getting things up and running could be tricky, but that isn’t necessarily the case. Right now, I just need to do three things to create a straightforward agent that can tell me about my customers and their history with my company.
#1 Create a data source – in this case, my company’s ‘sales’ folder in Google Drive – or any other storage container or application, it doesn’t have to be Google!
#2 Create an application that utilises both search and AI, linked to the data source I created in step one.
#3 Publish the URL for the app to those who want to use it.
Simple. Once I have done that, anyone in my company can interrogate the sales material, asking questions like “Can you summarise all the proposals to Acme?” and “Who was the guy from Universal Exports who ordered all the cigars?”.
Within the coming weeks, it will get even simpler. By issuing a spoken prompt like this:
“Create an Agentspace agent that connects to the sales folder in Drive and can tell me about all the material there.”
Yes, that’s right, you can use AI to create AI agents. With that sort of power, the sky’s the limit – and you’ll probably find yourself thinking, “The World is Not Enough”.
—
Next time, we will look at some more agent news from Google, this time taking your customer engagement to the next level.
If you want to know more about Agentspace, contact me, Matthew Wooller at Codento, and let’s talk about how a not-so-secret agent of your own can make your life easier and more productive.
Matthew Wooller, Data and AI Strategist, Codento
Contact us for more information on our services:
ctively participating to the event in Las Vegas with Ulf Sandlund and Markku Pulkkinen and remotely via the entire Codento team. Earlier on Tuesday Codento was awarded as 
ng embedded across a broad range of Google Cloud services addressing a variety of use cases and becoming a true differentiator, for example:
What is already known convinces us that Google Cloud and its AI approach continues to be completely enterprise-ready providing capabilities to support deployments from pilot to production. 
