Deploying an AI Agent Has Never Been Easier!

Arham: We’re beyond the point where we need to justify the value AI agents bring to businesses. The value and ROI that agents are bringing is quite apparent in the modern company’s workflow. However, there are a couple of tools that broadly achieve the same goal, but thrive in very different environments. I’m talking about Google’s Agentspace and AI Agent Builder.

To get more clarity on this, today I’ve brought on Yann Kronberg to dissect both of these tools and tell us what each of them is actually best for. So, Yann, my first question would be: what’s the difference between these two tools? They’re both agents, but how do you distinguish them?

Yann: To make it simple, Agentspace is the equivalent of enterprise search. You’re able to connect various data sources from many different environments. For instance, you can have a Jira instance or all kinds of Microsoft products you’re using, and you connect them all in one place: Agentspace. And you’re able to not only search all these things, but you can also add logic, create workflows, and perform actions. For example, you can search for information in Jira and then create a Slack message with that information directly from Agentspace. So it’s a very, very neat tool. It’s, if you want, a modern enterprise search.

Agent Builder is a bit different. It’s Google’s way of making it easy for people to build agents. First off, you have the Model Garden, and they also have an Agent Garden or Agent Gallery. You can choose a pre-built agent you want, such as one for customer service or quality assurance, and it’s mostly ready. You just need to connect it to your different sources. Alternatively, you can create one from scratch using complex coding and the SDK. You can build any sophisticated agent you can think of—like a customer service agent or a sales agent—using Agent Builder.

Where they intersect is that they can share data sources. So, you can have Agentspace connected to sources, and Agent Builder can leverage those same kinds of data sources.

Arham: There’s definitely a lot of work being done on constantly improving these tools. From your perspective, what do you see upcoming for both of these tools? Google has some announcements coming up soon, but do you have any thoughts on where these agents can still add a lot more value to their users?

Yann: It’s like the old AI stack; the agent is a new tool that increases productivity for most people. We’ve seen agents doing more and more things autonomously. In coding, it’s reaching the level of junior engineers, where you can execute on a roadmap with existing code. It still has some way to go, but they’re becoming more and more sophisticated in what they can do on the agent side.

For Agent Builder, Google’s direction is that you can connect to anything anywhere, including your own data, and you have one interface to get to this information. So these are definitely products that are going to increase people’s productivity, and it’s going to get better and better, for sure.

Arham: Now, Yann, I know you’ve worked with both these tools. Imagine I’m a company coming to you, and I want to get an idea of how much weight I really need to lift to get each of these working. Could you talk a bit about the management and technical setup that’s required for each one of these tools, and how can anyone looking at these tools get support from the ecosystem?

Yann: Agentspace is really plug-and-play. You can literally just turn it on because Google has built tons of out-of-the-box connectors. I think there’s more emphasis on what you can do with the information. So they basically let you create a little bit of workflow capability and creation capability. That’s really easy to set up; there’s almost no training.

Agent Builder is a bit more sophisticated. It’s not necessarily the building of the agent itself, because Google makes it easy by providing a lot of agents in their tools out of the box. It’s more about how you insert it so it’s useful within your company. Everyone has the customer service kind of agent in their mind; this is something that’s really well done. But beyond the customer service agent, you can do a lot of things: help you code, help you be more efficient on the automation side. We’ve seen quite a few companies improving their HR operations, their payroll operations, their bookkeeping operations using agents, and I think it’s just the beginning.

So you still need a human everywhere, but you need to make sure the agents are well-inserted into your company processes. That takes a little bit more work on the change management side, essentially. But as far as the easiness of building an agent, it’s never been easier. And as far as the implementation of Agent Space, it’s literally just turn it on, so it can’t be easier than that.

As far as building them, you don’t need any code. In the Agent Garden or Agent Gallery that Google provides, you can just use one and connect it to your data source. So there’s not a lot of support you need. If you want to do something more sophisticated, you’ll need some engineering support for sure. It’s little support unless you want to go into complex automation work. On the agent side, that’s where you need a team to help you a little bit.

Arham: All right, thank you to everyone listening. If there’s anyone who’d like to read more about this, you can read Yann’s latest Substack article about AI Agent Builder VS Agentspace. As far as the kind of work Zazmic does, we do have workshops that can help you get kick-started on Agent Builder and Agentspace projects. If you do want to dive deeper into these tools, do reach out to us.