Agent Bricks Multi-Agent Supervisor Demo

Key Takeaways

I'm going to walk you through our new product, Agent Bricks. Agent Bricks builds agents autooptimize on your data. You start by specifying your problem. So, you give us your data and tell us the task. That would be one of the four use cases listed here. We then optimize on your enterprise data. We build you the best agent system that balances cost and quality for your use case. Finally, we leverage a brand new research technique called agent learning on human feedback to guide the responses and optimize for the right quality for your use case in a way that's natural for humans and also natural for the model to understand. I'll give you a demo of this with our multi-agent supervisor. In multi-gen supervisor, you can combine the power of Genie for text to SQL for structured data with knowledge assistant for unstructured data to give you insights across all of your data sources. In this example, I have a knowledge assistant endpoint that can answer technical support queries, knowledgebased articles, etc. I also have several genie spaces to give me information about billing accounts and plans and products. This lets me start to ask more complex questions like the customer keeps being seeing speed reduced on their phone. What account plan are they on? How much data have they used the cycle? And how can we get their speed back? So the supervisor will start by coming up with the first question it wants to ask. So it's going to ask the account agent what plan is this customer on? So that's the first piece of information I need to collect to understand are they over their limit? have they exceeded that for this month or are there potentially other issues going on with their account? So, it reaches directly out to Genie and it comes back with them being on a premium individual plan. Next, it's going to check their usage data. So, it's going to reach out to that billing agent and understand how much data they've used in this current billing cycle so that we can start to get a grasp of what's happening with this customer. So we can see here that they have used 474,000 megabytes which might give us a hint of what's happening with this customer. And so now I'm going to reach out to the plans and products to understand what the data limits and throttling policies for premium individual are. So we can see that it has a 10 gigabyte limit and it gives me a description of what that plan involves. And then finally it's going to reach out to my support agent to understand what speed reduced might mean and how we could potentially improve that. And we can see in here with knowledge assistant that it's going to reach out and start to search my knowledge base as well as my support tickets to understand how we could increase that speed for our customer here. So it's going to structure the response in order to explain why the speeds are reduced, list immediate options and advise on escalation if issues persist or seem incorrect. So then we can see that they could potentially purchase an additional plan, upgrade their data plan or connect to Wi-Fi. And I even get footnotes here so that I can ensure that my agent is actually answering correctly. And then finally, I get a really nice summary here from my supervisor. So, it's collected all of this information across all of my different spaces. So, I can see that they're on a premium individual plan. They have a 10 GB limit and they've used approximately 463 GB. So, that's clearly why they're seeing speed reduced. And I'm also able to have some actions for the customer and for my support agent to take. And so this shows you that just in a few minutes, you can take some of the agents you've already had built and start to get some really deep insights and really enable data intelligence at your company at scale.