Resource requirements

- As you go through the considerations you need to take into account to build, use, and maintain an agentic system, a very important step is to think about your resources. When implementing AI agents for your business, you need to fully understand all of your resource requirements if you want to be successful. Now, I like to break the problem down into four separate categories. There's computing infrastructure, data resources, human expertise, and of course the financial investment. Let's start with the computing infrastructure. Agents, particularly those powered by LLMs, can be very computationally intensive. Your options will include cloud-based solutions that offer scalability with low upfront costs. There's on-prem infrastructure for greater control and potentially lower long-term costs, and there's the hybrid approach that balances both. The right choice ultimately depends on your existing infrastructure, security and privacy requirements, and the complexity of your desired agentic workflows. Next up are the data costs. Next up are the data costs. The types of data resource that you need to consider are things like the training data that will establish the baseline capabilities, the integration data from your business systems, the operational data that you gather as you use the agents over time, allowing you to improve them. You should also plan for both storage capacity and effective data governance. Remember, the quality of your data will directly impact your agents' performance. You cannot compromise on this. Human resources are often overlooked, but they are absolutely critical. Don't make the mistake of undervaluing them. You'll need technical experts for development, integration, and maintenance. You'll need domain specialists who deeply understand your business process. There's monitoring personnel for ongoing supervision, and there's end-user trainers to ensure smooth adoption. For most businesses, this means either hiring new talent, training existing staff, partnering with vendors, or some combination of all of the above. Finally, let's discuss the financial considerations. You do need to budget for initial development and deployment costs, as well as things like ongoing operational expenses, including API calls, compute resources, and all that. You'll also have to think about regular maintenance and updates, training and change management, and there's going to be surprise costs here. I find it's really good advice to allocate about 20 to 30% above your basic implementation costs to ensure that contingencies are covered and that optimization opportunities will be met. So as you plan your agent implementation, remember, resource requirements aren't just important, they also evolve over time. Keep a close eye on them, and it's always good advice to start with a pilot program to establish that baseline before you scale up. And of course, always keep a robust monitoring system to ensure that resource usage is optimized and that you have the nimbleness to adjust as your agents mature and your use cases expand. So in the next video, we're going to explore considerations when integrating agents with your existing systems.