Getting started with a POC
1. Getting started with a POC
Now that we've seen what AI is, what it isn’t, the value it can bring, and how to identify use cases - let’s move to the next step in an implementation strategy. That is building a proof of concept.2. What is a proof of concept?
A proof of concept, or POC, is a small realization of an idea to demonstrate its feasibility and potential. Through a POC, decision makers will be able to make educated, more informed decisions on value and potential risks. For example, a car proof of concept can demonstrate functionality.3. Phases of a POC
A proof of concept, or any AI solution, follows these phases. Identify, and confirm, the opportunities;4. Phases of a POC
Determine the goals of the POC and characterize the problem;5. Phases of a POC
Profile the data;6. Phases of a POC
Architect and design the system;7. Phases of a POC
Develop and test the AI model and entire system;8. Phases of a POC
Deploy the solution and evaluate its performance;9. Phases of a POC
and finally, scale and govern. We learned how to identify and confirm a use case in the previous lessons. We will cover setting goals, profiling data, architecting the system, and developing and testing in this chapter.10. A successful POC
With these phases in mind, a successful POC, or one that shows value and scales to a full implementation, has clear goals and exit criteria, uses available and representative data, is supported by an appropriate and efficient infrastructure, and is designed, created, and maintained by skilled employees or development partners.11. Defining clear goals
Clear goals are created from an intimate understanding of the value proposition,12. Defining clear goals
the end-user,13. Defining clear goals
the timeline,14. Defining clear goals
who is funding it,15. Defining clear goals
and business metrics and similar details. These should be defined in collaboration with the project team and stakeholders. Through these conversations and goal setting, an understanding of the exact problem will be developed.16. Understanding the end-user
Defining the end-user can help refine the value and goals as well as provide requirements on the architecture of the POC. Further, it can be a guide for how the AI solution either integrates into an existing system or requires a new tool. Here are some questions to help guide this process. Who will be using this solution? How many people will be using the POC? The full solution? How will they be using it? What decisions will be made based on the solution? How interpretable does it have to be? Why would this AI solution, over an existing or non-AI solution, bring more value to them?17. Understanding timelines
Each AI POC is different, so it's difficult to provide an average time frame. Knowing clear time constraints will help in feature prioritization. Here are some questions to think about. Is there a pending deadline that requires an outcome from this POC? If so, how much time is available? What would be the important milestones?18. Understanding costs and funding
Even a POC requires upfront funding to build. Knowing some high-level details on how the AI solution will be built - or bought - is helpful to determine if there is enough funding for the POC. Will you be buying versus building some or all of the components? There are pros and cons to both which we will learn more about in an upcoming lesson. Will it be cloud-based or on-premises? Is appropriate funding available? Who is funding the POC? As you can see there are a lot of questions to answer before designing the POC.19. Let's practice!
Before moving on to that phase, let's test your knowledge!Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.