AI Watch

Blog archive

AI Watch: Building AI Solutions with Scaled Agile in Government

This week I had the opportunity to join the leaders and practitioners implementing AI projects with scaled agile in government organizations. There was quite a bit of buzz around the topic of AI, especially with some of the new federal guidance that had been released. The federal government released a memo that will help federal agencies advance the efforts around governance, innovation, and risk management when it comes to AI. Here are three main points that echo what I have been working on with companies in the private sector.

The importance of AI governance
AI governance is a key component of an AI strategy that helps to identify and mitigate risk. It's important for companies to identify a chief leader responsible for spearheading this effort, establishing the vision, and delivering accountable results. 

Responsible AI in Innovation
It's critical to ensure that all AI projects start with responsible AI safeguards in place. This means that organizations must think about building in AI safety systems and implementing an AI Red Teaming strategy. This includes having human-ai experience testing, prompt engineering, model selection and evaluation guidelines as well as deploying to a well architected public cloud.

Establishing a framework for managing and mitigating risk
This can be a complicated part of the process, but organizations do not have to go through it blindly. NIST (National Institute of Standards and Technology has issued an initial public draft of the AI Risk Framework. This framework helps both federal agencies and commercial companies to have a common understanding of risk and what to be considering as AI projects begin and continue.

The good news is that there are long-researched approaches for managing risk and fairness in AI deployments. As a company, whether government-aligned or not, there are clear guidelines and best practices that have emerged you can take advantage of as you begin your journey.

Among the most needed skills in these types of projects are those practiced in the art and science of scaled Agile. In the past 2 years and 25+ generative AI projects, I have come to realize that the scaled agelist is the glue that holds the engineering and AI work together and sees it to a successful end. 

When it comes to AI projects, it's less about the technology as it is about what will be available through applied APIs, and much more about humans and change. In a recent "Change Management in the Age of AI" workshop, one of the attendees asked how they would be able to motivate their leaders to invest and innovate when they seem to be focused on cutting costs and saving money. I explained that we are now in a world that prefers show versus tell. People want to see the value before they invest in it. This is why I have coined the phrase "from the boardroom to the whiteboard to the keyboard." It’s important to get alignment between the business leadership and its technical leadership, and then to build something that is small, yet remarkable, to demonstrate the concept. I call it an "MRP," or minimum remarkable product

Innovation can be stifled by taking on too much too fast. A minimum remarkable product focuses on a small, meaningful feature that generates an impressive result. There are three benefits to starting with an MRP:

  1. You will create a project that has a specific goal with measurable results
  2. You will have a short timeline and be able to deliver results quickly
  3. You will generate momentum, which will empower you to take on more MRPs

One way to find an MRP is to study the activities in which your customers are engaged. Specifically, examine those difficult activities your customer must engage in. Look for that pain point. If you can find a process that's filled with friction, you'll have an opportunity to leverage AI to reduce that friction and delight the user. Always focus on creating a remarkable solution that delights the user you are trying to help. This will create the momentum you need to take on additional use cases and ensure that you stay focused on the most important thing: features your customers need and will love.

If you have questions about MRPs and building responsible AI at scale, reach out to me on LinkedIn any time!

Posted by Noelle Russell on 05/20/2024