The Pilot-Optimize-Rollout (POR) Process

When building and launching a generative AI solution

for your business, you might be tempted

to dive in head first. You might think, "Let's just build it and beat

our competitors. Speed, speed, speed, right?" Now, this kind of thinking

can be very risky. Generative AI is a

groundbreaking technology, yet is still very much

an emerging technology. Like with any

emerging technology, it's easy to get it wrong

and get it wrong fast. So before rushing into action, let's take a step back and create a disciplined

process of doing this. This will save you time, money, and resources by focusing them at the right place

at the right time. Now, I developed this approach by working with and analyzing the successful digital

transformation journeys of numerous companies across

many diverse sectors, including energy,

manufacturing, finance, education, retail,

and healthcare. What we've found is that successful AI transformation

projects follow a structured process

that breaks down the overall project into more manageable

iterative phases. With the PAD framework, you already got a

GenAI project plan tailored to your business needs. Now, I suggest you structure your subsequent actions

in three phases: pilot, optimize, and rollout. Here are the main

actions in each phase. In the pilot phase, we first

focus our energy on building the initial proof of concept

or minimally viable product. So gather your development team and start building right away. It doesn't have to be

pretty or super polished. What you need is a

working version that demonstrates the basic

capabilities of your AI solution. This is where your planning from the PAD framework

comes into play. Leverage your understanding

of the problem, the required abilities, and your data to shape

this initial build. After you have it, we focus

our resources on quickly testing this initial

product with a small controlled

test environment. The goal is to get

a picture of what an initial user experience looks like as soon as possible. Depending on whether your

solution is intended for internal use or for

external customers, you'll want to test it

with the right audience. Gather a small core

unit of users and start observing their usage

and get feedback right away. This stage will often

take some time, maybe a few weeks,

but don't skip that. Because for one, this

phase helps reveal any unforeseen technical issues

or user experience gaps. More importantly, the

feedback that you gather from this step

will be invaluable in helping your development

team refining the initial solution into something that the

users like to use. In this stage, the

optimize phase, you'll be focusing on improving

the user experience and finalizing a polished version 1 of your solution for launch. In fact, based on the findings

from the pilot phase, you might find that you need

to make quite a few changes. Perhaps integrating

new data sources, connecting to other systems, improving interface, or even reworking some core

functionalities. It's far better to

get these worked out early on rather

than after launch. Once you have fine-tuned

your solution, it's time for the rollout phase. This is where all your

hard work comes together. You launch version 1 of your generative AI solution

to a broader audience. But remember, launching

is just the beginning. During this phase, you

also need to define and closely monitor key

performance metrics. You need to track user

engagement, return on investment, and any other relevant

performance indicators specific to your solution. Of course, monitor

any risk factors that your solution

is exposed to. The rollout phase is a cycle

of continuous improvement. As your solution scales, you'll learn more about how it's used and where it needs

further adjustments. Stay agile and be

ready to implement changes as you gain

more insights. To sum it up, successful

GenAI transformation projects follow a structured process. Five main steps and three

main phases: pilot, optimize, rollout, or POR

if you like acronyms. This approach helps break down the complex task of deploying GenAI into manageable steps. Now, let's see how this action sequence

is implemented step by step in actual large-scale

project of GenAI adoption, the Maizey AI

project at Michigan. We'll look at each steps of the process from the

initial pilot through the full rollout and see what lessons we can learn to

apply to your own business.