Leading AI Implementation Projects

Contents

Introduction

AI implementation projects share many similarities with Citizen Developer implementations. Here’s a common Citizen Developer scenario: your organization just launched a low-code platform. The goal of the launch was to put development capability in the hands of business users so IT support isn’t required for every process improvement. The sponsor was energized, the demo was convincing, and the training and communication all went according to plan. Then, the six-month check-in arrived.

The six-month usage numbers for the developer hub have been met, but the resulting apps weren’t used. The ones being used didn’t undergo any review process, and IT expressed frustration because they hadn’t seen the expected lift. Business users expressed confusion about what tools they were allowed to build. And the resistance was expressed in comments like, “I don’t know what I’m supposed to own here.” Which wasn’t defined before launch.

In this situation, the practitioner spent more time identifying the root causes of the issues than building the original change strategy. However, this wasn’t a failure of the change management strategy. The typical strategy wasn’t built for this situation because the change was too dynamic for a strategy designed for stable outcomes.

While AI implementations start similarly to Citizen Developer implementations and support, they are far more complicated. AI implementations add layers that low-code platforms don’t have because AI technology participates in the work. The technology typically surfaces recommendations, informs decisions, and in some cases initiates actions. Add to that the fact that the tech continues to evolve after go-live, and the change strategy and plan must keep evolving too. An added wrinkle is that when something goes wrong, accountability gets really complicated, very quickly.

This guide is built around the identified layers of AI implementation projects. It won’t ask you to throw out what you already know, or replace your existing methodology. It will show you exactly where your existing approach needs to be extended and give you the tools to do so. By the time you’ve worked through the sections relevant to your initiative, you’ll know how to classify your AI initiative accurately. Once the classification is defined, you’ll be able to design a strategy around how the workflow changes, read resistance patterns that trace back to workflow, accountability, and trust gaps rather than awareness or skill gaps, and measure adoption in a way that reflects the actual workflow impacts.

This guide was written to close those gaps by extending what you already do to meet the conditions AI creates.

Every section is written to be used during an initiative, under pressure, with a specific problem in front of you. You should be able to open the right section, orient in a few minutes, and decide on your course of action. That’s the standard this guide was built to meet.

Who This Guide Is For

The guide was written for experienced OCM practitioners who have run implementations before and are now working on, or preparing for, their first AI initiative. It assumes solid change management fundamentals and focuses specifically on where those fundamentals need to be extended.

It is also written for practitioners who are mid-initiative and already facing issues they are unsure how to address, whose usage metrics are being met but whose workflow impacts aren’t being realized, or who are seeing resistance that typical change tactics aren’t resolving. If any of that sounds familiar, this guide can help you diagnose what’s happening and adjust before it becomes a bigger problem.

What’s Already Showing Up

Enterprise AI implementations are still in their early stages, so the field is learning in real time. But patterns are already emerging. Adoption metrics look fine while the real work quietly stalls. The activities practitioners rely on to drive adoption produce engagement without behavior change.

There’s more to manage than a typical rollout prepares you for. AI doesn’t hold still after go-live the way past technology did. The same prompt returns different results, the vendor ships new capability, people find uses no one planned for, and what counted as good use a few months ago no longer does. That movement shapes how much people trust the tool, how the work actually gets done, and how decisions get made, so the change keeps developing well past launch. The guide is built for that reality, and it shows you where to watch and what to do as the initiative keeps shifting under you.

Getting ahead of these patterns is one of the most practical things a change practitioner can do right now.

These are the conditions this guide was designed to address. Not as edge cases, but as the predictable reality of what AI change work looks like in practice.

What Prepared Looks Like

Each section in this guide is built around a gap that AI implementations create and standard OCM practice doesn’t yet close.

You’ll be able to accurately classify your AI initiative, not only by the type of technology being implemented, but by how it behaves in practice and what that behavior means for your change strategy. That classification changes what you assess, what your strategy emphasizes, and which tactics you’ll employ first.

You’ll be able to identify and respond to AI-specific resistance patterns before they take root. There are seven distinct patterns. Each pattern is associated with a specific diagnosis and a response. You won’t have to guess whether you’re dealing with a trust problem, a workflow problem, or a manager reinforcement gap.

Your change experience or change management strategy will define where AI belongs in the workflow, how decisions should be made with AI input, how accountability is assigned, and how governance gets translated from policy into steps people can follow.

You’ll know how to set adoption targets that reflect how AI works in practice. In most AI initiatives, target pressure arrives before the workflow is stable enough to support it. You’ll be able to shape that conversation by proposing targets that hold up to scrutiny, identifying when a target will drive the wrong behavior before it’s locked in, and making the case for stage-appropriate expectations when pressure arrives ahead of the work.

What makes AI change work more complex is not the technology. It is reading whether resistance is a trust problem, a workflow problem, or a manager gap when the indicators look identical, calibrating how much people should rely on a tool that is sometimes right and sometimes confidently wrong, and keeping a strategy intact when the system itself changes under it after launch. Those are judgment calls experienced practitioners bring to this work, and they are what this guide is built to support.

Governance on AI initiatives is harder than the governance you’ve translated before, because the rules need to cover a system that acts, sometimes decides, and keeps changing. Making those rules usable in real work and keeping them usable as technology shifts is part of what this guide prepares you for.

What Each Chapter Does

The guide moves you through each stage, from initiation and diagnosis to strategy, execution, and sustainment. Each chapter builds on the previous one when read in sequence, but each also works as a standalone reference when you’re entering mid-initiative. Here’s what you’ll find in each one.

How This Guide Fits Into Your Current Practice

This guide doesn’t replace your change management practice. It changes where you aim it. On most implementations, change activity starts with who is affected and how to build awareness and impact adoption. AI work starts somewhere else: how the work will be performed, how AI outputs get interpreted and trusted, and where validation and accountability sit once AI is in the workflow.

This guide uses Change Experience Design (CXD) as the lens for that work: designing how employees encounter, understand, adopt, and integrate AI into their daily tasks. CXD layers onto the change methodology you already practice rather than replacing it. The chapters that follow are organized around it.

You will still carry out familiar activities. You will still assess, design, enable, and reinforce. What changes is what you look for. You’ll surface more, and what you surface shapes what you act on and design toward.

The Through Line

Across every stage of an AI initiative, five things should stay connected. Workflow: how work is performed with AI. Decision-making: how choices are made, evaluated, and owned. Trust: how people judge and rely on AI outputs. Governance: how expectations are applied in real work, not only in policy language. Measurement: how you know whether any of it is working.

If any one of these drops out, adoption weakens. The chapters in this guide are organized around these five threads. Each chapter addresses one or more of them directly. Keeping them connected from intake through sustainment is the work this guide is designed to support.

How to Use This Guide

If you are preparing to implement AI technology, read this guide from front to back before the initiative begins. If you’re in the middle of developing a strategy, work through Chapters 1 through 5 before it is finalized because those chapters can be used to shape how you approach the work from the start. Chapters 6 through 9 are active references you can use during execution and sustainment.

The table below maps common situations to starting points. You don’t need to read everything before a section to use a section effectively; each section is written to stand on its own.

What Comes with This Guide

This guide does not stand alone. It is one part of a connected set of materials designed to work together. The chapters here provide the framework and the rationale. The materials below make it operational.

Before Chapter 1, you’ll find two unnumbered prefixes. The Start Here document sequences the core instruments and one entry path for practitioners who need to move immediately on a new initiative. It tells you which instruments to activate first and in what order, before you’ve read the full guide. The Ecosystem Map shows every instrument in the EmpowerChange toolkit, organized into three zones: the core sequence used on every initiative; the condition-based additions activated by specific signals; and the situational instruments activated by specific routing signals.

Two documents that accompany the guide show the framework in action.

A Note on How This Guide Was Built

The patterns in this guide stem from recurring dynamics in enterprise AI implementations. Not from theory. From the predictable places where well-executed change management still falls short, and the specific adjustments that make the difference. The frameworks here reflect what experienced practitioners encounter when AI change work hits reality.

The guide will evolve as AI implementations evolve. The technology isn’t static, the organizational challenges it creates aren’t static, and the change management response shouldn’t be either. What’s here now addresses the conditions most consistently shaping AI adoption challenges in enterprise contexts today.

Your experience matters to that process. If you find a pattern this guide doesn’t address, a scenario where the framework breaks, or a section that could be sharper, that feedback belongs in the next version. The guide improves the more practitioners use it in real-world conditions.