Digital transformation looks new. The change management underneath it isn’t. The same leadership, communication, and process discipline that drove past transformations — from ERP rollouts to cloud migrations — drives this one too.
Strategic Reminders Before You Start
These four principles have derailed more AI initiatives than any technical failure. Get them right before anything else.
Avoid "technology-led" projects — those that start with a tool and search for a use case. Successful AI initiatives start with a clear business problem and define measurable outcomes before a single line of code is written.
AI is "garbage in, garbage out." The quality of your outputs is bounded by the quality of your inputs — your data, your process documentation, your system integrations. Strong foundations aren't glamorous, but they're the difference between AI that works and AI that looks like it works.
Use high-impact pilot programs to generate quick wins before attempting full-scale rollout. A customer service team that handles tickets 40% faster is a story you can tell. That story builds the organizational confidence — and the political capital — to expand.
Automating a step is a good start — and sometimes it's all you need. But it's also an opportunity to ask: now that we're touching this process anyway, could we do it better? Not every workflow needs to be redesigned. But many benefit from a fresh look when AI enters the picture. The question isn't just "how do we automate this?" — it's "while we're here, is this the best version of this process?"
Two Types of Change
Not all AI initiatives require the same level of organizational disruption. Understanding the type of change you’re undertaking shapes how you manage it.
Adaptive change is incremental — adding AI assistance to one workflow at a time, improving without overhauling. Lower disruption, easier to absorb, faster to show results. Good for building confidence and learning before committing to bigger bets.
Transformational change means redesigning how the organization fundamentally operates — new roles, new processes, new ways of making decisions. Higher impact, higher risk, and it requires more deliberate management. The ADKAR framework below is especially important here.
The Change Process
Every successful transformation — whether AI-powered or not — moves through five stages. Skip a stage and it tends to come back and bite you later.
Assess readiness honestly. Surface resistance before it surfaces in production. Build the case for change with the people who will be most affected — not just the people who decided on it.
Define what the future state looks like in concrete terms. Align leadership on both the destination and the route. Ambiguity at this stage creates confusion and competing interpretations downstream.
Execute in stages, not all at once. Communicate constantly during this phase — more than you think is necessary. Keep feedback channels genuinely open and act on what you hear.
Update processes, job descriptions, training, and incentives so the new way becomes "how we do things here" — not a special initiative that fades when attention moves elsewhere.
Measure against the original goals — not the easy metrics. Learn from what didn't work. Adjust rather than declaring premature victory when demos look good.
ADKAR: The Human Journey Through Change
Technology is easy. People are complicated.
When you introduce AI into a team’s workflow, the primary emotions you’ll encounter are fear — fear of obsolescence, fear of making mistakes with a “black box,” fear of losing control over work that previously felt familiar. The ADKAR model maps the human journey that every individual needs to travel for change to actually stick.
People need to understand the business reason for the change before they can embrace it. "Our manual ticketing process is bottlenecking growth" is a real reason. "AI is the future" is not.
"The AI handles the copy-pasting so you can focus on the work that actually requires your judgment." Desire is personal. Generic benefit statements don't create it — specific, role-level answers do.
This is where AI literacy and prompt training come in. Not a one-hour all-hands. Role-specific guidance on the specific workflows that are changing.
The right UI and workflow integration matters. Employees shouldn't need to be engineers to trigger an AI workflow. If the tool requires technical skill that the team doesn't have, ability breaks down regardless of desire.
Celebrate wins where AI made work genuinely easier. Share those stories. And critically: don't use AI adoption metrics to justify headcount cuts — that's the fastest way to make the first four letters of ADKAR collapse.
“If it ain’t broken, don’t fix it” — but is it really not broken?
Japan famously continued using fax machines in government offices well into the 2020s. They worked fine internally — reliable, familiar, trusted. But as digital-first businesses and citizens increasingly expected email and web interfaces, the fax machine became a point of friction with the outside world. In 2024, the Japanese government officially retired the last government fax requirement.
The lesson: the right question isn’t “does this work?” It’s “does this keep working as the world changes around it?” Some processes are genuinely fine. Others are fine right now — and quietly accumulating technical and competitive debt.
📝 Key Concepts
- Business first: define the problem and the ROI before choosing the technology
- Two types of change: adaptive (incremental) vs. transformational (overhaul) — they require different management intensity
- Five-stage process: prepare → vision/plan → implement → embed → review; skipping stages costs more later
- ADKAR: the human journey through change; all five stages must be met for adoption to stick
- Reinforcement matters most: people sustain change when it visibly improves their work — not when it’s mandated
In the ADKAR model, what does the "D" (Desire) stage address?