1Chapter 3: Change Management¶
Figure 1:One person’s mindset shift becomes a team’s new normal, which becomes a department’s standard practice, which becomes an organization’s competitive advantage. Change radiates. It does not cascade.
“Culture does not change because we desire to change it. Culture changes when the organization is transformed — when organizations take new actions, new behaviors become the norm.” — Frances Hesselbein
Here is a scene that plays out in banks all over the country, right now.
The CEO sends an email to all staff: “AI is a strategic priority for us. We are committed to responsible adoption. Training resources are available. Please reach out to your manager with questions.”
Ninety days later: nothing has changed.
Not because the people don’t care. Not because the tools don’t work. Not because the strategy is wrong. But because a message is not a movement. An email is not a transformation. And announcing a priority is not the same thing as building a system for adopting it.
Chapter 2 gave you the personal operating system — the mindset. This chapter gives you the organizational operating system — the change architecture. How you take what you now believe about AI, what you’ve started to practice, and what you’re beginning to build — and turn that into something that outlasts you. Something that spreads.
Because the most important thing a BankUnited professional can do with their AI capability is not just use it better. It is multiply it — across their team, their department, and their organization.
That is what this chapter is about.
1.11. Why Change Fails — And the One Thing That Works¶
Let us start with the uncomfortable data.
We already know from Chapter 2 that only 5% of companies globally qualify as “future-built” for AI — meaning they have genuinely restructured how they work, not just added AI tools to their existing processes. Meanwhile, 60% report minimal or no measurable value from their AI investments despite real financial commitment.
Why?
The answer is not technology. McKinsey’s research is definitive on this point: 55% of organizations that are seeing AI returns redesigned their workflows. They did not add Copilot to the old way of doing things and expect it to be transformative. They asked: “Given that this tool exists, what is the new right way to do this work?” And then they built that new way.
The organizations that are not seeing returns are, almost universally, running the old process with a new tool sitting next to it. Copilot open in one tab, the same manual workflow proceeding in another. The tool gets used for occasional convenience. The fundamental way of working never changes.
This is not a technology problem. It is a workflow redesign problem — which is, at its heart, a change management problem.
Figure 2:The technology is not the variable. The workflow is. Organizations that redesign how work gets done — not just which tool they use — are the ones capturing exponential returns.
The good news: workflow redesign does not require a top-down transformation initiative. It does not require a six-month consulting engagement. It does not require IT to rebuild anything.
It requires one person — perhaps you — to ask: “What if we just didn’t do it the old way anymore?” And then to prove that the new way works. And then to show someone else.
That is how every lasting technology adoption in history has actually happened. Not from the top. From the middle. From the practitioners who tried something, discovered it worked, and couldn’t stop talking about it.
1.22. Three Frameworks That Actually Work¶
Change management as a field has accumulated an enormous library of frameworks, most of which are useful in academic contexts and largely ignored in practice. We are going to focus on three — not because they are the most academically sophisticated, but because they map directly to what is actually happening when AI adoption succeeds or fails inside a bank.
Figure 3:Three frameworks, one purpose: making change stick instead of fade. Each illuminates a different dimension of the same transformation.
1.2.1Kotter’s 8-Step Model — The Organizational Staircase¶
John Kotter’s framework, developed from studying hundreds of organizational transformations over 30 years, identifies eight conditions that must be met for change to take hold. Miss any one of them, and the change stalls. Understanding where your organization — or your team — is on this staircase tells you exactly what to do next.
Table 1:Kotter’s 8 Steps — Applied to BankUnited AI Adoption
Step | What It Means | What It Looks Like at BankUnited |
|---|---|---|
| People must feel the need to change — not just be told about it | Show the 1.7× revenue gap. Show what peer banks are deploying. Make the cost of inaction real. |
| Change requires a cross-functional group of champions, not just top-down mandate | Identify 3–5 respected practitioners from different teams who believe in this and will model it |
| People need to see where they are going, not just what they are leaving behind | “In 12 months, every RM comes to a client meeting briefed by AI. Every credit package is first-drafted by Copilot.” |
| You need early adopters before you need everyone | Don’t try to convert the skeptics first. Find the curious. Start there. |
| The biggest barrier is usually permission, not skill | Make it safe to experiment. Create a sandbox. Celebrate attempts, not just successes. |
| People need evidence it works before they commit | Showcase Projects (Part IV of this book) are engineered for exactly this step. |
| Early wins get used to justify slowing down. Don’t let that happen. | Use wins to expand the program, not to declare victory. |
| Change becomes the new default — not a project, but the way we work | AI-assisted workflows are in the onboarding checklist. New hires learn them on Day 1. |
The most common failure point in AI adoption at banks is steps 4 and 5: organizations try to enlist everyone at once (skipping the volunteer army) and fail to remove the real barriers (usually permission, not training). Fix those two steps and the rest follows faster than you expect.
1.2.2ADKAR — The Individual Change Model¶
Kotter describes the organizational staircase. ADKAR, developed by Jeff Hiatt at Prosci, describes the individual journey. Because organizations don’t change — people do. And they change one at a time.
ADKAR is an acronym for the five things an individual needs before they will sustainably adopt a new behavior:
card-carousel - Unknown Directive
card-carousel - Unknown Directive:::{card} **A — Awareness**
Understanding *why* the change is necessary. Not "AI is a strategic priority" — but *why*, specifically, your role and your work are affected and why now is the moment.
:::
:::{card} **D — Desire**
Wanting to participate in the change. Awareness without desire produces compliance, not adoption. Desire comes from connecting the change to the person's own goals and interests.
:::
:::{card} **K — Knowledge**
Knowing *how* to change. This is where most training programs start — and where ADKAR reminds us it's the third step, not the first. Skipping A and D and going straight to K is why most training doesn't stick.
:::
:::{card} **A — Ability**
Being able to apply the knowledge. There is a gap between knowing how to do something and being able to do it under real conditions. Ability requires practice, feedback, and patience.
:::
:::{card} **R — Reinforcement**
Having the change sustained and recognized. Without reinforcement — positive feedback, recognition, peer visibility — new behaviors revert. Reinforcement is not a nice-to-have. It is what separates a training event from a culture shift.
:::When AI adoption stalls at BankUnited — on your team, in your department — diagnose which ADKAR stage is the bottleneck. Nine times out of ten, it is not Knowledge (people don’t know how). It is Desire (people don’t see why it matters to them personally) or Reinforcement (people tried it, it worked, and then nobody noticed).
Fix the right stage. Don’t add more training when the problem is recognition.
1.2.3Bridges’ Transitions Model — The Emotional Truth¶
William Bridges made a distinction that most change management frameworks miss entirely: the difference between change and transition.
Change is the external event. A new tool is deployed. A new process is mandated. A new strategy is announced. Change happens on a calendar date.
Transition is the internal journey. It is how the people inside the organization move from the old way to the new one — emotionally, psychologically, professionally. Transition has three phases, and it does not start with the change. It starts with endings.
1.33. Why AI Adoption Fails — The Five Organizational Failure Modes¶
Figure 4:The five failure modes that consistently derail AI adoption at financial institutions — each preventable with the right organizational design.
With the frameworks in place, let us name the specific failure modes that appear most consistently in AI adoption programs at financial institutions. These are not theoretical. They are drawn from the pattern across organizations that invested significantly in AI tools and measured minimal returns.
Failure Mode 1: Tool-First, Workflow-Never The organization deploys the tool, runs a training event, and considers adoption complete. No one asks: “What specific workflows are we redesigning?” The tool becomes optional. The old workflows continue. The ROI doesn’t materialize.
Failure Mode 2: Compliance Without Conviction Employees complete the mandatory training, achieve their certification, and never open the tool again. Because the training addressed Knowledge (ADKAR step 3) without ever building Awareness or Desire (steps 1 and 2). The training happened. The change did not.
Failure Mode 3: Champions Without Authority The organization appoints an “AI Champion” who is enthusiastic but has no team, no budget, no recognition, and no power to remove barriers. The champion burns out. The program fades. The lesson incorrectly drawn: “AI champions don’t work.” The correct lesson: champions without authority are ceremonial, not structural.
Failure Mode 4: Perfection as the Enemy of Progress Governance teams spend six months building a perfect AI policy before anyone is allowed to experiment. By the time the policy is approved, the technology has moved twice and the policy is already outdated. Meanwhile, competitors who started experimenting six months ago have six months of learning that your organization doesn’t have.
Failure Mode 5: Showcase Without Scale A brilliant Showcase Project is built. It wins internal recognition. And then it lives on a SharePoint page that nobody visits. Because the program never built a mechanism for taking individual innovations and turning them into team or department standards.
1.44. Finding and Fueling the AI Champions¶
Every successful bottom-up AI adoption story has the same cast of characters at its center: a small group of enthusiastic early adopters who were empowered to experiment, recognized when they succeeded, and given a visible platform to share what they learned.
These are your AI Champions. And finding them is more important than almost any other organizational action in the early stages of an AI program.
Figure 5:AI Champions are not the most technical people in the room. They are the most trusted — and the most willing to learn publicly.
A common mistake: organizations look for their most technical people to be AI Champions. This is almost always wrong. Technical sophistication is not the primary qualification. The qualities that make an effective AI Champion in a banking context are:
Credibility with peers. The Champion’s colleagues already respect their professional judgment. When they say, “I tried this and it changed how I work,” people listen — because they trust the source.
Genuine curiosity. Not enthusiasm manufactured for a role — but authentic interest in figuring out what is possible. You can hear it when you talk to them: they are asking questions, not just reporting answers.
Willingness to learn publicly. This is the rare quality. Most professionals are comfortable sharing successes. The Champion who says “I tried this, it didn’t work, here’s what I learned” — in front of their team, without embarrassment — is the one who makes it safe for everyone else to try.
A real workflow to test on. The best Champions are not experimenting in the abstract. They are applying AI to the actual work they do every day and reporting back from the front line.
1.4.1The Bottom-Up Pattern — How It Actually Happens¶
The most instructive case studies in AI adoption are not the top-down strategic transformations. They are the grassroots ones — the ones that started with a single person who tried something, and spread.
ING Bank — one of Europe’s most successful AI adopters — did not mandate AI adoption. They created communities of practice: informal groups where employees who were experimenting with AI could share what they were learning. No policy. No certification. Just structured sharing. Within 18 months, AI-assisted practices had spread to over 60% of their knowledge-worker population — driven almost entirely by peer-to-peer sharing.
JPMorgan Chase deployed what they internally called a “floor captain” model: one identified AI enthusiast per department floor, whose job was not to train colleagues but simply to be available, to demonstrate, and to celebrate when someone tried something and it worked. The floor captain model accelerated adoption faster than any centralized training program.
Sears — the 50-app story from Chapter 2 — was built on a similar pattern. The employees who built those applications were not following a mandate. They were solving problems they personally found frustrating, using tools they had been given access to, and showing their results to their colleagues. The organizational contribution was not the mandate. It was the access and the permission.
At BankUnited, the same pattern is available. The question is: who in your team — or in your department — has that Champion spark? And what does it cost to give them access, time, and permission?
1.55. The Four Archetypes of Resistance¶
Not everyone is a Champion. And that is not a problem — it is a reality that effective change management accounts for rather than ignores.
In our experience working across organizations at different stages of AI adoption, resistant professionals tend to cluster into four archetypes. Each archetype has a different underlying concern, and each requires a different response.
Figure 6:Resistance is not uniform. The right move depends on which archetype you’re dealing with. One response does not fit all four.
1.5.1The Cynic¶
What they say: “We tried something like this three years ago. It didn’t work. This is just the new version of the same hype cycle.”
What they actually mean: “I have been burned before and I don’t want to invest emotional energy in something that will be abandoned.”
The right move: Don’t argue with the cynicism — validate it. Organizations have hyped technology and abandoned it. The Cynic’s skepticism is earned. What changes the Cynic is not more communication about strategy — it is a concrete, visible demonstration that this time is different. Show them a real workflow that actually changed. Show them the time saved, the result improved. Give them evidence, not enthusiasm. The Cynic becomes a convert when the proof arrives. And when they convert, they become among the most credible advocates — because everyone knows they were the hardest to convince.
1.5.2The Perfectionist¶
What they say: “I want to use it, but I need to make sure I’m doing it right. I don’t want to send something out that was AI-generated and be embarrassed.”
What they actually mean: “My standards are high, and I’m afraid AI output will undermine them.”
The right move: The Perfectionist is not resistant — they are risk-averse in a way that is actually professionally appropriate for a banking environment. Give them structure: the verification discipline, the review protocol, the clear rule that AI drafts and humans decide. Give them explicit permission to treat AI output as a first draft that they improve, not a finished product they endorse. The Perfectionist, once they see that AI raises their quality floor rather than lowering their quality ceiling, often becomes an extremely disciplined and effective AI user.
1.5.3The Territorial Expert¶
What they say: (rarely out loud) “If AI can do what I do, I’m not needed anymore.”
What they actually mean: “My professional identity and my job security are connected to specific expertise, and this tool appears to be a direct threat to both.”
The right move: This is the identity layer from Chapter 2, manifesting at the individual level. The wrong move is to dismiss the concern as irrational. It isn’t. The right move is direct, honest acknowledgment: “What AI is replacing is the mechanical part of your expertise. What it cannot replace is the judgment, the client relationships, the contextual knowledge, and the ability to apply all of it in real situations. Your value does not decrease when AI does the drafting. It increases — because now your judgment operates at a higher level with better inputs.” Then prove it by giving them an AI tool that visibly amplifies their expertise rather than substituting for it.
1.5.4The Overwhelmed¶
What they say: “I want to do this, I just don’t have time to learn something new right now. Can we revisit in Q3?”
What they actually mean: “My capacity is genuinely at its limit and I need someone to make the on-ramp shorter.”
The right move: Reduce the barrier. Fifteen minutes, one use case, right now — not a training program. Not a certification. Not a self-directed learning module due in two weeks. Sit next to them, open Copilot, and apply it to something they are actually working on today. The Overwhelmed professional, once they experience the time-savings themselves, becomes motivated to invest the learning time — because they have seen the return. The barrier is not willingness. It is the apparent size of the investment before the payoff.
1.66. Communication Patterns That Build Trust¶
Figure 7:Effective AI change communication flows from executive leadership through managers and champions to every employee — with feedback loops built in at every level.
The way leaders talk about AI inside BankUnited matters enormously — more than most leaders realize. The wrong communication pattern creates fear, breeds cynicism, and makes the resistance archetypes entrench further. The right pattern builds the psychological safety that allows people to experiment, fail, learn, and grow.
Here are the communication patterns that work — and the ones that undermine what they intend:
Table 2:AI Communication — What Works vs. What Backfires
What You Might Say | What People Hear | What to Say Instead |
|---|---|---|
“AI is a strategic priority for BankUnited.” | “This is mandatory and you’d better comply.” | “AI is a professional development opportunity I want to personally support.” |
“We need to stay ahead of the curve.” | “We’re behind and I’m worried.” | “We’re at an inflection point where early movers build lasting advantages.” |
“This will make everyone more efficient.” | “We’re going to do more with fewer people.” | “This will take the tedious parts off your plate so you can focus on the work that actually requires your judgment.” |
“I expect everyone to complete the AI training.” | “Check the box. Move on.” | “I’m doing the training too. Here’s what I learned. What did you learn?” |
“Our competitors are using AI.” | “We’re reacting to fear.” | “The professionals who build AI habits now will have advantages that compound for years.” |
(Silence after a failed AI experiment) | “Trying and failing is not safe here.” | “Tell me what happened. What would you try differently? Good experiment.” |
The most powerful communication pattern of all is the one that is hardest for most leaders: public modeling. Talking about your own AI experiments — what worked, what didn’t, what surprised you — in team meetings, in one-on-ones, in informal conversations. Not as a performance of innovation, but as a genuine sharing of a professional learning process.
When leaders do this, they create permission. Permission for their team to try things that might not work. Permission to talk about AI as a real, evolving practice rather than a polished deliverable. Permission to be in the Neutral Zone together, figuring it out collectively.
That permission is worth more than any training program you can buy.
1.77. The 100-Day Playbook¶
Every successful AI adoption program needs a structured beginning. Not because structure is more important than culture — but because structure creates the conditions in which culture can form. Without a defined sequence, adoption dissolves into the urgency of daily operations and never gains momentum.
Here is the 100-day playbook that has proven most effective in financial services environments:
Figure 8:The 100-day playbook. Three phases, one goal: transform AI from a topic people talk about into a practice people live.
1.7.1Days 1–30: Awareness — Ignite the Curiosity¶
The goal of the first 30 days is not adoption. It is curiosity. You want people asking questions, not completing checkboxes. The specific activities:
Week 1: Leader communication — personal, specific, and modeling-focused. Not a strategy announcement — a genuine “I tried this and here’s what I found” message from the most credible voice available.
Week 2: Identify Champions. Three to five volunteers across different functions and seniority levels. Give them access, time, and a clear mandate: experiment and share.
Week 3: First “AI Lab” session — informal, one hour, Champions demonstrate something real. No slides, no certification. Just: “Here’s what I tried. Here’s what happened. Here’s what you can try today.”
Week 4: Open channel (a Teams channel, a Viva Engage group, a SharePoint page — whatever exists) where people can share AI experiments, tips, and questions. Seed it with Champion contributions. Make sharing feel low-stakes.
1.7.2Days 31–60: Pilots — Prove Real Value¶
The goal of the second 30 days is evidence. Specific, measurable, visible results that make the case from the inside better than any external report.
Week 5–6: Champions identify one repeatable workflow in their area and redesign it with AI assistance. Not a demo — a real workflow that the team actually uses.
Week 7: First internal “what’s working” session. Champions present: “We changed how we do X. Here is what it looked like before. Here is what it looks like now. Here is the time saved / quality improved.”
Week 8: Expand access. Based on what pilots revealed, identify the next cohort of practitioners who should be building with AI.
1.7.3Days 61–100: Scale — Spread What Works¶
The goal of the final 40 days is institutionalization. Making the AI-assisted workflow the default, not the exception.
Week 9–10: Document the redesigned workflows as standard operating procedures. Add them to onboarding materials. Make AI-assisted practice the baseline expectation for new hires.
Week 11–12: First Showcase event — internal, celebrated, visible. Not a compliance exercise — a genuine recognition of the professionals who built something and changed how their team works.
Week 13–14: Review what the 100 days revealed. What worked? What stalled? Who surprised you? What should the next 100 days look like?
The playbook is a beginning, not a destination. Its job is to build enough momentum that the change starts sustaining itself — that AI practice becomes the norm from which deviation requires justification, rather than the experiment that needs permission.
1.88. Governance vs. Permission — Finding the Line Between Protection and Paralysis¶
Figure 9:The governance spectrum: over-restriction kills adoption just as surely as reckless deployment kills trust. The right line is governed innovation.
This is the hardest conversation in every AI adoption program inside a regulated institution. And it is worth having directly.
BankUnited operates in one of the most carefully supervised industries in the world. OCC guidance, Federal Reserve expectations, FDIC standards, state banking regulators — every workflow that touches client data, financial advice, or regulatory reporting carries legal and compliance weight. That weight is real. The people who carry it are doing important work.
And here is the tension: the same caution that protects the bank from compliance risk can, if miscalibrated, protect the bank out of its competitive position.
The question is not whether to govern AI use. Governance is non-negotiable in banking. The question is: what kind of governance, applied where, at what stage?
Table 3:Governance That Enables vs. Governance That Paralyzes
Governance That Enables | Governance That Paralyzes |
|---|---|
“Here is what you are permitted to do. Start there.” | “You must get approval before you do anything.” |
“Here is the approved use case library. Add to it.” | “AI use requires a formal risk assessment per workflow.” |
“Here is what not to put in AI. Everything else is fair game.” | “AI use is under review pending policy development.” |
Fast lane for low-risk experiments, scrutiny for high-risk ones | One governance process for all AI use regardless of risk |
Policies that evolve with the technology | Policies finalized before the technology is understood |
“Go. Here are the guardrails.” | “Wait. Here is the approval process.” |
The financial institutions that are winning the AI race — JPMorgan, Goldman Sachs, Morgan Stanley, ING — all have rigorous AI governance. None of them have governance that requires pre-approval for a professional to use Copilot to draft an internal memo.
The principle is proportionality: governance intensity should match risk level. Using Copilot to draft talking points for an internal meeting is a different risk profile than using AI to generate client-facing financial projections. Treat them differently. Reserve the full compliance apparatus for the workflows where it genuinely matters, and create a fast lane for the vast majority of daily AI use that is inherently low-risk.
The goal is: go with guardrails, not wait with approval.
1.99. Map Your Team’s AI Readiness¶
Here is the tool that makes everything in this chapter actionable at the individual leader level.
The AI Readiness 2×2 maps your team members across two dimensions: their current AI skill level (from low to high) and their willingness to change (from low to high). The four quadrants that result require four different leadership moves.
Figure 10:Every person on your team fits somewhere in this matrix. Each quadrant requires a different leadership move. The same intervention applied uniformly to all four will fail for three of them.
Table 4:The AI Readiness 2×2 — Leadership Moves by Quadrant
Quadrant | Who They Are | What They Need | Your Move |
|---|---|---|---|
Champions (High skill, High willingness) | Your early adopters — already experimenting, already building | Visibility, authority, a platform to share | Appoint them. Give them time and recognition. Let them run the AI Lab sessions. |
Eager Learners (Low skill, High willingness) | Enthusiastic but untrained — asking questions, showing up to everything | Structured access, quick wins, a clear starting point | Give them Chapter 1 of this book. Sit beside them for 30 minutes. Remove the blank-page barrier. |
Coachable Stars (High skill, Low willingness) | Technically capable but emotionally resistant — often the Territorial Expert archetype | Psychological safety, acknowledgment of loss, visible proof | One-on-one conversation. Acknowledge what they’re protecting. Show how AI amplifies it. |
Skeptics (Low skill, Low willingness) | Haven’t started and don’t want to — often the Cynic or Overwhelmed archetypes | Evidence, not training. A win they can see, not a process they have to follow. | Don’t start with them. Start with Champions. Let their results create FOMO. |
The strategic insight in this matrix: do not start with the Skeptics. It is a natural impulse — the resistant are visible, and organizational culture is sometimes held hostage by them. But starting with skeptics is fighting uphill. Start with your Champions and Eager Learners. Build real results. Make those results visible. The Skeptics will move when they see evidence — and some will not move until they see their colleagues succeeding without them, which creates the only motivation that actually works for the deeply resistant: social proof combined with mild competitive pressure.
1.10🧪 Try This — Build Your Team’s AI Change Map¶
These exercises are practical change management tools you can use immediately. They work whether you are a team lead, a department manager, or a senior individual contributor who cares about your team’s trajectory.
1.1110. The AI Readiness Questionnaire — Your Culture Diagnostic Tool¶
Everything in this chapter — the frameworks, the archetypes, the 100-day playbook — works only if you have an honest picture of where your team actually stands today. Not where you hope they are. Not the official narrative. The real picture.
That is what this diagnostic tool gives you.
The AI Readiness and Integration Questionnaire is a 15-question instrument designed to surface three things simultaneously: where each individual sits on the AI maturity spectrum, how the team perceives the organization’s AI capability, and where the highest-value AI automation opportunities are hiding inside your day-to-day work.
→ Take the AI Readiness and Integration Questionnaire
1.11.1Understanding the Five Dimensions the Questionnaire Measures¶
The 15 questions are not random. They are organized around five diagnostic dimensions, each revealing a different layer of readiness.
Dimension 1 — Personal AI Sovereignty (Questions 3, 4, 5)¶
Where are you on the AI maturity scale?
These questions assess your personal relationship with AI — not your organization’s, yours. This matters because organizational AI adoption is always, at its foundation, the aggregate of individual journeys. You cannot build an AI-capable culture from people who have never personally experienced AI doing something useful.
The AI maturity scale for individuals runs from five recognizable stages:
Table 5:Personal AI Maturity Scale
Level | Stage | What It Looks Like | Score Range |
|---|---|---|---|
1 | AI Unaware | Has not meaningfully engaged with AI tools. May have heard of ChatGPT but hasn’t used it for real work. | 1–2 |
2 | AI Curious | Has experimented — tried a few prompts, played with a chatbot, seen a demo. But nothing has changed how they work. | 3–4 |
3 | AI Functional | Uses AI tools weekly for specific, repeated tasks. Has found at least one thing AI does reliably well. Getting value, but still dependent on the AI’s defaults. | 5–6 |
4 | AI Integrated | AI is embedded in daily workflow. Prompts intentionally. Has developed personal prompt libraries, personas, and workflows. Would feel meaningfully slower without it. | 7–8 |
5 | AI Sovereign | Designs AI solutions. Builds agents, automations, and custom workflows. Thinks in systems, not just tasks. Could teach this to others. | 9–10 |
Most professionals at BankUnited will score between a 3 and a 6 — functionally aware, selectively capable, but not yet integrated. That is not a failure. It is the starting point. The gap between a 5 and an 8 is not talent — it is deliberate practice, structured exposure, and a permission environment that says it is safe to experiment.
Question 4 — What’s the most impressive thing you’ve done with AI that actually stuck in your workflow? — is the most revealing question in the instrument. An honest answer reveals whether someone has crossed the threshold from curiosity to functional use. If someone cannot answer it, their score is probably a 3 or below, regardless of what they selected in question 3.
Question 5 — What’s been your biggest barrier? — tells you what is between your team and their next level. Not enough time, not knowing where to start, distrust of outputs — each answer requires a different intervention. This is where your Champions become critical: peer-to-peer credibility removes barriers that no training deck can touch.
Dimension 2 — Organizational AI Perception (Questions 6, 7, 8)¶
How does your team see BankUnited’s AI readiness?
These three questions reveal something leaders often do not know: the gap between the official AI narrative and the experienced reality. An organization can have a formal AI strategy and still have a workforce that experiences the environment as ad hoc and unsupported. That gap is a trust problem, and it is a change management problem before it is a technology problem.
Question 6 asks what percentage of the team is genuinely AI-capable. The responses here reveal something important: how visible AI adoption is to peers. If Champions are working in isolation — getting results but not sharing them — their colleagues will underestimate the actual capability level of the team.
Question 7 assesses whether people experience the organization’s AI strategy as real or rhetorical. The difference between “we’re exploring” and “we have a formal strategy with KPIs” is not semantics — it is the difference between employees feeling that AI initiatives have institutional backing or that they are on their own.
Question 8 — the Bus Factor question — is the most uncomfortable one in the instrument. If your top three people left tomorrow, what would break first? This question surfaces something that AI cannot fix: the organizational fragility of undocumented expertise. But it is also the clearest indicator of where AI-powered documentation, knowledge capture, and process codification would create the most immediate institutional value. If the answer is “almost everything would break,” you have just identified your first AI pilot.
Dimension 3 — Role Context (Question 9)¶
What is your role?
This question replaced the original “industry” field because inside BankUnited, the relevant variable is not industry — it is function. A relationship manager, a compliance officer, an operations analyst, and a branch manager all work in banking, but they have radically different AI use cases, different risk tolerances for AI outputs, and different workflows where AI can create value.
Segmenting responses by role allows you to do targeted analysis: where are the Commercial Lending professionals experiencing friction? What are Operations teams identifying as their biggest bottlenecks? Role-based patterns produce role-specific pilot candidates.
Dimension 4 — Pain Points Addressable by AI (Question 10)¶
What workflows are high-volume, error-prone, or expertise-dependent?
This is the question that separates diagnostic surveys from actionable intelligence. Question 10 asks people to describe 2-3 critical business workflows that are either high-volume, error-prone, or require expensive expertise to execute — specifically those with the most manual handoffs or where errors cost the most.
The responses to this question are, in effect, a crowdsourced AI opportunity map. When multiple people from different roles identify the same workflow as painful, that convergence is your highest-confidence pilot candidate. The three characteristics that make a workflow AI-ready — high volume, clear structure, and tolerance for imperfect output — often align directly with the workflows that frustrate people most.
Thematic analysis of question 10 responses across your team will almost always surface 2-3 workflows where a well-designed Copilot prompt library, an agent, or an automation would create immediate, measurable time savings.
Dimension 5 — AI Automation Opportunities (Questions 11, 12)¶
Where would AI create the most value if it worked perfectly?
Questions 11 and 12 access something that structured workflow analysis often misses: employee aspiration and pain. Question 11 — If you could wave a magic wand and have AI handle one thing perfectly starting tomorrow, what would it be? — reveals where people are most frustrated with their current cognitive load. The “magic wand” framing removes pragmatic filters and gets to the honest answer.
Question 12 asks for the single biggest bottleneck or process failure, with an estimate of cost or time lost. This question produces the data you need to build a business case. When employees themselves are estimating the cost of a broken process, the ROI conversation becomes grounded in lived experience rather than consultant projections.
Together, questions 11 and 12 give you the data to answer the three questions every AI implementation requires: Where does AI create the most value? What does success look like? How do we measure it?
1.11.2Conducting a Thematic Analysis — Turning Responses Into a Culture Change Roadmap¶
The questionnaire is most powerful when aggregated. Here is how to run a basic thematic analysis that turns 15-30 individual responses into an organizational action plan.
The goal of this exercise is not a perfect analysis. It is a shared, evidence-based starting point. When a team of 20 people has collectively named their pain points, their aspirations, and their capability gaps — and those responses are reflected back to them as a group — something important happens culturally: people feel heard, and the AI initiative stops feeling like something being done to them. It starts feeling like something they helped design.
That shift — from AI as imposition to AI as response — is the foundation of sustainable culture change.
1.12Chapter Summary¶
Change management is not the soft side of AI adoption. It is the whole game.
The technology works. The people are capable. The question — always the question — is whether the organizational conditions exist for individual action to become collective behavior. Whether the Champions have platforms. Whether the permission is real. Whether the governance enables rather than paralyzes. Whether the leaders are modeling rather than mandating.
BankUnited has something that most organizations never get: a workforce that already operates with high professional standards, genuine client commitment, and a culture of excellence that has been building since 2009. That culture is not an obstacle to AI adoption. It is the raw material that makes AI adoption sustainable — because when professionals with those standards apply AI to their work, they do not lower their standards to match the AI’s output. They raise the AI’s output to match their standards.
The frameworks in this chapter — Kotter, ADKAR, Bridges — are not the answer. They are the map. You are the navigator. And the destination is a BankUnited where AI-empowered professionals are the standard, not the exception.
The 100 days starts whenever you decide it starts.
AI Champion A credible, curious, and publicly-learning professional who models AI adoption within their team or department, making it safe for others to experiment by sharing both successes and failures.
ADKAR A change management model (Prosci) describing the five individual conditions required for sustainable behavior change: Awareness, Desire, Knowledge, Ability, Reinforcement.
Kotter’s 8-Step Model A framework for organizational transformation identifying eight sequential conditions required for change to take hold — from creating urgency to institutionalizing the new behavior.
Bridges’ Transitions Model A change model distinguishing between external change (events) and internal transition (the psychological journey) — comprising three phases: Endings, the Neutral Zone, and New Beginnings.
AI Readiness 2×2 A team diagnostic tool mapping individuals on two axes (AI skill level and willingness to change) to identify Champions, Eager Learners, Coachable Stars, and Skeptics — each requiring different leadership responses.
Permission Gap The organizational condition in which employees want to engage with AI but have not received clear, explicit permission to experiment, fail, and learn in that domain.
Proportional Governance The principle that AI governance intensity should match the risk level of the specific use case — applying full compliance rigor to high-risk workflows while providing a fast lane for inherently low-risk AI use.
Floor Captain Model An AI adoption approach (pioneered at JPMorgan Chase) in which identified enthusiasts are available within each department to demonstrate, support, and celebrate colleagues’ AI experiments without a formal training mandate.