Living Between Two Worlds

Part 1 of a series on the moral dilemmas of learning AI as a designer

The conference room in Overland Park, Kansas had the kind of fluorescent lighting that makes every conversation feel like a performance review. The rest of the team had gone on break, which left two of us: me, a consultant on the team redesigning how Sprint served its customers, and a project manager whose buy-in the whole program needed.

We hadn't gotten along. Consultants walk into a building carrying a reputation they never earned personally, and I inherited mine at the door. Sprint's culture had a favorite phrase for change, "Who moved my cheese?", borrowed from the business fable everyone kept on a shelf somewhere. People said it with a laugh. Nobody laughed when someone actually reached for their cheese.

She was stressed, and she didn't hide it well. A merger with T-Mobile hung over the building, and the platform we were designing would make several departments far easier to manage on their own. For someone worried about her job, our work offered the opposite of comfort. Most subject matter experts never name that fear so plainly. She came close to saying it out loud.

So I asked her: "What would an outcome look like that would ensure you didn't feel at risk in this transition?"

She didn't have an answer.

Something shifted anyway. After that conversation, she changed who she was willing to introduce me to and how she introduced my team to others, and she started going to bat for our design decisions without anyone asking her to. I'll come back to why she couldn't answer. First, the seat I was sitting in.

Between the worlds

I knew about the merger. As a consultant, I lived between two worlds for the entire engagement: leadership's plan on one side, and on the other, four teams who had spent years perfecting how they served customers. Telesales, support, in-store, and web each ran on their own tools and their own instincts. Telesales alone juggled more than twelve of those tools, and agents had to learn the tools, the devices, and the scripts before they could help a single customer.

Those teams were experts, and experts have earned the right to resist change. They had built workarounds that worked. Then a consultant showed up to tell them it could all be simpler, in the shadow of a merger some of their leaders knew about and they didn't.

Call it merger paralysis. Fear of irrelevance slowed everything down, and it worked the same way at both scales. Inside each team, people protected workflows they had spent years perfecting. Across the business, the company braced for a merger that could rewrite everything it had built. Nobody at either level wanted change, because change meant an unknown impact on what they had worked so hard to achieve. Any change that made a team's work too easy read as a threat to the team itself, and threatened people stop cooperating with the designers studying them.

Every designer who has worked on enterprise tooling has sat in this seat. You study someone's job closely enough to improve it, and improving it changes the job. Sometimes it changes the headcount. The merger only made the stakes legible. (The full project lives in my Sprint case study.)

What designers do in that room

Earning trust from people whose jobs your work will change takes more than good intentions. Over the years I've leaned on a handful of moves, and Sprint tested all of them.

Show the potential in their own day. We sat one of the telesales team leads in front of the updated experience with no training and asked him to work through it as if he were on a live call. He explored, narrated what he'd do next, and talked through how he'd use each piece. By the end he had watched his own workflow get simpler and a pile of problems from his scattered tools disappear. He went straight to his team and became the loudest champion the work had. No deck could have done that.

Give real ownership. We ran service blueprint workshops that put every team's voice in the room. People defend what they helped build, and that alignment sped up delivery right when the organization needed speed most.

Credit people in public. The same project manager from that conference room, along with the telesales manager, facilitated contextual inquiries and made the introductions that uncovered the opportunities we focused on. I called that out every chance I got. Commendation has nothing to do with flattery. This work tends to go thankless, so the effort to say thanks out loud reliably earns goodwill and cooperation.

Ask the human question out loud. When fear sits in the room, name it, and ask what a safe outcome would look like. The question helps people separate what they can't control from what they can.

These moves work, and every one of them borrows against leadership's motive. When leadership reinvests the gains in its people, the excitement a designer builds turns into retention. When leadership cuts, that same excitement turns into something the designer helped sell. Designers control the trust. Leadership controls what the trust turns into.

Both chairs

AI changes the math because it puts designers in both chairs: the observer taking notes in the ride-along, and the person whose work someone else studies.

At Sprint, we trained a model on the existing sales script to automate parts of the "next best action" experience. That was 2019. I designed AI into someone else's enterprise workflow years before anyone asked me to learn AI for my own. Now leadership teams everywhere ask designers to build fluency in tools that may reshape the design team itself. That's the moral dilemma of learning AI as a designer: you master the thing that might shrink the room you work in.

That dilemma has company. Design exists to improve the lives of the people who experience it, and AI's appetite for energy and water works against the planet those same people live on. The models also trained on what amounts to stolen art, much of it from the same creative community designers grew up in. Each of those deserves its own post, so consider this part one. It starts with the people in the room.

I learn it anyway, for the same reason I asked that project manager what a safe outcome looked like. Fluency earns a seat where motive gets decided. Designers who opt out hand that decision to people who never sat in the ride-along. Needless to say I have further opinions on the matter: here.

For what it's worth, my workarounds have always looked a lot like today's. I've spent a career trying to work smarter. The difference now: I build my own tools where I used to settle for simple automation of repetitive tasks.

Ownership shapes motive

The same technology produces opposite outcomes depending on who owns the saved hours.

Most companies answer to shareholders, and shareholders reward margin, so efficiency gains flow toward margin by default. Nobody needs to play the villain for that to happen. The incentive works like gravity: no one has to push.

IKEA shows what happens when an owner points the gains somewhere else. In 2021, IKEA launched an AI assistant named Billie to handle routine questions like store hours, order tracking, and returns. Billie assisted 47% of the customers who used it in its first two years, and that share has since grown to 74%. IKEA kept the people Billie freed up and spent about two years retraining roughly 8,500 call center workers into complex service roles and remote interior design advisors. Its remote sales channel grew from €1.08 billion to €1.25 billion in a year, and customer happiness climbed from 60% to 89%. (People Matters)

The structure behind that choice matters. Ingka Group, which runs most IKEA stores, belongs to a foundation with no owners of its own, and it reinvested 85% of its net profit in the business last fiscal year. No quarterly shareholders wait for the saved hours to show up as margin. The foundation structure also served tax and succession goals, so I won't pretend it's a monument to altruism. It still proves the point: change who owns the gains, and you change what the gains buy. (Ingka)

Public companies still get a vote. Gravity sets the default, and leaders can push against it. Klarna leaned on AI to shrink its customer service staff, then started hiring humans again over service quality. Skipping the choice costs more than making it.

The cost of an undeclared motive

I've been on the receiving end of two AI-attributed layoffs in two years. Neither company lacked smart people; both lacked a declared reason for the experiment. They funded unbounded AI experimentation before they had a plan for the people doing the work.

Layoffs come with a career, and I hold no grudge against the people who made those calls. What I carry forward is a lesson. When leadership keeps its motive to itself, people fill the silence with the worst version, and the fear shapes the work long before any decision lands. I watched it happen at Sprint, and I've felt it from the other chair since.

The question goes back to leaders

The project manager couldn't answer my question because the answer never belonged to her. She didn't control whether her job survived the merger. She controlled whether the program succeeded, and once she started operating from what she controlled, fear gave way to optimism about what we could accomplish together. Designers can help people find that footing. Only leaders control the job safety underneath it.

So I'll ask the question again, this time to the people who can answer it. What would an outcome look like that would ensure your people don't feel at risk in this transition?

The Sprint platform cut average handle time 45%. Every one of those saved minutes could have funded reinvestment or consolidation, and the platform didn't decide which. Neither does AI. Leadership decides who owns the saved hours, and the most useful thing a leader can do before an AI rollout is say so out loud: where the hours go, and what the people doing the work can count on.

Your designers can earn trust in the ride-along. They can't earn it on your behalf.

I don't know what happened to those four teams after the merger closed. I shipped the work and got moved to another engagement, the way consultants do, but I do know that what we built gave the company an incredibly robust view of their customers, and a much simpler path to service them.

The work gets more achievable

The Sprint program took sixteen people under a year to collapse twelve tools into one. Most teams facing a problem that tangled assume they need a consultancy's bench to touch it. AI changes that math in the direction IKEA chose. It compresses the slowest parts of enterprise tooling work: synthesizing hours of contextual inquiry, prototyping a flow a team lead can run cold, testing variations before engineering commits a sprint to them. The team you already have can take on problems that used to sit out of reach, and your experts stay in the room doing the part only they can do. The hard conversation in the fluorescent room still takes a person.

That's the work I do. I bring two decades of design leadership into deeply complex organizations, where the tooling runs deep and the stakes run human, and I help leaders use AI to make their people more capable. I'm available for fractional and advisory engagements through Design by Proxy. If that sounds like your situation, or you've lived between these two worlds yourself, connect with me on LinkedIn.