Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

Your AI rollout has an identity problem

Three steps to move beyond capability.

[Photo: Getty Images]

Profile photo of Roger Wong

Head of Design · BuildOps

When I posted an essay arguing that software design is changing faster than most designers want to admit, the reaction that struck most didn’t come from a critic. It came from Julian Quayle, a creative director I worked with in my agency days, who left one line under the post: “So much for years of craft and imagination … I didn’t sign up for this.”

None of us signed up for it. There is a real grief in watching a craft you spent years building collapse into typing in a chat box. I’ve been working as a designer for 30 years, honing a craft of millimeters and pixels most people never notice. That work is still worthy, but it’s being repriced, in public and at speed. The people who’ve built careers and identities on being tastemakers and problem solvers are feeling the squeeze.

Every leader I talk to is living through some version of the same letdown. The artificial intelligence upskilling was rolled out. Tools like Claude and ChatGPT were deployed. Employees tokenmaxxed. Velocity doubled. In some shops it jumped tenfold. Dig deeper and the outcomes aren’t as significant as they appear. People use AI for the easy work. On the work that matters, they don’t.

THE FIX ISN’T TRAINING

McKinsey’s 2025 State of AI report put a figure to the gap: 88% of organizations now use AI in at least one business function, up from 78% the year before. Roughly 1% describe their deployments as mature. Many businesses have adopted AI, and almost none have seen meaningful results.

The knee-jerk explanation: This is a capability problem. People aren’t writing good prompts. The tools are confusing. They just need more training. So, the budget goes into reskilling, AI literacy, and AI consultants.

Organizations believe that once people understand the tools, the value will follow. It won’t. The distance between “we use AI” and “AI changed how we work” is a gap in identity, and in the workflow that identity unlocks.

WHAT THE LAST GREAT DESIGN DISRUPTION TAUGHT US

We have been here before. Graphic design production went from X-Acto knives, photostats, and rubber cement to the computer. In the mid-1980s, shortly after Apple launched the Macintosh, Adobe debuted PostScript. Together with Apple’s LaserWriter printer and Aldus PageMaker, desktop publishing was born.

Designers who survived were not the ones who learned the software first. They were the ones who understood that the paste-up boards had never been the job. The job was making a page work: the hierarchy, the dance between type and image, the judgment about how to direct the eye across the page. The rubber cement was the technique. The judgment was the craft. Anyone who had mistaken one for the other was out of the trade within a few years.

Vikram Sreekanti, who co-founded RunLLM, wrote recently that AI tools “attack user identity in a way that most software doesn’t.” Every earlier wave, from computer-aided design to Photoshop to Figma, asked a worker to learn a new tool. AI asks whether the skill that paid someone’s rent can be done cheaper. Desktop publishing took a decade to force that reckoning. AI is forcing it in months.

The pattern Sreekanti describes is not abstract to me. I lead design at BuildOps, where we make software for commercial contractors. Across organizations adopting these tools, some people take to them quickly, reading the tool as one more thing to be good at. Others resist for months, not because they can’t learn it, but because the tool is asking them to concede that the skill they were proudest of is now cheap.

Three things move an organization from using AI to getting something out of it. I have watched each one work and watched the absence of each one stall everything around it.

1. NAME WHAT IS BEING RENEGOTIATED

Most rollouts talk about augmentation and doing more with the same headcount, but they hide the “what” and “how.” Instead, say plainly what AI is about to make cheap and what each person will have to do differently to stay in their seat. Employees can’t work through a shift in identity that isn’t acknowledged by leadership.

2. DEFINE THE NEW ROLE BEFORE YOU DISSOLVE THE OLD ONE

Reskilling teaches the tool and assumes the new identity will assemble itself. It doesn’t. People need to see the shape of their new job, what judgment and orchestration look like inside their function, before they will let go of the job they have. Intercom is the clearest case I have seen. After going agent-first, chief technology officer Darragh Curran reported his top 5% of developers produced six times the median throughput, and were the same people spending more than a thousand dollars a month in tokens. Everyone had the same tools. The gap was between engineers who rebuilt their job around directing the machine and those still doing the old job slightly faster.

3. STOP MEASURING ADOPTION BY USAGE

Logins, application programming interface calls, and tokens per employee per week measure outputs but reveal nothing about whether the work is any faster or better. The real test is whether your folks are providing more value to your customers than they were a year ago. Companies that have made AI work abandoned the vanity metrics and started rewarding people for improving customer outcomes.

None of this is an argument for moving slower. Competitive pressure is real, and so is the productivity case. But the conviction that one more training program will finally unlock AI adoption is exactly what stands between most companies and the returns they promised their board.

Graphic designers who flourished through the last industry shift survived because they reinvented themselves first, then rebuilt how they worked. Some did it alone. Most needed a leader who understood what was really being asked of them. The 1% who call their AI deployments mature have done the same work. Everyone else is running one more course and waiting for adoption to catch up—and it won’t.

Roger Wong is head of design for BuildOps.