Most ideas don’t fail.
They drift.

If an AI gives you the right answer, how do you prove it?

The problem

Hospitals and universities weren’t rejecting AI. They were refusing to approve what they couldn’t defend. No one could show where an answer came from, so pilots stalled: 95% of them never reached production. Bast was seven people selling trust to the most skeptical buyers there are.

The call

Don’t staff the gaps. Close them. Kill the services business and go product-led. Design for the person who bears the risk, not just the person who uses the app. Product decisions are made, then built, in the same repo as the engineers.

What changed

The product, end-to-end. Positioning, the admin platform, a design system, and the front end that carried them. 60 changes shipped in a month, we had planned 17. An iOS app in eight days.

The outcome

Bast made trust operational. Customers can now control the knowledge, test the behavior, approve a deployment, inspect the evidence behind an answer, and see where the system falls short. By spring 2026, it was live at Maryville University and Northwestern Medicine/Lurie Cancer Center, with completed pilots at Craig Hospital and MSU Denver.

When AI shows up in your software, how do you know?

The problem

IBM had 200 software products and one new mandate: put generative AI into them by THINK. Our audit found teams inventing their own answers in parallel. Worse, designers sometimes couldn’t identify where AI was active in their own products. If we couldn’t tell, our users didn’t stand a chance.

The call

Don’t give teams guidelines. Give them working parts. One shared language for identifying AI, explaining what it’s doing, and keeping the human in control, built directly into Carbon so the answer traveled with the software.

What changed

We built the intelligent layer, AI Label, explainability patterns, and shared chat foundation, then developed them openly with early-adopter teams. Their edge cases came back into the system as reusable patterns and components. Instead of 40 teams inventing 40 answers, each new implementation improved the shared answer.

The outcome

Carbon for AI went from no shared standard to the THINK main stage in eight months. It spread to 40+ product teams, was open-sourced for anyone to use, and won three major design awards. AI transparency stopped being guidance teams had to interpret and became behavior their software inherited.

How do you build a design discipline for something nobody knows how to design yet?

The problem

In 2016, “cognitive” was on every IBM roadmap, and nothing anywhere said what these systems owed the people using them. There was no shared language. The methods were built for software that behaved the same way twice, and the ethics conversation lived in a few people’s heads instead of product teams.

The call

Define what it means to design for AI. Full stop. No Watson, no specific technology. I gave up a 19-person team to become a team of one, and built the discipline in layers for the population that would carry it: principles, ethics, people, craft.

What changed

IBM’s first practical AI ethics guide, published free and still cited. A guild that grew from two dozen volunteers past 250. A method that onboards watsonx teams, and an AI-specific design language, Carbon for AI, standing on all of it.

The outcome

When the generative wave hit, the discipline was already standing. The skills ladder made AI designer a role IBM hires for. The material still teaches people who will never know my name.

How do you scale a new way of thinking without it turning into a process?

The problem

IBM Design Thinking 2.0 had been embraced by thousands, but the content was too dense and still felt built for software product teams. Static PDFs and training workshops could never meet 380,000 IBMers where they were.

The call

Rebuild it so it could be used by everyone, every day, everywhere. We rewrote the method in plain language, opened it to work far beyond software, and replaced the PDFs with a public digital experience built around the Loop.

What changed

IBM Design Thinking moved beyond theory and training sessions and into the day-to-day work of teams. A decade later, the Loop is still at the center of the practice.

The outcome

Teams using the practice cut initial design time by 75%, development and testing by 33%, and design defects in half. Products reached market twice as fast. Design Thinking had become more than a way to work. It changed what IBM could deliver.

How do you make 1,000+ designers consistent without taking away their choices?

The problem

OneUI had made IBM’s products consistent through reusable components and prescribed patterns. It worked technically, but the experiences were generic, inflexible, and disconnected from the IBM brand. Designers were designing around the system instead of through it.

The call

Describe what makes something feel like IBM, then give designers enough guidance to make their own decisions. A language instead of a library of answers. That’s essentially the original brief, which called for products that could “look, sound, think and perform like IBM” without patterns and templates.

What changed

Consistency became an outcome instead of a constraint. Designers could make very different products while drawing from the same principles, typography, iconography, color, and expression. Even the earliest adopters showed how far a small amount of shared guidance could go.

The outcome

A language ages better than a library. A decade later, it still underpins how IBM looks and behaves, from Plex to Carbon to the physical spaces the company builds. The descendants are open source, so the outcome no longer belongs only to IBM.

How do you design a workspace when the thing you’re really trying to change is the company?

The problem

IBM was hiring hundreds of designers, but the real estate playbook still called for cubicles. In our temporary space, I watched designers dismantle the furniture and rebuild their surroundings around whatever they were working on.

The call

Treat the studio like something we were designing for the designers themselves. Everything needed to be configurable. If it couldn’t roll, two people had to be able to lift it. The space needed to change as quickly as the work did.

What changed

We built 50,000 square feet around movement, collision, and reconfiguration. Furniture rolled. Whiteboards became sliding walls. Teams could open, close, combine, and remake spaces around the problem in front of them.

The outcome

It became home to 350+ designers and countless internal and client Enterprise Design Thinking sessions. Google, Apple, and others came to study the fluid spaces for their own workplaces. The environment became part of how the work happened, instead of simply where it happened.

How do you teach a clinician to ventilate a child when the expert is half a world away?

The problem

Some of the most important pediatric critical-care knowledge lived with the people who had learned and developed it through years of practice. If you weren’t in the same hospital, on the same shift, or even in the same country, getting access to that expertise could be difficult or impossible.

The call

Don’t turn that expertise into another pile of courseware. Let clinicians teach the way they actually teach: by showing, explaining, and coaching. Then make that experience available anywhere, anytime.

What changed

OPENPediatrics recreated the apprenticeship model online. Video, guided learning, expert access, and collaboration let physicians and caregivers learn directly from Boston Children’s Hospital specialists and from each other, regardless of time or location.

The outcome

Within six months, more than 1,000 caregivers in 20 countries were using it, and techniques delivered through the platform were credited with saving multiple lives. By 2016, it had reached thousands of clinicians at 343 hospitals in 74 countries.

How do you get better ideas by giving people less time?

The problem

IBM iX needed to help clients find market-defining ideas faster. Too often, the conversation started with features and technology instead of the people being served and what might actually be possible.

The call

Compress the engagement to eight weeks. Put IBM and the client on one small team, understand the problem together, explore the art of the possible, and start building immediately. Keep what works. Throw out what doesn’t.

What changed

The vision stopped being something we documented at the end. We made it in real time. Clients left with an interactive “concept-car” they could experience, a Vision Blueprint, and a roadmap for turning it into something real.

The outcome

IBM iX has sold hundreds of Accelerated Visioning engagements. The method has been in the market for more than a decade and has changed the way the business works with clients.