Johannes Seemann had the idea for a financial planning tool built around someone's relationship with money—how they think, feel, and act around it—years before he had a way to build it. Then GenAI came along, and the idea finally seemed possible. But the models kept smoothing away the exact complexity he needed them to hold. 

In this conversation, IDEO's Becca Carroll talks to Johannes about taking a human-centered AI design approach, the technical challenges he's wrestling with, and why he made sure he had the right problem before he ever went looking for the technology to solve it.

Johannes is the founder of Sooner and an IDEO alum. He was IDEO's first data hire and spent two years running ethnographic research inside Wells Fargo before starting the company. Most recently, he was part of IDEO’s Startups-in-Residence program, where he shared space with the IDEO team while building and prototyping Sooner. If you're building something with AI right now (or thinking about it), you’ll find advice and insights you can apply to your work. 

Listen on Apple, Spotify, YouTube, or wherever you get your podcasts.

Article Summary

• Money is an identity issue, not a math problem

• The real barrier is trust, not expertise 

• Starting with human needs before building the technology

• From goals to intentions: designing for messy, multi-part decisions 

• Designing for how it should feel, not what AI can do

• Three technical challenges of holding complexity when building with AI 

• Put it into practice

Money is an identity issue, not a math problem

Over two years of ethnographic interviews at Wells Fargo, Johannes kept hearing the same pattern: people assumed a financial advisor wasn't for them. So in the moments that mattered most, they made major decisions alone. The gap wasn't information—most people can look up an interest rate. It was that something other than 'what's the best investment' was driving the decision, and no one was helping them work through that part.

What convinced Johannes that this wasn't a numbers problem was how much identity shaped their decisions. Someone who calls themselves frugal will tell you so unprompted and act accordingly, whether or not it's the "rational" choice. Someone living a lifestyle they can't quite afford is usually doing it for a deeper psychological reason. Sooner starts from that financial psychology perspective instead of starting with the math. 

The real barrier is trust, not expertise

People looking for financial advice want expertise, but trust plays a large part, too. Money conversations carry real vulnerability: embarrassment about past mistakes, fear of being judged, discomfort admitting what you don't understand. When someone can't get past that with an advisor, they hold back. The advisor ends up working with incomplete information, which works against a good outcome.

The clearest illustration of the trust problem is a woman Johannes interviewed called Rose. She'd paid off her house and built savings, staying out of the stock market entirely because it felt too risky. Then a friend from the gym who worked in finance recommended a stock to invest in. She put $25,000 into it and lost it all. 

It’s so relatable to trust a person you have a good relationship with, even if that is not the best choice logically. Johannes saw the potential for Sooner when he saw in late 2024 that people began to trust in AI and confide in general-purpose chatbots about money. 

Starting with human needs before building the technology

AI didn’t generate the idea for the business. Johannes spent years validating the human insight before he started building a product. What AI unlocked was feasibility and viability—the ability to deliver something at the scale financial advisors never could, 24/7, without the complexity of finding the “right” financial advisor. 

From goals to intentions: designing for messy, multi-part decisions 

Sooner's product starts with a psychographic profile—a conversational alternative to a personality test—that builds toward a set of "intentions" rather than goals. The distinction matters: a goal is binary, you hit it or you don't. An intention is closer to a direction you're moving in, one that can flex when life does.

The clearest example from the conversation: an unplanned $2,000 vet bill. A goal-based budget treats that as a failure if it comes out of money earmarked for savings. An intention—take care of your dog, stay aligned with your values—absorbs it and adjusts. 

Designing for how it should feel, not what AI can do

Johannes is deliberately not building a robo-advisor that makes financial decisions for you, even though the technology could probably get there. What mattered more to him was how the experience felt to the person using it, not what the AI was capable of automating. When his team tested a free-flowing AI conversation versus a push-to-talk format, people preferred push-to-talk, because it handed control back. The AI's job is to surface options and reasoning. The decision stays with the person.

Three technical challenges of holding complexity when building with AI 

These three problems specific to AI product design came up for Johannes as he was building out Sooner. If you're building a product or service with AI right now, you might find them relevant to your own work.

Normative contamination

A model trained on broad cultural data will project value judgments onto what people say, and those judgments get baked into a profile that shapes every future recommendation, whether or not they’re correct.

False coherence 

Models want to resolve contradictions into a tidy narrative. One user said he wanted to be his own man making his own decisions in New York, while his parents were still covering part of his rent. Both things are true. A model under pressure to be coherent will flatten one of them out, usually losing the nuance that actually matters.

Context rot 

The longer a conversation runs, the more a model's early, hard-won understanding of someone tends to fade, right when the person is trying to make a bigger decision that depends on it.

None of these are prompt-engineering problems you solve once. They're ongoing design work, and Johannes is candid that he doesn't yet know how much of the fix belongs to Sooner and how much is a problem the underlying models still need to solve.

Put it into practice

Chase the problem, not the technology. If a new tool excites you, go experiment with it. That's how you find out what it's capable of. But before you build an AI product, make sure you're solving an actual human need. It's easy to confuse "this is possible" with "this is worth building."

Design for control, not just automation. If your product makes decisions on someone's behalf, first try to understand how much control people desire. Do they want it done for them, or do they want more intervention? 

Talk to people. Johannes spent two years running in-depth interviews before he built anything. You don't need two years, but when you're unsure what to build next, talk to real users. That's where the next insight almost always comes from. 

Be willing to let go of what you've built. Sticking with a direction because you've already invested in it is what Johannes calls sunk cost fallacy. Not every new signal deserves a pivot—some caution is smart. But when the evidence says your product isn't the right solution anymore, the discipline is letting it go, not protecting it.

Explore More

Listen to more episodes

This is the first of two conversations profiling founders from IDEO's inaugural startups-in-residence program. [Link to the second episode once it's live.]

Try Sooner

Take the Sooner Money Mindset quiz or sign up to be a beta tester. Follow along on Instagram.

Take a course

IDEO U's AI x Design Thinking Certificate — Learn the core concepts of design thinking, and bring AI into your process to expand your perspective, generate more ideas, and unlock efficiencies.

IDEO U's Designing a Business course — Learn how to make a business desirable for customers, financially viable for stakeholders, and feasible to build and deliver to the world.

About the Speaker

Johannes Seemann

Founder, Sooner

LinkedIn

Johannes Seemann is the founder of Sooner, a fintech startup built on the premise that better financial decisions start with understanding yourself, not with a spreadsheet. He is an IDEO alum and was IDEO's first data hire, and has taught in the interaction design department at CCA. Before starting Sooner, he spent three years as design director at Wells Fargo, building a strategic design function and running long-form ethnographic research into how people actually relate to money. He holds a master's in economics.