Designing with AI around the "personal touch" you deliberately keep

Yogoro: One Media is a company built around creative production and advertising agency work. We handle a wide range of production projects—planning, producing, and distributing SNS ad content centered on short-form video, casting influencers, and more—and in the creative side of things, we believe differentiation and added value are extremely important.

Until now, we hadn't really thought about using AI for the creative work itself. That's because the creative domain demands flexibility and human sensibility, and it's the very heart of how we differentiate. So precisely so we could spend our time on that differentiation, we wanted to use AI to make our other routine, repetitive work more efficient.

Yogoro: Our first step was Deep Research. When I paid for it personally and gave it a try, it had a clear effect on sales research and finding casting candidates, so we decided to roll it out company-wide. We also hold a volunteer-run AI study group once a month, where we create a chance to share things like "this kind of prompt works well" and "I tried using ChatGPT this way."

After about three months everyone had grown comfortable with using AI, but in practice it stayed within "replaceable territory" like brainstorming ideas and drafting rough outlines—we hadn't gotten to the point of building it into our workflows and dramatically changing how efficient our work was. I felt there was a limit to how far things could spread by just dropping in tools one at a time.

On top of that, talking with AI and keeping up with trends both cost time, and there were moments I doubted whether it would really lead to major efficiency gains. In particular, ChatGPT, which we mainly used, kept conversations closed off, making it hard to make them broadly useful or to share what we'd learned. Before I knew it, I'd taken on a kind of "AI point person" role, and I sometimes felt the pressure of having to produce output—which threw off the balance with my actual work.

Yogoro: More fundamentally, I reached out to Wolkin so we could build "systems" with AI embedded in them ourselves. AI evolves incredibly fast, and keeping up with best practices on your own or on the front lines is hard given the time constraints. That's exactly why I thought asking Wolkin, who is always tracking the trends, was the best move. I knew firsthand their ability to keep up with AI trends and their willingness to share it generously, so I decided to ask them.

Around the time I was consulting with them, a team member who had served as a director left due to a change in their life stage, which became an occasion to feel the risk of key-person dependency much more personally.

As a first step toward AI, I thought we'd start with the casting work of proposing influencers active on social media in line with client requirements. We had a history of organizing the flow and data for this work in-house, so I had hopes that this area could be handled without becoming dependent on specific people—and bringing this consultation to Wolkin is how it all began.

Yogoro: Over three months, Kagawa—an executive officer who leads the production team—and I took the lead in learning Cursor and n8n while implementing. In the first month we put requirements into words and designed the data; in the second month we rapidly built lightweight prototypes; and in the third month we polished it into a form that could stand up to real operation.

Wolkin was always online running alongside us, and whenever we got stuck, they solved it together with us. It was incredibly reassuring that they showed us, at a practical level, "this part can be done with AI" and "this part should be done by people." Watching Nakazono actually share his own screen and walk through the troubleshooting process with us made me feel I could put it to use myself, so being taught in a hands-on style really was a good thing.

What I felt was especially good about how this support was designed is that, while valuing the short-term goal of automating the casting work, they designed it by working backward from One Media's overall business goals and future challenges. I think it was designed with very high extensibility, so that once we could introduce this goal, it could be repurposed for other work too.

Yogoro: The biggest thing was becoming able to use Cursor. It understands context and assembles code for you, and you can accumulate knowledge and files within the project. The more you use it, the more you get this sense that your "workspace is growing," and I became able to imagine that "maybe we can build the systems out in the world ourselves too." That was a huge step forward, and it was honestly exciting.

After getting the lessons, we became able to take on, ourselves, areas that used to depend on outside help—calling a bookkeeping tool's API to streamline back-office work, or collecting social media data to build the skeleton of our own reporting. Through efforts like these, a habit took root of the leadership team spending 2-5 hours a week on development, and I feel we've grown a real feel for what matters when it comes to AI.

Yogoro: Without rushing, I want to keep stacking up lightweight connections first. Connecting data and tools across the board with BigQuery and n8n raises the real-time quality of reporting and decision-making. Down the line, we're also looking at turning things like ad reporting into our own products and delivering new value to our customers.

As One Media, we also want to value not eliminating key-person dependency entirely, but designing how we deliberately keep some of that "personal touch." By striking a balance between what we entrust to AI and what we entrust to people, I believe we can raise the density of our creative work.

And I'd love to spread the sense that "we can build it ourselves too" throughout the company. AI takes the first step, and people refine the core of the value. If we can make this cycle stick, we can raise the density of creative work while reducing key-person dependency. That's the kind of organization I want to aim for.