One year since launching a new business, results have grown 131x — but it wasn't "easy with AI"
Hello, I am ushiro, a BizDev responsible for launching new businesses at SmartBank.
As a BizDev in the AI era, I am working on 0→1 business launches.
The financial service for freelancers and sole proprietors that I launched in March 2025 has grown to 131 times its original scale in exactly one year.
*Comparison between March 2025 and March 2026
The launch coincided with the time when "AI agents" began to attract global attention, and it overlapped with a period where AI adoption spread dramatically, such as the explosive popularity of Claude Code. However, I would like to write down some lessons that are a bit different from the idea that "AI made it easy."
One year since my last note
Eight months ago, I wrote a note titled "Survival Strategy for BizDev in the AI Era." It was about how I, as a BizDev, took the lead in setting up an "AI x SaaS" stall and launched a new business without building a full-fledged product.
At that time, it was the 4th month since launch, and the business scale was 16 times that of the first month. In the 8 months since then (as of April 2026), the team has grown, the stall has become a solid building, and the scale has reached 131 times the original. It has grown about 8 times in the last 8 months.
The growth after bringing in-house systems and marketing staff, and having each section start functioning, was incredible. I realized once again that "teams are amazing" (simple thought).
The BizDev model for new businesses in the AI era: Dash → Hand off → Operate
After doing new business development for a year, I realized there is a pattern to launching in the AI era.
Dash — Dive into a new area and do the work yourself to the limit
Hand off — Once the resolution is perfected, hand it over to the team and AI
Operate — Create a systemized state that runs even without you
And then return to 1 in a different area. This is not a story divided into first and second halves; I am still doing this while changing areas. The 131x growth was the result of cycling through this process many times.
The point is that AI has two roles: AI as a tool to "hand off" and accelerate the relay to humans, and AI as a partner to take over the handed-off area and "operate" the system. If you use these without distinguishing them, your AI utilization will spin its wheels. I will write about them in order.
Dash
Screening, receivables management, KPI management, credit design, OPS improvement, partner negotiations, inquiry handling, GAS development, development issue creation, marketing design, hiring, management reporting...
During the launch phase, almost everything was on my plate. Now, most areas are handled and driven by the respective team leads.
I was the only one fully committed, but I want to emphasize that there were colleagues who supported me at key moments.
Expert knowledge, consultations, partial accompaniment—it was because of such presence that I was able to push myself to the limit.
I believe that 'resolution' accounts for 90% of launching a new business.
What are users struggling with, which operations are the bottlenecks, and what needs to be built to make the business work? You can never understand this from stories heard from others.
It only becomes visible after you touch it with your own hands, see it with your own eyes, and push to your physical limits.
Resolution reaches its peak when you have done everything yourself to the limit. Once you cross the threshold, 'what needs to be realized to make the business work' suddenly becomes clear.
How quickly can you reach that point?
Handing off — Handing off effectively to humans, using AI
When resolution peaks, you know 'what to build' and simultaneously 'what to discard'.
This was most effective in the phase of 'handing off' issues to others.
The person you are handing off to is a human. However, in the AI era, you can use AI to hand off effectively to humans. Externalizing resolution, verbalizing judgment criteria, working mocks—translating what is in my head using AI before handing it off to the team.
Handing off to engineers
During this launch, as engineers joined the team and we began full-scale in-house system development, there were almost no specification reworks or misunderstandings between Biz and Dev.
The fact that I had been running the manual operations myself, making it clear 'what to build and what to discard,' was a major factor.
Because I ran it myself and produced results, I can say with certainty, 'This is the right answer.' Because I can say it with certainty, the person receiving the hand-off doesn't get lost.
As a result of conveying the operations I touched every day and my gut feelings—such as 'if this breaks, the business stops,' 'this is still fine to do manually,' and 'this won't withstand the scale next month'—the development team built the system at maximum speed.
As a major premise, I want to emphasize that the team's engineers and the company's infrastructure team are extremely talented.
It is no small feat that they accompanied me during the initial operation phase, where we were forcing things to work by connecting spreadsheets and external SaaS, kept conceptualizing 'what we should build in the future,' and built the system at top speed from a 'street stall' situation.
It was a year where I, coming from an entrepreneurial background, thought, 'It's nice to be in a company where you can work with talented people while getting paid.'
Handing off to marketers
From the beginning, we built a system to visualize not just new acquisitions, but how they lead to profit. We built a custom mechanism using GAS to send data to the Conversion API to pass granular values to ad platforms, and used vibe coding to set up tools for KPI measurement.
Although it was a handmade spreadsheet, by laying the measurement foundation first, the data was ready when the marketer joined.
While it was a high-difficulty situation for the marketer to 'join a new business mid-stream as a new employee,' I believe I was able to hand it over in a state that was at least easy to work with.
— tsukky/つっきー_株式会社スマートバンク (@tsukky739) March 12, 2026
Currently, we have a system where the marketer takes responsibility for everything from initial customer acquisition to CRM, repeat promotion, and LTV.
In my team, we are expanding the marketer's scope of influence by saying, 'If you want to increase the ad budget, let's go increase LTV! Let's also improve the products and OPS necessary for that!'
I am of the opinion that since marketers are the first to come into contact with users, the business will grow even more if there is a structure and power that allows them to be involved in the latter half of the business funnel.
AI absorbs the shock of the handoff
Some who have read this far might think, 'In the end, isn't this just a story about humans working incredibly hard?' That is half right and half wrong.
It is truly difficult to accurately convey to others the 'if you do this, it will work' logic that is in your head.
Even if you document it, things get missed, and even if you explain it verbally, the context is lost. Previously, it took a huge amount of time to bridge the gap in resolution between yourself and the other person.
Now, AI absorbs this for us.
Structuring operational procedures, verbalizing decision criteria, and instantly creating mocks that can be understood by running them, not just by looking at specifications. I externalize what is in my head through dialogue with AI and convert it into a form that can be handed over to the team.
This process has become overwhelmingly faster and easier.
Because AI absorbs the shock of the handoff, the long-standing hurdle for BizDevs of 'having high resolution but being unable to convey it' has quietly been lowered.
Before handing it over, you must be ahead of the curve
The most important thing when 'handing over' is 'to be ahead of the curve yourself before handing it over.'You must have a resolution that allows you to speak with confidence and facts, even to excellent members. If you hand it over without that, you will face endless rework and misunderstandings.
Even if AI helps with the 'handoff,' if you hand it over while the BizDev's resolution is vague, the AI will just be translating an empty bucket.
Only when you have exceptional resolution will AI accelerate the communication.
In particular, since startup BizDevs have nothing to offer but customer resolution, they tend to 'resolution-mount' by forcing other professions to increase their resolution (personal opinion).
However, I feel that with the evolution of AI, it has become easier to convey things smoothly to others.
Running — AI Takes Over Routine Operations
Once the "handover" is complete, a state arrives where the domain continues to run even without you being there. This is the "running" phase. The protagonist here is AI.
In the world before AI, the team receiving the results from the BizDevs who were running at full speed to launch a new business had to work just as hard to keep up. This is because high-resolution requirements are also requirements with a high execution load.
But now it's different. AI handles part of the operations, boosts development productivity, and automates routine tasks. Because AI absorbs the shock, humans can meet high demands while maintaining a sustainable way of working.
"Running" does not mean not being involved yourself. You look at the numbers, detect deviations, and make small course corrections if necessary. However, you do not perform the daily execution yourself. If this line becomes blurred, you end up in a state of "thinking you've handed it over, but not actually having done so."
What AI is Actually Running
Specifically, AI is taking over these areas.
-
Standardizing Operational Judgment: While brainstorming with AI, I verbalized the thought processes for the screening and customer support tasks I handled daily, and turned them into documents that allow the team to act without hesitation using the same criteria.
Currently, thanks to the efforts of the development team, we have a system where AI makes judgments based on these criteria and automatically executes tasks.
Automation of Measurement and Reporting: Handing over areas that "quietly keep moving every day," such as daily KPIs and web advertising reports, to AI.
-
Product Specification Review and Development: Devin stands at the center of the team, handling everything from brainstorming product issues with members to creating specifications and implementing development.
AI is running a system where collaboration between non-engineers and engineers has become smooth, and everyone can face business challenges.
The engineering team set up the system and environment where Biz and Dev can smoothly coordinate with Devin to face business challenges. This was a very successful initiative, so I want to write a note from the Biz perspective.
Because AI takes over these routine operations, each team leader can focus on the "next move" in their area, and BizDevs can focus on the "next place to break" in the business as a whole.
A State That Can Be Called "Running"
I can confidently say it is "running" when it reaches a state like this.
Even if I am away for a week, the numbers do not collapse
Leaders are thinking about and acting on the "next move" for their own areas themselves
AI calmly executes routine judgments, and only exceptions are escalated to humans
My role changes from "approval/confirmation" to "observation"
Once you reach this point, you fade out of that area and head to the next frontier.
When 'running' ends, you go back to 'sprinting'.
The essence of the model is not to end at 3, but to go back to 1 once 3 is finished.
BizDev fades out of the area that is now running
With free hands, dive into the next area that is about to collapse
Sprint to the limit there again
Once the resolution is maximized, hand it over to the team and AI
Once it is in a running state, move on to the next
With each cycle, the business becomes more solid, and the areas BizDev can touch expand.
Every member of the team is running the 'sprint → hand over → run' cycle.
And it is not just BizDev running this cycle. The leaders of each area are also running their own more specialized 'sprint → hand over → run' cycles in their respective domains. I am just overseeing the whole thing and taking charge of the frontiers that no one else is touching.
A business moves forward not through individual sprints, but through a cycle that the entire organization runs simultaneously and repeatedly.
In my own team, we keep increasing the number of things we 'run with AI'. Therefore, no matter how much we hand over to AI, there is always more work for humans. We couldn't have reached this point without AI, but even with it, it is still a lot of hard work. The result of fully utilizing AI and the hard work of the team members was a '131x growth in business scale in one year'.
The most important thing in business launch × AI utilization is the 'order'.
In new areas, humans must dive in first. There is a decisive difference in resolution between feeling like you understand something by asking an AI and the experience gained by touching it with your own hands.
The difference in the distance run becomes the difference in the quality of judgment. That is why I touch it myself to maximize resolution before breaking it down into reproducible rules to hand over to AI. I always keep the results returned by AI in a state where humans can reverse-calculate and explain them.
If you don't follow this order, you end up with a state where 'the numbers are moving, but no one knows why'.In a Fintech business that handles money and risk, that is unacceptable.
For example, even if you leave user screening decisions for financial services to AI, if humans cannot reproduce why that conclusion was reached, you will not be able to fulfill your accountability regarding economics and compliance, and as a result, business progress will slow down.
That is precisely why, in this business, we never compromised on 'humans take on all the risks first and sprint'.
You should only leave things to AI after you have made the same judgment so many times that you can confidently say, 'I will always judge it this way every time.'
Because I followed this order, this past year wasn't about 'making things easy with AI,' but rather 'using AI to dive deeper.'
Moving from exploratory to steady-state systems
Many AI use cases in the world focus on exploratory topics like 'Look what we can do now!' That in itself is interesting.
However, to run it as a business, it is crucial to be able to integrate it into a steady-state system where 'hundreds of outputs of the same quality continue to run silently every day'.
Rather than having one special output, scaling a mechanism that continues to run quietly every day is overwhelmingly more effective for business growth.
How quickly can you integrate the methods found in the exploratory phase into a steady-state system? This 'speed of the cycle to operationalization' determines the launch speed of a new business.
I hand over the creation of steady-state operational mechanisms to AI and the managers of each area, while I go off to find the next place that might break or the next North Star.
Because AI exists, we can jump even further
After doing this for a year, I am convinced of one thing.
Launching a new business ultimately depends on the BizDev (business manager)'s commitment to 'making it win.' This hasn't changed in the AI era; if anything, it has been strengthened.
At first glance, it seems like we can do things smarter with AI and the human role has diminished.
But the reality is the opposite. The gap will widen depending on whether you stop at 'it became easy with AI' or use AI to 'push your own limits further forward'.
Because AI catches us from below, humans can jump even further. We can jump further ahead than we could ever see before.
To those who want to do BizDev for new business launches
'I want to do BizDev for new business launches'
'I want to race through as a BizDev in the AI era'
'I want to take responsibility for a business from 0 to 1'
'I want to experience not just 0 to 1, but 1 to 10 and 10 to 100'
For those who want to try 'racing through, handing over, and running' in their own business, there is opportunity at SmartBank right now.
While we have our main product, 'One Bank,' we also have multiple new businesses in motion. Each business is at a different phase, requiring a diverse range of skills and experience.
Since our main focus is on toC products, there is also a path for marketers to transition into business development.
Please feel free to reach out via X DMs or other channels!