Genspark × Bloom: The Closer

We held our second Bloom event with Genspark, and this time the theme was sales. I was really curious how people in sales actually use AI in their day-to-day work. So this time we invited Yong-gi Kim, known as "the 43 Trillion Won Man," who came out of Shipley Korea, the Korean arm of a global bid consulting firm, for a condensed sales lecture. After that, people used Genspark Slides to build a deck in 20 minutes, pitched their own products, and competed for prize credits.

What stuck with me most was something he said. People decide the final ranking in their heads before they even look at the scoresheet, and then fill in the scores to match. In other words, people pick the side they're already drawn to before they look at the details. So, paradoxically, slides are only a visual aid, and you shouldn't over-focus on them. That message hit me hard. What matters isn't the slides. It's connecting with the audience more than 90% of the time and winning them over.

The person who raised this to an art form is Steve Jobs. Since him, almost every tech company's presentations have followed his style of storytelling and comparison. That's less a trend than an evolution in how presentations are done.

So sales doesn't come down to making a good deck and speaking fluently. Three days later people remember only 10% of text and 20% of images, but when text and images are combined, they work in "Dual Mode" and about 68% sticks. How to use a well-built story so the data is still running through someone's head as they fall asleep, how to plant the message that matters more to the other side even when you're objectively the underdog and turn it into a win: in the end, that's the territory of messages that AI and humans make together.

Slide proposals and presentations are an essential part of sales. Thanks to FuriosaAI, who once again gave us a great space, and Genspark, who sponsored generous credits, we were able to put together a meaningful event for the AI community. Thank you, sincerely.

Sell to the people in this room in three minutes

THE CLOSER, hosted by Bloom, Genspark, and FuriosaAI, took place on Tuesday, October 27, at 7 p.m., at FuriosaAI on Dosan-daero in Seoul. Most of the crowd were in sales and business development, and they'd all come straight from work. Junsu Kim from the Bloom team hosted and ran the event, and I helped out in the background. There was one mission: sell your product to the people in this room in three minutes. If you didn't have a product to sell, you could pick an imaginary one, or try selling the Cry Cheese Burger we served that night, or Genspark.

The flow was a 30-minute talk from the speaker to learn sales basics, then 20 minutes to build a deck of up to five slides in Genspark AI Slides. AI would screen the submitted deck links first and send five people to the final. In the final, the five would each pitch for three minutes, and instead of judges, everyone in the audience would vote BUY or PASS. Whoever got the most BUYs would be the night's CLOSER. The preliminary round was judged on the deck and the final on speaking, so you couldn't win with only a good deck or only a good mouth.

Junsu said the decks made that night would be published online like open source. Developers build better careers by open-sourcing what they make, and now non-developers need to put their work out there and make themselves known too. Anyone who didn't want that could opt out of having their deck published and of being photographed, and people from the same company were asked to sit apart if they could.

Someone from FuriosaAI's business planning and product side, who had given us the space, introduced the company. FuriosaAI designs chips that handle AI inference in data centers, and this year it started mass-producing its second-generation chip, RNGD. The card-shaped accelerators displayed in the back go into servers supplied to Korean AI data centers and internet company data centers, running various open-source models, and FuriosaAI also directly runs services for Korean foundation model developers like Upstage. They said the space was built for outside community events, and if anyone was curious about the product, it would be right there in the back during networking.

Thirty minutes with the 43 Trillion Won Man

The speaker, Yong-gi Kim, CEO of Brave Sales Labs, spent 15 years as head of the Korean branch of Shipley, the world's largest bid consulting firm, founded Shipley Korea, and was the first to bring bid consulting to Korea. Junsu said they had heard his lecture before and loved it, so they asked him, and he gladly said yes.

His first job was car salesman and his second was brokerage salesman. For about 30 years after that, he sold invisible things to companies, like consulting and training. He said that coming to an event like this felt like coming to see close family, with a sense of fellowship, and also like coming to see some truly pitiable people. The audience laughed. Then he added that startup sales is probably talking endlessly, ending the day without a sale, and having to forget it and start over the next morning.

He's known as "the 43 Trillion Won Man" (about $30 billion), but he made clear up front that he didn't sell all 43 trillion won himself. It's the total of bids he was involved in as PM, PD, or director on big projects. In large-scale sales you don't say you made a sale, you say you won a contract, and the opposite is a lost bid. In English, he explained, it's just Win and Loss. The experience that helped him most in 30 years in the field was about ten years as an evaluator for the Public Procurement Service, Korea's government procurement agency, because the materials he looked at while making sales materials and proposals were completely different from what he saw sitting in the evaluator's chair.

He said he'd picked only eight points to squeeze a three-day course into 30 minutes, and started with what he used to think every time he left a procurement evaluation. Looking back at who he'd given the most points to, the answer was always the same. Everyone followed the same table of contents and said similar things, so nothing stood out, and the points went to whoever made him think, "Ah, I get what that person is saying." Whether you're pitching alone at a client or competing in a bid against several companies, a sales pitch always happens in competition, and his premise was that in competition the memorable pitch wins.

Slides are only a visual aid

So the most important skill on stage, in his view, was not letting the slides become the star. He admitted it was a paradoxical message to give at an event where people came to build decks. He pointed out that the original name for a slide deck is a visual aid, but in the field the presenter often becomes the slides' assistant, reading the deck aloud.

The flow of the slide deck isn't what matters. The flow in my head is what matters. That's how I become the star.

The basics are to glance and confirm the slide you want is up (touch), turn toward the audience (turn), and spend 90 to 95% of the time looking at the audience while you speak (talk). Engineers and experts in the field do almost the opposite. With their backs to the room, they glance around briefly to check if people are listening, then turn back and read the slides line by line, spending 70 to 90% of their time with their backs to the very customer evaluating them. He asked if we knew what the hardest part of evaluating for the procurement agency was. His answer was sleepiness. When three or four presenters with their backs to you go 30 minutes each for two or three hours, after a few dozen minutes you're too sleepy to listen.

That's why he believes restoring the relationship with the audience is the most important thing in a presentation. A good proposal is one you can fully understand without the presenter explaining it. But with presentation slides, the moment the audience fully understands them, they have no reason to listen to the presenter. In front of a presenter reading proposal-style slides, evaluators all keep their heads down, busy scoring. Only when they meet a presenter who is the star do they make eye contact and want to hear what's being said.

He said that when evaluators have their heads buried in scoring, he turns the slides off completely. When the screen goes dark, people have no choice but to look at the presenter, and if you start your story then, they listen closely.

Restore the relationship. Never hand control to the slides.

People who write the ranking before the scores

What he called a surprising secret was the scoresheet. Big government projects come with an RFP, and the RFP comes with a scoresheet. You'd think evaluators score Company A and Company B item by item and add them up, but many actually write the total at the bottom first, deciding who's first and who's second. He told us to watch how we judged when we pressed BUY and PASS later.

He calls those who score item by item analytical evaluators, and those who write the total first holistic evaluators. Analytical evaluators are practitioners, specialists, engineers, and juniors, good, diligent people who check each thing one by one. Holistic evaluators are people with big decision-making power and lots of experience, like large-company CEOs or senior professors who've been doing procurement evaluations for 30 years. In his words, "people who can't be analyzed."

The problem is that those of us who make proposals and presentation materials are all analytical people, so we miss what holistic evaluators look at. What he saw in the field was that holistic evaluators look at the big picture and the story: what strategy this organization will use to deliver the solution. Even if you present alone to a client today, another company comes tomorrow and another the day after, so the essence of a sales pitch is competition, and to win, what you said has to be remembered first.

Story sells, data tells

To explain how to be remembered, he brought up "Dokdo Is Our Land," a well-known Korean song about the disputed islets. Many people would know the distance between Dokdo and Ulleungdo, he said, and pulled 87 km from the lyrics. Some would know the address too, he said, and as he moved on to the address he joked that one person in the audience must be getting on in years, which got a laugh. People in good shape even know the latitude and longitude, he said, and that's no accident.

When you sing, whether you're angry, sad, or fired up, your right brain is working, and information that comes in through the right brain stays much more strongly than information stacked straight into the left brain. In an experiment by the educational psychologist Bruner that he cited, children remembered the same data 18 times longer and 18 times more strongly when they sang it than when they simply memorized it. That's why industries that persuade the public, like advertising, presentations, and marketing, are obsessed with storytelling and study it.

He named Steve Jobs as the person who brought this into industry. On the stage where he unveiled the MacBook Air, Jobs compared its thickness with the Sony TZ series, the thinnest laptop at the time, showed again that the thickest part of the MacBook Air was thinner than the thinnest part of the Sony, and then pulled the laptop out of an ordinary office envelope. After we all watched the clip together, he said that before Jobs, even Samsung Electronics would have talked mostly about specs, meaning data, when launching a laptop, but now Samsung and LG launch products differently. Pant legs getting wider and narrower is fashion, he said, but this kind of progress, once it happens, doesn't go backward.

For those wondering if it works in B2B and not only B2C, he shared a case he consulted on. The client was the defense company LIG Nex1, which at the time only made guided missiles and had never built a radar. The rival was a joint venture between Samsung and a French defense company said to make the best radars in the world. It was a fight between an absolute underdog and a team it had no business beating.

So the script he wrote with the client said this project was to develop an advanced radar called AESA, only the fourth country in the world to do so. A homegrown company had to develop it so the technology wouldn't leak abroad, and that was the only way to export the radar to the world. Even with its shortcomings, since this was an early R&D project, the homegrown company LIG Nex1 deserved a chance. While that was being said, the slide showed a Korean flag waving, as video, not a photo. As a consultant he couldn't go into the presentation room, but he heard afterward that the evaluators all rewrote their scores at that moment, and it ended in a win.

Story sells, data tells. Data is what informs, and story is what sells.

Customers buy with the heart and explain with logic.

One success story beats 100 references

Among stories, he stressed success stories most. What a client really wants to know is how good your solution is and how good your team is, but there's no way to ask that directly, so they ask for references. Asking where and how many times you've done it is a quantified approach that infers from numbers. Telling a success story is a qualified approach that shows how well you did it. From the client's side, he said, they don't really care how many times you've done it.

To show how much the customer's point of view matters, he had us try the E test from psychology on the spot. You face your partner and each draw a capital E on your own forehead with your finger. He joked that if you fail, you should look for another line of work, and people laughed before it even started. The room split between people who drew the E so it was readable to themselves and people who drew it so the person facing them could read it. He got another laugh by saying quite a few people here should be changing jobs.

He found the reason visuals matter in presentation materials in imagination. When kids read a lot, they imagine, guess, and predict, and that develops their brains. Visuals stop that kind of imagining and guessing, and that's exactly what you need if you don't want the customer guessing on their own. The share people remember after three days, what he called the retention rate, is around 10% for something read in a book and around 20% for visuals, but Dual Mode, combining visuals with text, goes up to 68%.

He put up a page from an actual proposal that made his company's name. Under the image was a sentence saying that the Hyundai Rotem wheeled armored vehicle project, where he served as PM leading everything from bid strategy to the proposal presentation, led to export wins for derivative models, with revenue topping 1.3 trillion won and more than 4 trillion won expected. Korean proposals usually leave out a sentence like that because it was already explained above. But under the law of reading gravity, where every eye flows top to bottom and left to right, a busy evaluator will never look at the image and then go back up to read the text, he explained.

When he ran a business unit as vice president, he raised its win rate from 70% to 90% in one year, even though it was already number one in the industry. As revenue went from about 70 billion won to about 90 billion won, net profit jumped from 5 billion to 25 billion won. At a company where anything over 300 million won was a big project, all he changed was having people write a one-page strategy for every project over 300 million won, having them rehearse the presentation in front of him and coaching them, and once a year building one really solid proposal for them and having them reuse it more than 300 times. It cost very little. So many people worry they have no references, but with one well-made success story, you don't need to envy anyone.

One well-made success story beats 100 references.

Delete the rest, on purpose if you have to

What frustrates him most is when a good solution dies because someone dutifully followed the usual table of contents. The moment you pour it into the same format as everyone else, it's dead, he said. He put up the familiar five-slide structure, Problem, Solution, Proof, Commercial, CTA, and said that without today's lecture, everyone would have written it that way. Then he flipped through a deck he'd made in 10 minutes with an AI model and pointed out the slides he deliberately left unsaid. The bragging and the context, the presenter can say out loud. He'd keep only two slides, Proof and one showing through a case that our solution works, and delete the rest, on purpose if he had to.

The flow of a proposal is logical, so it's called an outline. The flow of a presentation is emotional, so it's called a storyline. That storyline should live in the presenter, not in the sales deck. The presentation materials should hold only the key points you want to stress, so the deck supports the presenter instead of leading. For competitive pitches, where you win by being remembered, his conditions were: never hand control to the slides, and win on strategy and storytelling. His last words were that today it was enough to share the problem rather than specific methods.

Junsu then asked everyone to judge whether they remembered Junsu's PowerPoint more or Junsu more, and got a laugh by promising to explain the credits while looking at the slides as little as possible. The reason for adding a hands-on session was that if you only listen to a lecture and go home, you forget everything, and it sticks only when you build something yourself from what you heard. Participants got Genspark credits through a QR code and built decks for 20 minutes until 8:05 p.m., then submitted them via share links anyone could open. Junsu said that if the story is well set, the PPT comes together quickly, so no need to rush. Piano music played, and while AI judged the decks we left the time open for networking.

BUY or PASS

Five people made the final. After each pitch, the audience had 30 seconds to choose BUY or PASS on a voting form Junsu built personally, and the results would be revealed after all the pitches. Junsu said five people competing alone wouldn't be fun, so we added Koreans' favorite game, guessing who'll win. Guess the winner and you get extra credits. The audience let out a "wow," though with one pick out of five, the odds were pretty good.

The final opened with KakaoTalk screenshots received by a six-person food distributor in northern Gyeonggi Province (the province around Seoul). It has 170 restaurant accounts, and orders start coming in over KakaoTalk at 7 p.m. Manager Kim at this company retypes them into the system 150 times a day, until 1 a.m. Over a year, that's an employee's salary vanishing into dictation. The CEO tried installing an ordering app, but the restaurants wouldn't use it, so they had to take app orders and KakaoTalk orders separately and the work only doubled. So instead of trying to change the restaurants, they decided to change only the distributor's side.

It started from the idea of taking last night's 20 orders, which took Manager Kim three hours to enter, and finishing them in 0.01 seconds. In practice, of 65 lines in 20 KakaoTalk orders from 8 accounts, 59 went into the system as is, and not one line went in wrong. The remaining six lines, where it wasn't sure whether it meant 2 radishes or 2 bunches, weren't entered, and the person in charge was asked to check. In an industry where one wrong line starts returns and payment fights the next day, they chose not being wrong over being fast. The US and Europe already have two paid services with recognition accuracy up to 97%, but Korea, where KakaoTalk dominates ordering, has no such tool, and big companies have no interest in visiting small distributors one by one to clean up their product lists. That was the gap the presenter saw.

The offer was to run it free for two weeks, charge nothing if it didn't clear 85% automatic processing and 30% account conversion, and charge around 990,000 won a month if it did. They said they'd continue validating with that northern Gyeonggi company in November, and closed by saying that the day Manager Kim gets to leave at 10 p.m., that CEO will be their first success story. They'd taken the success story lecture they'd just heard and used it as is.

The presenter who brought Claude Enterprise worked in sales at the Korean subsidiary of a Japanese IT company. The company resells AWS infrastructure and, as an Anthropic reseller, helps companies adopt Claude Enterprise. It started B2B sales in Korea this June, and contracts for more than 600 seats are coming through, mostly with large and listed companies. They handle billing in won instead of the dollar-denominated charges and issue tax invoices, explain the strengths that come only with the Enterprise plan, like audit logs, despite the higher price, and consult on setup for companies that want to plug multiple LLMs into their backend.

They also pointed out that they make a custom quote for each client to fit Korea, where expense processing matters. Clients include asset managers and semiconductor, manufacturing, and biotech companies. They filled their three minutes by saying anyone looking into the Enterprise plan could reach out anytime for a consultation based on their guide documents.

The presenter who built a real-time AI interpretation and captioning service started with why they have to be good at English.

I went to college in the US. A lot of money went into me. So I have to be good at English.

At their first job, at an American company, they kept getting stuck interpreting because they spoke English. But speaking English and interpreting well are different skills, so they wondered if AI could do it and started building it themselves. Last month's revenue was about 2 million won. A global security company and its distributor use it, and a company in the crypto industry used it at a 300-person international seminar.

Human simultaneous interpretation requires two interpreters taking 15-minute turns because of fatigue, so it costs 300,000 won an hour, 2.4 million won for eight hours, and with a 1-million-won booth and receiver rentals it comes close to 5 million won. A quote for an AI solution came to about 4 million won for eight hours. They said that with one laptop and fast Wi-Fi, they'd do it for 150,000 won an hour. Put a QR code up for the audience and 300 or 500 people can each read captions on their phones, and you can show the slides and captions on one screen. On a four-person team, two developers, a marketer, and themselves, they're the one who built it, but their main job is sales, which is why they were up there. They stepped down asking for lots of inquiries at 150,000 won an hour.

A presenter who said they don't work in sales came up with a brand-new product: event partnerships with Cry Cheese Burger. They opened with a scene: you're meeting a friend on the weekend and want to go to a cafe, but not Starbucks, somewhere special, so you stand in the cold searching for minutes, and if it were a date, you'd even end up fighting because you didn't plan ahead. Their read was that people want to consume experiences and hate the ordinary. If you're planning an event, you might look for the most expensive fine dining your budget allows, but there's a better option.

At a department store, the food floor is packed but the higher floors are empty. In a downturn, small luxury is in, buying the most special experience you can afford, and Cry Cheese Burger can own that middle ground, not too ordinary and not out of reach. That was the logic. It's more price-competitive than pricey brands, it has run so many B2B events that it can handle 1,000 guests without things falling apart, and it already planted the image of being the burger from the Genspark workshop here. After laying that out, they closed by asking that if you're planning a special event this quarter, give the person in charge just 30 minutes for a casual call.

THE CLOSER

Soohwan Yoon took the stage saying they'd thrown it together and didn't think it would get through, but it did. They started from the point that these days, when American companies expand into Asia, the first country isn't Singapore or China. It's Korea. Most of these are AI products, and impatient Korean users adopt them quickly and heavily. To enter Korea, you need a country manager, a recruiter, and a team for that recruiter to hire. But American recruiters don't know what "SeoSeongHan JungGyeongWeSi" means (shorthand for a tier of Seoul universities), don't know which companies besides Samsung and SK to hire from, and people from Google or Meta don't bother joining an American startup that just arrived in Korea.

They have twice been country manager in Korea for American unicorns. Now they meet people in person, screen them, and when someone is truly great, persuade them to come on board. They pick people to the standard the US headquarters wants, so they personally guarantee the quality. Of the roughly 10,000 outside recruiters in Korea, they have partnerships with 400 and can send a search to all of them at once, so the partners guarantee the speed. A month after starting, they signed Rakuten and are in talks with a legal AI company and a collaboration tool company. They closed their three minutes the way the event's name demanded: sign today and within a week they'd connect you with three people who perfectly fit the role, and sign today and you get a better price.

After five rounds of voting, the person with the most BUYs was Soohwan Yoon. Junsu said there was no grand reveal prepared and just shared the screen to show the results, announced Soohwan Yoon as the night's closer, and said those who guessed the winner would get extra credits. Asked how they made the deck so pretty so fast, Soohwan said Genspark Slides has a Creative version separate from the regular one. They picked a template there and kept it to about five slides, had it read the homepage they'd built to pull in the design system, and fed in text written to match the lecture, and it came together fast. They wrapped up with "Genspark, as expected," which got another laugh.

Junsu split the night into a lecture, deck building, and pitching because, as someone who goes out and sells too, Junsu believes a sale happens only when the deck is solid and the speaking is strong as well.

A pretty PPT alone doesn't make the sale, and neither does just being a good talker.

Originally written in Korean on Brunch, October 30, 2026.