AWS × Cliwant × LG U+ × Bloom: Winning Proposals in the AI Era

Before I get to this Claude Bloom recap, there's something I want to say first.

Every time I prepare an event, it feels like I'm being tested. People coming for the first time have no expectations, so I have to paint something good on that blank canvas. People who've come before probably won't be satisfied unless I give them more than they imagined. A lot of people ask why I go that far. Reputation is part of it, sure. But if I'm doing it anyway, I might as well do it well. Mostly, I feel I owe more than the bare minimum to the people who drag themselves across the city after work, and to the speakers who show up for the community with basically nothing in return.

I'm also running events for the first time in my life, so the hardest parts are still time management and no-shows. However carefully I build the timetable and prep the questions in advance, something unexpected always happens on the day. The bigger the crowd, the harder it is to keep things on track. Everyone comes in rooting for us, which helps me stay less nervous. Still, to do it well, you need a healthy amount of nerves.

So even though this was our third event this week alone, I'm nowhere near used to it. Each one is a new challenge. I think that feeling of being challenged squeezes you and makes you grow at the same time. People around me say I look more relaxed as time goes on.

All good so far. But yesterday's event felt a bit messier than usual. About half the room seemed happy and half seemed let down. Honestly, it was the first time I'd felt that at one of our events.

When I think about why, it's because the speaker talks and the fireside chat ran long, so the small-group sessions, the real heart of Claude Bloom, started more than an hour late. By then everyone was already worn out. People always enjoy the small groups, but some left before they got to them, and I felt bad and sorry about that. The room was packed and the mood was great at the start, and it bothered me that things came apart from the middle on.

Maybe I'm just greedy, but I always try to pack in as much content as possible. I'm not sure it's true, but people tell me what actually draws them in, and gets them to recommend us and bring friends, is the organizers' energy, the raw way we run things, and the honest content. I know they're not telling me to do it halfway, but that hit home.

As always, recaps went up on Discord after the event, and everyone wrote only nice things. Part of me still felt conflicted. So I just told our Discord, with its 1,900 members, honestly how disappointed I was in the day. It was 12:25 a.m., so plenty of people probably missed it, but keeping it bottled up felt too stifling.

Luckily, supportive posts and personal DMs kept coming in, and for a second I wondered if this is how celebrities feel when comments cheer them up, haha. Anyway, nobody wants to do better more than I do. I'll keep preparing the next events step by step, and I think the more I do this, the more I'll figure it out.

The core of a winning proposal hasn't changed

On Tuesday, June 9, we held Claude Bloom × AWS × Cliwant × LG U+ at DreamPlus Seolleung, a startup center in Seoul's Gangnam area. It was our third event that week.

Some days I wonder why I'm doing this. Strangely, on the morning of an event, the energy comes back. I start looking forward to who will show up, what I can give them, and what I'll learn myself.

I picked this topic because, as the events piled up, I kept hearing that sales in particular is a field where "AI has a hard time getting in" and people still have to do more of the work. That made me want to dig in. I wanted to look at which parts of sales AI can automate and which parts people should focus on more.

There was a reason behind the three speakers, too. Jihoon Park from AWS covered global standards and proposal design. Junho Cho of Cliwant covered AI automation and redesigning the process. Sangmyung Yoon from LG U+ (one of Korea's big three telecom carriers) covered presenting and communication. Together they made one value chain, from sales all the way to winning the contract.

DreamPlus sponsored the space and the catering. The room was big, and it was still full, with more than 150 people.

Claude Bloom is a community run by an AI company, so unlike a typical event organizer, our bar is to make events we'd want to go to ourselves. We always plan around topics we care about and whatever is hottest right now, and the format goes beyond a one-way seminar. Like this time, we add small-group sessions and have each group present at the end. We're also planning a sales hackathon for salespeople, where they build a sales deck in 15 minutes with a tool like Genspark and pitch it, and an event that pits professional lawyers against AI models.

At the opening, I announced one big plan. In September, I want to rent out an entire pop-up street in Seongsu (a trendy Seoul neighborhood) and bring in well-known AI companies from abroad for a three-day, two-night event. If people come from the US, Europe, and Southeast Asia, and not just Korea, nobody has to fly to Silicon Valley to find insight. The idea is to grow our current hundred-person scale 30 or 50 times and put on a show of Korea on our home turf. I said it out loud because I think you have to announce a dream for it to come true.

1. Why the sales team has to be in the field

Jihoon Park, Proposal Manager at AWS, led the first session.

His core point was one idea: TITO, Trash In, Trash Out. Put in poor information and you get a poor proposal out.

The proposal team only knows what's written in the RFP. An RFP is a formal written document, so it doesn't carry what the customer really thinks. It isn't honest. Public information covers 20 to 30 percent of the picture, and the other 70 to 80 percent you have to go and ask for in person. That's why the sales team has to be in the field. If the sales rep didn't collect a need, not one line about it can make it into the proposal.

There are standards for what to collect. MEDDPICC, a sales framework, is an acronym for the kinds of information you need to gather, and three of them form the backbone that decides your win rate. First, you have to prove in concrete numbers what economic value the solution gives the customer, because once a price is attached, the customer wants value bigger than that price. Second, you have to find the hidden needs, the real worries that aren't written in the RFP, because the actual reason they put out the bid sits outside the 20 to 30 percent the RFP covers. Third, you have to confirm who the final decision maker holding the budget is. Without these three, a proposal is just castles in the air.

He also pointed to the death triangle that leads to failed proposals. The writer wants to show off, so the analysis of the customer's pain points comes up short. The proposal lists features from the supplier's point of view instead of customer benefits. And the information the customer actually wants to know goes missing. Problems always show up at these three corners.

Proposals are almost always winner take all, so differentiation is everything. Before, under time pressure, the work was modularizing documents and assembling them, checking for typos, and matching the requirements scoresheet. In the AI era, people have to add something AI can't come up with. Everyone uses AI, so if you only submit what AI produced, A, B, and C all end up looking the same.

Designing a win theme has principles too. Combine proven core ideas so they fit together, back them up with quantitative evidence from solid data, and target hot buttons that neutralize the competition. Neutralizing competitors is a technique used a lot in the US. A competitor always beats you in some area, so you kill that strength by framing it as "not important for this project." It's called ghosting, because you make the thing vanish like a ghost. As for hot buttons, the customer never shows them first. Like a hot potato, there's an issue they need solved but won't bring up, so you have to dig it out through information gathering.

He told us to try putting the customer's pain point into one sentence. Fitting all your thinking into a single sentence is harder than it sounds. It gets long, picks up flowery language, and drifts off course. A sentence that's compact, simple, and easy for anyone to understand is the best form.

In the AI era, once everyone uses AI, proposals end up with similar structure and similar sentences, and differentiation disappears. So he said, "Whoever reinvents the process wins." Beyond using AI well, what matters now is how you build AI into your own process, and the question of what the human role is starts here.

2. Being good at AI and having a company that grows are different things

Junho Cho, CEO of Cliwant, started with his conclusion: "Even in the AI era, the core of a winning proposal doesn't change. What changes is the complicated process."

First he gave a quick explanation of APMP, the Association of Proposal Management Professionals. It's a global association of proposal professionals headquartered in the US, with about 18,000 members worldwide, and it spent decades defining concepts like the win themes and hot buttons Jihoon Park had just talked about. Cliwant runs its Korea chapter.

He talked about changing the question he was asking. He's building a product as a startup, and he was afraid the product would get eaten by AI in the end. Swap the product for himself and the fear was the same. Negative thinking like "it won't work, it'll get eaten anyway" just goes on forever. So he flipped the question. Instead of looking for a business AI won't eat, he asked which business AI would make bigger no matter what. That's when answers started to show up.

He said being good at AI doesn't set you apart anymore. His son, who's in elementary school, makes dozens of games with AI on his own. If elementary schoolers can do it, what should professionals be doing? His conclusion was that you have to be able to redesign your business with AI.

He was candid, too. Starting this January, his company rolled out AI across the whole team, set aside Thursday afternoons entirely for AI experiments, and switched every tool to an AI-based one. The company didn't grow. He felt it in his bones that better AI skills don't make the business grow, and he concluded the company only grows if you can redesign the business itself.

In that context he introduced Contrl, which supports the entire public bidding process with AI. It connects everything from finding the right contacts for pre-sales to bid analysis, strategy, and generating the proposal deck. The bidding process APMP defined is the same for big companies and small ones, in the US or Singapore. In that process, AI handles the repetitive work, and people make the judgment calls: Go or No-Go, how to build the strategy, and the presentation. The presentation is the one thing AI can't replace.

He gave a live demo. He opened with the question, "Can you make a hundred winning proposals a day with AI?" He said it's possible because they've extracted all the text from RFPs and statements of work across seven years of bid data. Ask it to "search for officials on AI projects, sorted by most projects," and you get a list of officials who actually ran AI projects. Pick a person and you see the projects they handled, which makes it possible to meet a hundred pre-sales contacts in a day. Ask for "the last month of AI-related bids," and the notices come up, and you can narrow down one level at a time, to public AI services and then to agentic AI. Pick a specific notice and you get everything at once: an analysis of the hidden needs in the RFP, a 3C analysis, a conditions checklist, a bid simulation showing your expected score, and the list of documents to submit. He said you can't do that in GPT or Gemini.

The proposal strategy step was the key part. What APMP cares about most is the sweet spot, where the customer's needs overlap with your strengths and the competition has nothing. Everyone sees what's inside the RFP, so to build a win theme you have to bring in information from outside it. The AI asks your company questions to draw out strengths, differentiators, and evidence. Upload interviews and your company deck, and that information builds up in the 3C analysis. From there it points out your differentiators and weaknesses, how to respond, and what information gaps you still need to fill, and it turns your win theme into one sentence. Last, it pulls the table of contents the RFP requires and maps your strategy onto it, because a lot of the time the strategy is set but never actually makes it into the deck. AI-written proposals also tend to have an oddly casual or awkward tone in Korean. So it pulls frequently used components from your company's templates and polishes the proposal until it looks like your own team lead made it.

The Korean B2B SaaS market is at most 2 trillion won, while the public bidding market is 200 trillion won. In decades, not one company has succeeded with B2B SaaS in Korea, and Cliwant was one of those companies too. When they flipped the market and said they'd go into the 200 trillion won market together with their customers, the size of the pie changed. They changed direction from selling a tool to going into this market alongside customers. He said the definition of a startup is breaking the existing order of a market with new technology and an innovative business model.

There was also an offer he'd turned down at first. Since the company was founded, customers kept asking him to "just do all of it," to write the proposals and win the contracts for them. He said no. Winning depends on a lot of factors, so it's hard to measure as performance-based pay, the output per resource was unpredictable, and he believed a startup should compete on its product. Then last week he changed his mind. If you have an AI loaded with the entire bidding process, which doesn't change, one person can do it. Write dozens of proposals a day, enter lots of bids, and you get a lot of hits. His biggest misjudgment was thinking a startup has to compete on product. Traditional consulting was out, but AI-maximized consulting that finishes a two-month process in a week seemed worth trying.

The on-site story from Singapore stuck with me most. When they bid on a local interactive display project, they were the most expensive option of all the competitors. Right before the presentation, they went to the site early and found a display on a restroom wall that was turned off. They took a photo, put it in the presentation, and said, "We'll get this connected for you." The customer was moved by that and picked the most expensive option. That was the last 1 percent AI can't make for you.

3. Trash In, Jewel Out

Sangmyung Yoon from LG U+ introduced himself as a national-team-level presenter. He works full time while going to grad school, he's written two bestsellers, and he's a content creator. He brought copies of his new book, "The 1% Leader's Language Formula", as prizes for the best small-group presenters. He said he wasn't a born presenter. He was a quiet kid, and he lived outside Seoul for 33 years before learning to speak with a Seoul accent, so he's living proof that people who grow up speaking standard Korean start ahead. His one-liner was "Trash in, Jewel out." Even with the same information, who produces the output and how is the human part.

He started by defining the times we live in. Professor Jinnam Choi of Seoul National University's business school borrowed a military term and called this the VUCA era: volatile, uncertain, complex, and ambiguous. AI makes this even stronger. In a meeting, a team member clicks once in AI and gets information from the whole world, so a leader can't hand down every right answer anymore.

In the AI era, presenting in person matters more. Some professors are dropping written reports and grading by presentation instead. Cover letters have all gotten better with AI, but more and more candidates don't live up to that level in the actual interview, so he expects more interview rounds to become a hiring trend.

He also talked about accountability. Self-driving technology is done, but commercialization is slow because when there's an accident, the question is whether the driver or the manufacturer is responsible. Documents written by AI are the same. Responsibility ends up with a person. So judging and taking responsibility is the human's job.

He listed things only people can do. During his talk, he stopped speaking for a few seconds and you could feel the air in the room change. AI can't create that atmosphere, he said. Same with nonverbal cues like hand gestures, eye contact, and footsteps. Humans are the only animals whose eye whites show when you look at them head on, so you can tell right away where someone is looking. AI can't imitate that.

He introduced two speaking frameworks. OREO is Opinion, Reason, Example, Opinion, a formula for logical persuasion that he said Harvard has tested for over 100 years. STAR is Situation, Task, Action, Result, which works well for explaining a situation in a structured way. He pointed out that Junho Cho's talk had followed the STAR structure, and said well-prepared talks fall into these formulas even when nobody plans it. Structure decides how well a message lands, more than content, and good presenters use these formulas naturally without thinking about them.

He also flipped VUCA into words for what only people can do. V for Vision, U for Understanding, C for Clarity, A for Agile. Only people can set out a vision, deliver it more clearly, and respond to a situation more nimbly.

He talked about stamina, too. AI runs 24 hours a day and people have physical limits. He wasn't saying we should compete with AI. To bring out the sensory parts, like atmosphere and reading the room, your body has to hold up. When you're too tired, kind words don't come out. Your language gets clearer when your body is healthy.

Asked what the last weapon people have to hold onto is, he said humor. Humor is what makes us human, and only people can give that human touch. Because he thinks analog human touch will be worth more the further we get into the AI era, he's been practicing his handwriting lately. "Submit your proposal handwritten. Maybe that world is coming." It was a joke, but it sounded like he meant it.

4. What came out of the fireside chat

These are the things that came up in the fireside chat with all three speakers.

Each picked a one-liner from their talk. Jihoon Park's was "It takes human design and human thought," meaning a person has to think through what AI suggests and submit it under a person's responsibility. Junho Cho's was "The core of a winning proposal doesn't change." Sangmyung Yoon's was "Trash in, Jewel out."

They shared what they'd learned from each other. Jihoon Park said evaluators look at the presentation deck before the full proposal, so he needs to learn more about presenting. Junho Cho said he got tense when Sangmyung Yoon suddenly paused mid-talk and wants to use that silence himself, and added that Jihoon Park's talk put into a system the proposal work he'd been doing by gut feel. Sangmyung Yoon picked Junho Cho's sweet spot idea: the key is catching the thing you have, the customer needs, and the competitor doesn't have.

Someone asked how a global company like AWS runs its proposal process. Jihoon Park laughed and said it's a company secret, but he described the common framework. Before deciding whether to write a proposal, you decide whether to bid at all. A lot of money is often at stake, so they put a lot of time into gathering information from the start. APMP has something called color reviews: five color-coded stages, from the team's first meeting to build the outline to a review by two or more executives in charge, where every two days you go through what to put in and what to cut. The submission window is only three to four weeks, so you have to run this whole standard procedure inside it.

Asked how to find a customer's hot buttons, Sangmyung Yoon said that with a private-sector customer, you have to think about your customer's customer. If you're pitching Shinhan Bank, you have to keep talking about how satisfied Shinhan Bank's customers will be. That language is already on the customer's website, in their mid- to long-term strategy reports, IR materials, and executive interviews. The words that come up a lot there are the hot buttons. In the public sector, the language changes to public benefit and resident satisfaction.

Jihoon Park also summed up how bidding differs in Korea and overseas. In Korea, you submit a PowerPoint deck and get judged on a presentation, and the questions are direct: can you do this, how will you do it. In Australia and Singapore, you submit in Excel. If around 100 questions come in, about 15 are general company information and the other 80-plus are specific technical and security items. You write your answers in Excel, down to naming specific services, and they grade how precisely you meet the conditions. He supports Australia and Singapore from Korea, and the people in those countries take the parts they know better, so the collaboration runs around the clock. Deals where compliance matters, like public projects, are led by people like him, while private-sector proposals, for a company like Samsung, are led more by sales.

An interesting relationship came out, too. Sangmyung Yoon added that for LG U+, AWS is a super VIP customer that uses its network and many of its services. So when LG U+ pitches AWS, it submits in Excel the American way, sometimes with over a thousand requirements. AWS isn't comparing presentations. It looks at how precisely you fit.

To sell globally, knowing the country's culture comes first. Jihoon Park said that with AI, the written language barrier isn't a big problem anymore. Even when the language works, though, the way people work in that country is different, so jumping in blind without knowing the culture won't get good results. You have to understand how people work before the language.

Asked how companies train presentation skills, Sangmyung Yoon said personal drive comes first. He took every in-house PowerPoint course as soon as he joined, but it wasn't enough. He asked an instructor, found out there was a certification course, became the first person at the company to get certified, and now he's the one teaching it in house. When personal growth and the company's growth goals line up, and it's clear that leads to results, the company is glad to support you.

Tacit knowledge came up as well. Junho Cho said that even three years after leaving his old job, he still gets calls asking, "Where's that proposal from back then?" The proposal file is still there, but the context of how it was made is gone. You can see the proposal the coworker next to you made, but not the context it came from. Methods are appearing that use a knowledge base or an LLM wiki to build up this tacit knowledge. You have to keep the context of the process, and not only the output.

Asked how to handle a situation that throws you off mid-presentation, Sangmyung Yoon said the answer is to never let it happen. That day, someone needed a clicker, so he lent his, and it didn't work. That's why he carries two. He backs up twice, three times, practices 200 or 300 percent, and prepares answers to expected questions at 500 percent. He'd heard Steve Jobs kept three generators at product launches. It might not be true, he said, but that's the level of preparation he means. If something still goes wrong, sincerity is the best answer. If you're not prepared, say so honestly, and promise to email it by 3 p.m. that day. That's how you win back trust.

Someone asked what their hit rate actually was and what mattered most in winning contracts. Junho Cho said that when you and the competitor offer about the same thing, it comes down to price, and winning on the lowest price isn't a good proposal. When things are close, you have to go to the site. That's where his Singapore restroom display story came from.

There was a sharp question about pre-sales, too. A proposal starts and ends with looking people in the eye, so if you just take what an official told you over the phone and put it into the proposal, how solid can the pre-sales work really be? Junho Cho said he completely agreed. You can't know the hidden needs without meeting the customer ahead of time. AI doesn't do all of that for you. It helps you meet the right contacts for pre-sales, tells you what questions to ask, and helps you put what you learn into your proposal strategy, all the way to shaping the RFP in a direction you can win. The core doesn't change.

5. Small-group presentations

The small groups discussed one topic: how am I doing sales and proposals differently in the AI era?

One thing came up across groups: AI gives you a proposal worth 70 points, and the last 30 points are up to people. How deeply the presenter has absorbed the proposal decides the outcome, so the reviewer's skill matters more now.

Someone also said that with AI, junior people can sell like senior people. That's a change to the structure of the organization itself.

A group focused on building up data said many people already store their day-to-day work data well and call it up with tools. They agreed the direction is to build up data first, whether it's company data or personal data.

A group focused on finding the hidden key person first talked about using LinkedIn to find people at the target company and pin down their pain points. Finding material AI hasn't learned and can't produce, and using it early, gets you closer to a winning proposal.

A group comparing sales in Korea and abroad said Korea has a lot of spec-focused, transactional proposals, while overseas, pushing specs at people actually gets you rejected. The relationship comes first. You understand the other person's story and become friends, and only then does a proposal get through. Either way, the last mile is human touch.

One group brought up the invisible hand in bidding. Nobody had an answer for how to manage it as a risk, but everyone agreed that finding a way to win in that situation anyway is the proposal team's job. The same group also wrestled with how to turn technical language into language judges can understand, since non-engineers often write and present the proposals.

One group presented in English. It was the group of Ali, the founder of Screenshow, a tool for making demos. They talked about how it's easy to build an app these days but hard to promote it and make money. Designers and founders don't like dancing on TikTok or putting their faces on social media, so the idea was an engine that uses AI to automatically produce product promo videos, leaving marketing to AI so they can focus on building.

The last group's presenter left us with this: "Nobody can tell you the textbook way to bid and win contracts. Which is exactly why we, studying here this late, might become the people who can't be replaced."

After the presentations, Sangmyung Yoon picked three people who stood out and gave them his book. One opened with "To sum it up in one line," which naturally used the first letter of the OREO formula. One used the most meaningful gestures, like sweeping a hand from "New Zealand" to "Korea." One said "There are three things," and structured it as first, second, third. The judging showed exactly that structure and delivery stay with people more than content does.

Originally written in Korean on Brunch, June 10, 2026.