Why Salespeople Aren't Afraid of AI

I'm no sales expert. I just pulled this together from interviews I did. But looking at the result, I think there's a lot in here that salespeople will relate to. I'd love to hear feedback and stories from people who actually work in sales.

People often ask me why I run the Claude Bloom community. Honestly, it started pretty organically, so we're still figuring out its purpose. But for me personally, the biggest draw is one thing: I can easily meet people from all kinds of fields whenever I want. When someone at an event catches my interest, I send them an email, and most of them gladly make time. They talk openly with me about everything from their personal worries to what they hope to get out of the community.

Why salespeople don't fear AI

Recently I had the chance to talk with people who work in sales. "Blue" and "Bloom" in the AI era are the main themes of our community (see Claude Blue and From Claude Blue to Bloom), so we covered a lot of ground, and I noticed something interesting they had in common. People in sales aren't very afraid of AI.

AI clearly helps with existing work, and the gap between people who use it and people who don't will keep growing. But at its core, sales is about understanding people and persuading them. And people are, at heart, deeply irrational.

One of the abilities that sets humans apart from other animals is reading feelings and thoughts from someone's face. That's why important job interviews happen in person, and why people get on a plane for a big negotiation. When you share a room and breathe the same air and pick up on the other person's body language and nuance, an invisible battle of nerves and the craft of negotiation play out.

Humans keep control

We'll work to keep control of important decisions for ourselves and not hand it to AI. Small decisions with little impact can go to AI. But even if AI can reach more neutral and rational conclusions, having someone else be the one who decides, instead of me, is a different matter entirely.

It's like being the person at a company who gives the final sign-off. Even if other staff have already reviewed something, you still get briefed and make sure you understand it. You aren't fact-checking because others might be wrong. You do it because you're the final decision maker.

That's why big business deals always end up going through human hands, and irrational human nature always affects those decisions. Salespeople seem to be born with an uncanny ability to pick up on exactly this, almost like it's in their DNA.

The two sales experts in this story

There are two people I want to introduce here.

The first is Choi Jong-il. He's been in presales and worked as a solutions architect for 27 years, and now runs PreSales Lab for people thinking about a career in presales. I've known him for a long time, but I asked him for an interview because I was curious where sales skills are headed in the AI era. He readily agreed and shared a lot of deep know-how and insight.

The second is So Byung-wook. He handles presales at a Japanese IT company. He came to a recent Claude Bloom event, and his title was literally "PreSales Engineer," so I couldn't help being interested. He used to be a developer, then read Choi Jong-il's book on presales, fell for sales, and switched fields. I only found out later that the two of them were connected that way. He has an unusual position that spans both tech and sales, and he's also on the organizing team of the Cloud Club community.

1. The conclusion of 27 years: the buying journey doesn't change, even with AI

Choi describes the customer's path to buying something in four stages: awareness, consideration, evaluation and purchase. He says this flow stays the same in any industry and any era.

Awareness is when the customer has a vague sense that something is wrong. Sales' job at this point is to turn that feeling into a clearly defined problem. In consideration, the customer looks for a way to solve it: what technology they need, where to start. What helps here is designing the direction together with them. In evaluation, the customer has already narrowed down the candidates and is looking for criteria. And right before purchase, even after the choice is made, a risk remains: "Can we actually make this project succeed with this solution?" What the customer really needs is different at each stage, and if sales talks about the wrong stage, it won't land no matter how good the pitch is.

What AI changes is how the customer gets information at each stage. Customers ask ChatGPT first, which compresses awareness and consideration. Some show up with a comparison chart of competitors they've already made. Choi summed it up this way: we've entered an era where customers make about 60% of the decision before they ever call a vendor. For sales, the moment to step in keeps getting pushed later. There are two ways to respond. One is to design your content so your company's information shows up in AI search. The other is to take another look at the problem definition the customer brought in. The bits of information AI gives out are full of errors, and defining the problem properly is still a person's job.

There's a reason presales is getting attention in this shift. In the past, sales would say "let's do it" and the tech team would follow. Now presales has the authority to filter deals at an early stage. Their role is to spot deals that are technically impossible or don't fit the budget before sales does. As the information gap shrinks, what customers want is judgment, and information matters less. That's the conclusion of 27 years in the field.

2. By the time it's public, it's too late

Both of them said the same thing: by the time an RFP is published, the sale is already over.

So put it this way: "Once the RFP is out, the game is already over." When a procurement notice goes up, it means someone has been building the relationship that led to that notice. If an RFP closes in a week, someone already got there first. An RFP is the result of selling, and the starting gun comes long before it. At first there is no RFP. It takes shape as the customer's needs get clearer over countless emails and meetings.

Choi said the same thing differently: "If it's public, you're already late." The window where sales actually works is awareness and consideration. You have to be in the room when the customer first starts to feel the problem, when there's no budget and no decision yet. A salesperson who first shows up at evaluation or purchase is just going through the process.

Sales organizations at companies that have survived in the global market for decades aren't strong for no reason. They have solid methods and playbooks, tested in the field over a long time. So boiled down the core like this: the ability to figure out who pays, meaning who the real decision maker is, remarkably fast. Reaching the decision maker before anyone else matters more than competing on specs, and global sales organizations have spent decades sharpening that skill.

3. The one thing humans are good at

I think one of the things humans do better than other animals is reading faces. Tiny movements of facial muscles, where someone's eyes go, shifts in their tone of voice. We put all of that together and read in real time what the other person is thinking and how they feel. That ability was shaped by millions of years of evolution.

Talking with So confirmed this for me. What matters most to him is reading the context in the room. Who nods in the meeting, who crosses their arms, who puts their phone down: these show you the real decision-making structure. You get more information from how people talk and react than from what they say.

Human decisions are irrational. People don't buy because the specs are good. They buy because they can trust you. The feeling of "I want to work with this person" is what gets a contract signed. The essence of sales is drawing out pain points the customer hadn't noticed yet, and sometimes making them aware of a need they didn't have. That's a matter of instinct and experience, and AI can't replace it. That's why enterprise sales isn't too worried about AI.

So has a phrase for his work: "Presales is a fight with unstructured logs." Emails, meetings, recordings, handwritten notes: all of it piles up as unstructured data. How you turn it into something reusable decides how productive presales can be. He solved this fight with a system.

4. Designing how work becomes data

The core of the AI harness system So designed himself is a structure that stores work context as data. The way he put it stuck with me: "I didn't organize my work. I designed the way work becomes data."

Here's how it works. It starts with entering every work log from the day (emails, meeting notes, recordings, things he learned) into a daily note in Obsidian with markdown tags. #history is for records about client companies, #company for internal governance, #solution for information about his company's products, and #learning and #project separate what he's studying from deals in progress. As the raw logs pile up, they're automatically committed to GitHub, and a cron job runs regularly to store them in chronological order in a predefined folder structure. An agent reads these raw logs, summarizes them, and turns them into a reusable knowledge layer.

He explains this flow with an ETL frame. Extract pulls data from emails, meetings, recordings and learning. Transform is where markers and an LLM combine it with existing context and process it. Load puts it into reusable layers called History, Knowledge and Word. The data layers build up from Daily to History, Knowledge, Word, Governance and Output. Daily raw logs accumulate into history, patterns pulled from history become knowledge, and knowledge turns into actual deliverables.

For security-sensitive work, he uses LM Studio to run an on-premises LLM, so he doesn't have to send company data to an outside model. These days an agent handles all the committing, pushing and organizing.

His automation roadmap is also designed in stages. It evolves from observation to raw logs, a reuse layer, rules, automation, delegation and finally a work OS. There's a distinction he stresses. The goal of automation is to remove people. The goal of delegation is to leave only the points a person needs to look at. The philosophy is to focus human involvement on the moments where it's needed most, without taking people out completely.

5. Relationships follow people, not companies

Both of them saw the same thing over long careers in the field: trust builds up in people, and the company brand matters less. Choi told me about a financial client that one person handled for nearly 20 years. That salesperson changed companies several times, and the client relationship came right along. Enterprise sales is a relationship game through and through, and it's hard for an information service or AI to quickly catch up with someone who has covered the same client for more than ten years and built that trust.

So is extending this relationship capital with a system. People, context and the flow of conversations are kept on record, and those records help him prepare better for the next meeting. It's a way to manage relationship capital in a structured way without relying on memory.

He explained the core of the system like this: "Frontier models are the CPU. If you build a good OS, there's no vendor lock-in no matter which model comes in." Whether it's GPT or Claude, models keep changing. But the work context, customer history and knowledge structure he has built become assets he can plug into any model. "Models change. The harness stays."

He also said, "Systems compound." At first it felt like things were just piling up. Now the context he's built up actually saves him time. He spends less time reviewing past conversations before meeting a client again, and less time hunting for past cases when writing a proposal. He can focus on what matters more.

6. Why humans need to push ahead

People who worry that AI will replace sales seem to see sales as passing along information. From that angle, the worry is fair. AI passes along information faster and more accurately.

People who have done enterprise sales for a long time define it differently. To them, sales is discovering, together with the customer, a problem the customer hasn't clearly recognized yet, and designing the solution with them (co-creation). The most important thing in that process is trust, and trust comes from repeated meetings and judgment built up over time.

The more AI can do, the higher the bar rises for what people need to do. Organizing information, preparing materials, first drafts: these move to AI quickly. With that time freed up, people need to meet customers more often and have deeper conversations.

So describes his role as "compiling between tech and sales." He translates technology into the language of sales and customer needs into the language of technology. AI can help with that compiling, but understanding both languages at once and making calls in real time in the room is still a human job.

What I took away from talking with them is that good sales was already doing work only people could do, long before AI. That's why they're so calm about it. They can tell what changes from what doesn't, and they've spent a long time focused on what doesn't.

The people who turn that instinct into a system will get ahead in the AI era. They use AI to get even better at what humans are good at. I think that's the direction humans need to push in.

Originally written in Korean on Brunch, May 24, 2026.