Skip to story
Benjamin Oh
BackgroundAI workProjectsCoachingNewsletterCoaching siteGet in touch

hi :)

Benjamin Oh

I'm Ben.

I work where AI meets value creation. I started in finance, restructuring and private equity, and now work on taking AI offers to market at BCG. I like starting from a real problem a team has, not an abstract use case, then building something small that works and checking whether it actually saves time or money.

Read the backgroundLinkedIn

Background

I'm Ben. This is how I got here. I also coach people who feel stuck.

Singapore

I grew up in Singapore in a family that didn't have much. My parents worked seven days a week. I'm the eldest of three, and I trained in finance and accounting. That's not a very exciting start, but it got me my first job. I repeated a year of junior college because my grades weren't good enough.

2015 SEA Games

In 2015 I got cut from the SEA Games softball squad, went to a training camp in Japan anyway, and made it back. We won bronze with my whole family in the stands. Then average uni results and a lot of rejection emails.

Singapore men's softball team at the 2015 SEA Games

got cut, then made it back

Singapore men's softball, 2015 SEA Games. Bronze. (match report · roster)

The career fair

I applied for every internship and got none of them. I went to the career fair anyway and talked to the recruiters. One conversation became an interview at a local bank, then a summer internship in banking. I also turned down a written offer from a Big 4 firm in audit when I had no other role.

Sierra Leone

That job was in restructuring at Borrelli Walsh. For three years I worked on companies that were in trouble across Southeast Asia, and once on site in Sierra Leone. When a company is running out of money, nobody wants a slide deck. They want to know what happens on Monday. I learned more there than in all of uni.

Benjamin Oh on site in Sierra Leone in 2019
On site in Sierra Leone, 2019.

An operator-first PE firm

I left Borrelli Walsh in the middle of Covid to join an operator-first PE firm. My MD texted to ask if I was sure. I'd got to the end of what I was going to learn there, so I went.

Since then

Then I went into private equity as a sort of mini-CFO for a business in the portfolio. Fewer people, more hats. I did the numbers, and also anything else that needed doing.

In 2021 I joined BCG in Singapore, in the transformation team.

Part of that work was the transformation of a two-billion-dollar company. The move to London was a long run of nos, from external recruiters, internal recruiters, and my own manager. I kept at it. In November 2024 I moved to London.

From November 2024 to February 2026 I was at BCG X. I'm now a Global Offer Senior Manager - AI Transformations at BCG.

Somewhere in the middle of all that I got a coach. I was good at the job and a bit lost on everything else. It helped so much that I trained to do it myself, and got my ICF Associate Certified Coach credential in 2025.

My other certificates are on Credly, if you're the type who checks.

None of it went the way I planned. I just kept turning up.

The GPT I gave up on

I'm a finance person. I don't write code for a living. But in 2023 I was one of the early users of ChatGPT at work, and I've been tinkering ever since.

The project that taught me the most was one I gave up on.

In early 2025 I built a few custom GPTs for my team, for things like looking up assets and finding colleagues with the right skills. They got about 80% of answers right. That sounds fine until you're the one explaining the other 20%. Nobody changes how they work for a tool that's wrong one time in five. Nobody did.

I spent a while working out why. The model was reading our Excel files less like a table and more like a long list of labels. It did well on the first few columns and got worse the further across it went. Better prompts didn't fix it. So I shelved it and felt a bit silly.

Around the middle of 2025 the models got a lot better. I went back, gave the same GPT the same spreadsheet and asked the same questions. It got about 95% right. I hadn't changed a thing.

Early 2025

80%

Mid 2025

95%

Same GPT, same spreadsheet, same questions

it worked the second time

The lesson I took from it: sometimes your idea is fine and the model just isn't ready yet. So build it anyway, park it, and check back in a few months. Plenty of people try AI once, decide it doesn't work, and never look again. I nearly did.

Since then I've taught myself enough Python to turn a two-hour weekly Excel job into a few seconds, and most of what I build now happens in Cursor. The rest is in Projects, just below.

Things I've built

Most of these were built badly first, then less badly. A few are still in the batting cage.

Personal OS

In progress

A set of AI agents that keep track of my bio, my AI notes, my projects and my coaching. Built in Claude Code. This site is one of the things it looks after.

Kith

In progress

An app for keeping in touch with the people who matter. You keep notes on each person, log your chats with help from AI, and get a quick prep sheet before you meet. Each relationship lives in a room of a house, depending on how close you are.

Email workflow

Cursor

Pulls the data, does the analysis and drafts the email, in that order. I used to do all three by hand. Now I just read it and hit send.

Pipeline dashboard

Replit. Contents blurred.

A pipeline tracker you can filter by system and region. It replaced the meeting where someone read an Excel tracker out loud. Nobody misses that meeting.

My coaching site

Built in Lovable from a few prompts. The first time I made something that looked like a real product without writing any code.

coachee

My ICF coaching session log. Session prep, reminders, hour tracking, and calendar invites. Built for me for now, and for friends later.

Revenue script

Python

Splits our weekly revenue across practice areas and regions. It used to take me two hours in Excel. I learned Python in a week to write it, and now it runs in seconds.

Team GPTs

Custom GPTs

The ones from the story above. Built too early, shelved, then brought back when the models caught up.

Coaching

I got a coach before I became one. I'd built a career that looked right on paper and never stopped to ask if I actually wanted it. My coach didn't give me answers.

She asked better questions than I was asking myself.

Who it's for

People who feel stuck. Maybe you're thinking about a career change, or a move, or you've just got a new role and you're not sure how to show up in it. Maybe things look fine from the outside and you can't work out why they don't feel fine. You don't need a clear question to start.

What I do

I listen and I ask questions. You do most of the talking and all of the deciding. I'll help you get clear on what you want and what's in the way, then we work out a first step you'll actually take.

What I don't do

I don't give advice or tell you what to do with your life. I'm not a therapist either, and I won't rewrite your CV.

I'm ICF-accredited (Associate Certified Coach, 2025).

If something's been on your mind for months and you still haven't said it out loud, book a chat. Worst case, you get to think out loud with someone who has no stake in what you decide.

Book a chat

Newsletter

One email a month. What I've been building with AI, what's working, and what I'm learning from coaching. Short enough to read with your morning coffee.

It hasn't launched yet. Leave your email and the first one comes to you.

Say hi

Email is best. LinkedIn works too. I reply to everyone, eventually.

  • Email
  • LinkedIn
  • Coaching site
  • Credly
BackgroundAI workProjectsCoachingNewsletter

← tap the can

and it stays on the bar

StartEnd of the bar

Since you scrolled all the way down: I also hold the Singapore record for the longest bar slide with a beer can. That was 2018. It was a big year. Here's the proof.

© 2026 Benjamin Oh