Three years of convening

Mid-afternoon at the first one, Tokyo, 2024. I step out of a session to check on something and the corridor is fuller than the room I just left.

Not a few people. Most of them. Standing in clumps of three and four, talking, cups going cold, nobody moving toward the doors. Inside, four people on a stage are working through a question I had written months earlier, and it is a good question, and almost nobody is in there.

My first thought is that something has gone wrong. It takes me until the evening to understand that nothing had. The corridor was the event. I had built the room facing the wrong way.

I have convened the Tech for Impact Summit in Tokyo since 2024. Three of them so far, and a fourth on 18 and 19 May 2027. What follows is mostly the list of things I got wrong.

2024: I programmed it like a syllabus

The first summit ran from nine in the morning until after six. A dozen sessions between the opening ceremony and the closing one. I built that day myself and I was proud of it, and it was far too much.

Look at what I called things. One panel was “Impact Transparency Dilemma: Balancing Disclosure and Communication in the Age of Accountability”. Another was “Sustainable Innovation: Mitigating Unintended Environmental and Social Consequences of Technological Advancements”. Nobody can say those out loud. I wrote them the way you write a grant application, which is what I knew how to write.

I also put a session at one o’clock, on top of the lunch. It is still there on the archive page if you want to see it. So the one hour of the day where people were actually meeting each other, I interrupted with a panel, and then I wondered why the corridor was so busy later.

And I scheduled two sessions in the same three o’clock slot, which meant every person in the building had to choose, and plenty of them chose neither.

The closing ceremony was at quarter to six. By then the people I most wanted in the room had gone off to a bar somewhere to have the conversation the day was supposed to produce. Good for them.

Takuya Hirai spoke that year, and Charles Hoskinson. Real people, real substance. What people told me about afterwards was the lunch.

2025: bigger was the wrong direction

So the second one was larger. Many more speakers, many more people, several times the room.

I thought scale was the fix, because scale is what the outside world reads as success. Audrey Tang came. Kathy Matsui came. Every metric a summit reports went up.

And the thing I actually wanted got harder, not easier. In a big room you cannot find anyone. You spend the day in a state of low-grade searching, half-listening to whoever is in front of you while scanning for someone else. Nobody has a conversation that lasts twenty minutes, so nobody has a conversation at all.

Attendance and value are unrelated, and I had spent a year confusing them. That one cost me a full cycle to learn.

2026: one day, and the first quiet room

The third was a single day, 26 April 2026, at Kioi Conference in Tokyo. Taro Kono spoke. It was tighter, and I finally did the thing I should have started with: private sessions at small tables, held under the Chatham House rule and not recorded or written up afterwards.

Those rooms were the best part of the day and I have thought about why ever since.

Part of it is that nobody performs at a table of ten. On a stage, even an honest person is aware of the room, of the recording, of the colleague in row four. At a table with a rule that nothing leaves it, a senior person will say the thing they actually think about their own industry, which is usually more interesting and always more useful than the version with the logo behind it.

Part of it is that a table has no audience, so everyone has a job. Twenty minutes in, somebody who came to listen is arguing. That does not happen in a lecture, ever, and it is the whole reason to gather people in one country.

What invitation-only actually means

We say invitation-only, and I know how that sounds. For a while I was uncomfortable with it, because I have spent most of my life on the outside of rooms like this one, and I did not want to build another door.

Here is the honest version. The people you want in a room have the most alternatives. Their day is expensive. If they come and half the people there want something from them, they will not come again, and neither will anyone like them.

So the invitation is a promise, and the promise is not to the person being invited. It is to everyone else already in the room, about who they are going to be sitting next to. Every seat given away carelessly is taken out of somebody else’s day.

That is the whole of it. It is a curation problem dressed up as a status one, and I dislike the way it reads, and I have not found a better way to do it.

What the convener actually does

Not the programme. I used to think the programme was the job.

It is the seating. Who is next to whom, who is walked over to whom, which two people I put at the same small table because I have a hunch. It is unglamorous and slightly manipulative and it is most of the value I add on the day.

And it is the door. Deciding who comes. Saying no to good people, which I am poor at, and which I still get wrong in both directions every single year.

The 2027 summit is on 18 and 19 May in Tokyo, co-hosted with the United Nations University. It is smaller on purpose. I expect I am currently wrong about something in it that will be obvious to me in about eighteen months.

I will be at the door.