How to Book an AI Speaker in Japan: Criteria and Process

If you are organising an event on AI, the hard part is not finding a speaker. It is writing down what you want the audience to do differently the next morning. I give more than 150 talks a year, so I see the same organiser mistakes from the receiving end over and over. This article covers what to decide before you reach out, the criteria that actually separate AI speakers, the booking process and prep timeline, the differences between online and in-person sessions, and what changes when a non-Japanese company runs an AI session in Japan.

Decide these things before you contact an AI speaker

Write the takeaway as one sentence

“The latest in generative AI” is not a topic. It is the name of a field. Hand a speaker only that, and the safest thing they can do is deliver a broad overview nobody acts on. Narrow it to something like “an IT department lead removes one blocker from their internal rollout” or “executives leave with a way to judge AI investment,” and the structure of the deck and the choice of examples both change. The briefs where I can predict a good outcome are the ones where this sentence is already in the first email.

Describe the audience by state, not headcount

How many people attend matters less than where they are. Is everyone already using AI at work, or is half the room new to it? Has the company issued a direction, or are individuals experimenting on their own? When this is vague, the talk lands in the worst possible middle: boring for the people who know, too fast for the people who don’t.

Fix the talk-to-Q&A split before the deck exists

Sixty slides do not fit into a forty-minute talk. If you want a live demo, budget double the time you think it needs. And if you genuinely want the Q&A to work, shorten the talk and give the questions twenty minutes. That trade almost always raises satisfaction.

How to choose an AI speaker: criteria that actually separate candidates

Do they touch primary sources themselves?

In AI, information decays quickly. There is a visible difference between someone repackaging other people’s explanations and someone who uses the products and shows up where announcements happen — and it shows up in the Q&A, not the talk. I attended Google I/O in 2025 and 2026 as an invited guest, was invited to Samsung Galaxy Unpacked 2026 in London, and covered BoxWorks Tokyo 2026 on a press pass. Being in the room lets you say what was emphasised and, just as usefully, what went unmentioned. Ask candidates where they have seen things first-hand.

Do they know where deployments stall?

Explaining a product and knowing why a rollout stops inside a company are different skills. At Digirise, the company I run as CEO, we have supported AI adoption across more than 600 companies and over 50,000 people, and we hold the No.1 position in number of client companies for corporate AI reskilling services (Tokyo Shoko Research, as of the end of March 2026). What that experience tells me is that audiences rarely want a feature list. They want to know where other companies got stuck and how they got unstuck.

Are they close to your audience?

What lands with executives, with front-line staff, and with engineers is not the same material. Look at whether a speaker’s past venues resemble yours. I have spoken at GMO’s large conference two years running, at a seminar co-hosted with Anthropic in June 2026, at IVS Kyoto, and at SHIFT AI’s anniversary event — and I restructure the same subject for each room.

How often do they publish?

Someone who publishes continuously keeps their material current without being asked. I post on X as @masahirochaen with over 200,000 followers, run the YouTube channel チャエンのAI研究所 (Chaen’s AI Lab), and wrote the book 『AI脳』 (KADOKAWA, March 2026). A public track record is also your evidence: it shows what the person has actually looked at.

How is neutrality handled?

Many AI speakers hold vendor advisory roles or public positions. I serve as an AI & Web3 advisor at GMO, as a Google Gemini AI advisor, as a councillor of the Generative AI Utilization Promotion Association, and as a NewsPicks Pro Picker. The issue is never whether those roles exist — it is whether they are disclosed, and whether the speaker can still make comparisons. Ask about affiliations at booking time and, where relevant, have them stated up front. That avoids awkwardness later.

The booking process and the prep that matters

Six things to put in the first email

• Date, and whether it is a venue or online
• Audience profile and experience level, expected headcount
• Talk length and Q&A split, and whether there are other speakers
• Whether it is recorded, whether it will be published, and to whom
• Whether a deck must be submitted in advance, and the deadline
• Fee terms, and how travel and accommodation are handled

With those six, a speaker can answer yes or no and propose content in a single reply. “Could you speak about AI for an hour?” on its own burns two weeks in clarification.

Book a 30-minute prep call

Once the date is locked, schedule a short call two to four weeks out. The single most useful thing an organiser can bring is three real questions or complaints already circulating among the audience. That is what moves a talk from general commentary to something specific.

Let the deck move until the last minute

AI is a field with a continuous announcement cycle. Set the deadline too early and the speaker will be presenting stale material on the day. Share an outline early, then allow replacements up to the day before. If you need printed handouts, treat the handout and the projected deck as two separate artefacts.

Decide who runs the room

Whether the Q&A works is usually decided by the moderator, not the speaker. Collect questions in advance and have the moderator hold one opener for when no hands go up. Don’t leave time management to the speaker either — agree on a signal for remaining time.

Online vs. in-person: what actually changes

Online

• The shared screen is the whole event: large type, one point per slide
• Live demos break on connectivity — keep a recorded fallback ready
• You cannot read the room, so use polls or chat to create checkpoints
• Drop-off is visible, so state the conclusion in the first three minutes

In-person

• Size type for the back row, not your laptop
• Ambient light flattens screen contrast more than people expect
• Check whether the speaker can move, see their notes, and has a clicker
• Breaks and networking time contribute more to satisfaction than organisers assume

The one thing both share: test the connection and projection with the actual equipment you will use. Something goes wrong precisely on the days this gets skipped.

What changes when a foreign company runs an AI session in Japan

When an overseas AI or tech company holds an event or announcement in Japan, translating the headquarters deck into Japanese is rarely enough. I sit on both sides of this — attending announcements abroad, and speaking in Japan — so the gap is visible to me.

• Spend less time on features and more on who inside a Japanese company uses them, and how
• A session built only around an English-language headquarters speaker tends to produce a silent Q&A. Pairing them with someone who adds the local context gets questions
• In Japan, adoption decisions move through the organisation. Attendees need material they can use to explain the case upward
• Many companies cannot be named publicly. Prepare a way to describe cases anonymously while keeping the structure legible
• A press announcement and a customer-facing talk have different jobs. Reusing one deck for both makes both weaker

For organisers, the practical design is simple: if you include a session from an overseas company, place a local explainer in the same session or immediately after it.

How to apply this inside your own organisation

Choosing a speaker is the same exercise as articulating what your own company is stuck on with AI. Write the objective as one sentence, write down the real state of the audience, and decide the behaviour you want them to leave with. By the time you can write those three, you have already settled half of your internal training design and half of your project priorities. Whether an event ends at “that was interesting” or ends with a decided next step is usually determined by the first email you send.