How to Choose an AI Keynote Speaker

Why this market got weird

Since late 2022, the number of people calling themselves AI keynote speakers has gone through the roof. Plenty are great. Plenty spun up a deck in a weekend and started quoting five figures. For an event planner, the downside isn't a slow hour on the schedule. It's an audience that walks out with the wrong mental model of what this stuff actually does.

1. Vet depth, not vocabulary

Anyone can say "transformer" or "agentic." Ask a candidate to explain, in plain English, why models hallucinate, or when you'd reach for RAG instead of fine-tuning. If the answer turns into a word salad or a pivot to a client story, move on. Someone who has actually built with these tools can explain the mechanics without hiding behind the jargon.

2. Ask what they've shipped

Real practitioners have receipts: products, internal tools, MCP servers, automations, agents that broke in interesting ways. Ask what they've built with AI in the last six months. A vague answer about "advising Fortune 500 clients" is a yellow flag. A specific answer about a workflow, a prompt architecture, or why they picked one model over another is a green light.

3. Match the speaker to the room

A room of engineers doesn't need "What is ChatGPT?" A room of board members doesn't need a live demo of function calling. Most speakers have a natural altitude they operate at, whether that's strategic, operational, or hands-on technical. Be explicit about who's in the seats and what you want them doing differently on Monday morning.

4. Look for a balanced take

Watch out for the two extremes. The evangelist who thinks AI solves everything by Q3, and the doomer who thinks it's all a bubble. Both are fun to book and both age badly. The speakers worth paying for hold real optimism about specific use cases and real skepticism about the marketing wrapped around them.

5. Watch a full talk before you book

Sizzle reels are built to hide the weak spots. Ask for an unedited recording of a recent full-length keynote, ideally in front of an audience that looks like yours. Then pay attention to how they handle Q&A. Live questions expose depth faster than any slide can.

6. Questions to ask on the discovery call

• What's something about AI you were wrong about a year ago?
• Which tools are you actually using every day?
• What's a claim in the industry right now you disagree with?
• How do you tailor a talk to a specific audience?
• What do you want our attendees doing the next morning?

Red flags

• Bio leans on media logos instead of shipped work.
• Every talk sounds the same regardless of audience.
• Won't share a full-length recording.
• Can't name a specific project they built or led.
• Every prediction is either utopian or apocalyptic.