Your attendee used AI before they finished their first coffee. They drafted something with it, asked it to summarise a report they were never going to read, and argued with it about a fact. By the time they walk into your venue they have specific, hard-won opinions about where these tools work and where they fall over. A session that opens by explaining what a large language model is has already lost that room.
That cuts two ways for you. The content you programme has to clear a bar that moved. And the tools available to plan the event can now absorb a real chunk of the busywork. Both shifts are genuine. Both are easy to get wrong, in opposite directions.
Your audience is past the explainer
The fastest way to lose a 2026 room is to book someone who still treats AI as a novelty. Assume every person in the seats has used these tools in their own job and formed a view. They are not there for the tour. They want to know what to do on Monday, in their function, with the budget and the compliance constraints they actually have.
Four things separate a session that lands from one that empties the room.
Sector specificity. "AI is transforming everything" is a non-statement. A talk about how a regional hospital network handled clinical documentation, or how a mid-market logistics firm rebuilt routing and what it cost, gives people something to copy or argue with. When you brief a speaker, push them off the generic deck toward the case that maps to your audience.
Honesty about failure. The most credible voices right now are the ones willing to describe a deployment that came in over budget and under-delivered. Everyone in your room has lived the messy version. A flawless success story reads as marketing. A post-mortem reads as useful.
Workforce awareness. A large share of your audience is quietly worried about their own role. Programming that ignores this is tone-deaf. Programming that only stokes it is exhausting. The talks that work are specific about what changes for a given job function and which skills hold their value.
Range. A lineup drawn from one network produces one take, repeated. Spread across sector, company size, seniority, geography and lived experience and you widen the questions that get asked on stage. That is a quality argument before it is anything else.
Choose the format for the job, not for the calendar
Most agendas overuse whichever format is easiest to book. Each one does a different job.
| Format | Best for | Fails when | Typical slot |
|---|---|---|---|
| Keynote | Setting the frame, one sharp argument | It becomes a long, unstructured ramble | 30–45 min |
| Panel | Surfacing real disagreement across sectors | Four people agree politely | 40–50 min |
| Workshop | Hands-on skill people take home | It gets watered down into a lecture | 60–90 min |
| Fireside | The candid story nobody tells from a lectern | The host reads pre-agreed questions | 30 min |
| Emcee-led segment | Holding the day together between blocks | Treated as an afterthought, booked late | Throughout |
A workable default for a single day: one keynote to open, two or three panels, one or two workshops running in parallel, and a fireside to close on a human note. Practitioner-heavy crowds want more workshop time. Executive audiences want fewer slides and more fireside.
Panels deserve particular attention, because they are the format most often booked and least often designed. If you are running more than one, the composition and the moderator matter more than the names. We keep a separate track for panels and moderators, and the mechanics are worth reading in full in the guide to running a great AI panel before you lock a cast.
Where AI helps the planning work, and where it does not
The planning side is where the tools genuinely earn their keep, as long as you are honest about the boundary.
They are good at writing and structuring. First-draft session descriptions in your event's voice. Speaker one-pagers assembled from public material. Turning a messy brainstorm into a candidate agenda with time blocks. Drafting the sponsor prospectus, the reminder emails, the run-of-show template. Clustering last year's feedback into themes instead of you reading 400 free-text comments. Producing eight title options for a session so you can pick one rather than stare at a blank line.
They are bad at judgement. Whether a speaker is actually credible. Whether a room will tolerate a particular argument. Whether two panelists will spark or just clash. Which sponsor will be difficult. What your audience is quietly anxious about this year. Those calls are why you have the job.
One warning matters more here than anywhere else. If you use an assistant to research speakers, verify everything it hands you. These tools state invented credentials with total confidence: a conference that does not exist, a book that was never written, a title the person never held. We check every speaker individually before a booking goes through, for exactly this reason. Anyone can claim to be an AI expert, and now the software will claim it on their behalf. Ask for proof either way — what they shipped in the last twelve months, what went wrong on it, and who from a recent audience will take your call. The full checklist is in our step-by-step booking guide.
A calendar that works backwards from the date
Speakers are the piece organisers most often start too late and finish too early. Here is a workable shape for a mid-size event.
| When | What has to happen |
|---|---|
| T-6 months | Theme and audience defined. Rough format mix decided. Budget split between headline and supporting slots. |
| T-5 months | Start sourcing. Brief on outcome and audience, not on names. Build wide before narrowing. |
| T-4 months | Shortlist reviewed. Fee conversations with real numbers, not "on request". Headline slot confirmed. |
| T-3 months | Supporting speakers, panelists and moderator confirmed. Contracts signed, including recording rights. |
| T-8 weeks | Agenda published. Session titles and descriptions final. Promotion starts. |
| T-4 weeks | Individual speaker briefs sent: audience, what runs before and after their slot, what to avoid. |
| T-2 weeks | Panel pre-call. Tech check for every remote speaker. Run-of-show circulated. |
| T-1 week | Final headcounts, AV walkthrough, a named contingency for each speaker. |
| Week of | Confirm arrivals, re-send timings, brief the emcee on every change. |
| T+1 week | Feedback collected while it is fresh. Recordings edited and released if the rights allow. |
Two notes on that table. The T-5 sourcing window is longer than it needs to be if you brief well. A written brief covering audience, budget band and outcome is enough to produce real names, and we return a shortlist within 24 hours with no account and no call required, so you can start a brief at any point in that calendar and still be early. Second, the T-4 week speaker brief is the highest-leverage hour in the whole schedule. A speaker who knows what the session before them covered will not repeat it.
Build the remote and hybrid plan in, not on
If any part of your event is streamed, or any speaker is joining by video, decide that in month one rather than month five. Remote sessions have their own speaker requirements, their own production stack and their own timing rules. Retrofitting them onto an in-person agenda is where hybrid events fall apart.
A speaker who is magnetic in a ballroom can be flat on a webcam. It is a different skill, not a lesser version of the same one. We keep virtual keynotes as a distinct format for that reason, and the production and interaction mechanics are covered in the virtual keynote guide.
The same discipline applies to time zones. If you want a speaker from outside your region — and you should — build the slot around their clock before you publish the agenda, instead of asking them to present at 3 a.m. and then wondering why the energy was low.
Contingency is a planning artefact, not a panic response
Speakers cancel. Flights are missed, people get sick, a company drops a communications freeze on an executive two weeks out. Plan for it as a normal event rather than a disaster.
For each confirmed speaker, write down three things: a named backup or an alternative use of the slot, who makes the call and by when, and what happens to the money. Get cancellation terms into the contract on both sides at signature, not into an email thread the week of. On our side, if a speaker cancels we find a replacement within 72 hours or you pay nothing, and our 17.5% commission is printed on every quote so you can see what the speaker receives and what we receive. Whoever you book through, ask for that in writing.
Then do the boring part. Name the owner for each risk: AV failure, a no-show, a session running fifteen minutes long, a panelist who will not stop pitching their product. Each one needs a person and a first move, written somewhere your team can read at 8 a.m. on the day. The events that stay calm under pressure are not the ones with fewer problems. They are the ones where somebody already decided who handles what.