Two jurisdictions. One inject at a time. Half of it is a guessing game on a hidden board, and half of it is the part that matters.
Two jurisdictions, seven sites each. Every round one team plays the adversary, draws an inject and picks a target. The other team writes how it responds. Then the inject lands, and any site under it goes offline. Whoever has knocked out more of the other jurisdiction when the rounds run out wins.
There are no dice. A hit is a hit. The quality of a response is judged in the debrief, not by the game.
The injects run in a fixed order, the one set in the scenario workbook. Every inject carries an AI assignment: a team has to build part of its attack with an AI tool before it can commit.
In simultaneous play there is one inject a round and both teams work it at once, first as the adversary and then as the office. Both boards face the same footprint, so neither side gets an easier deck than the other.
One inject a round, the same one on both screens. Both teams write an attack from it on the first clock. The attacks swap, and both teams write a response on the second clock. Then the round resolves on both boards at once.
A screen cannot move to the next phase until the other team's packet has landed. A slow team holds the round rather than pushing the two screens out of step.
Nobody is cut off when a clock hits zero. Whatever is on the page is committed and the round moves. If no target was picked, the shot goes somewhere at random and the log says so.
A round has two halves. Team 1 attacks in the first, Team 2 in the second. Each team defends once per round.
Draw an inject. It carries an operational concern and a communications concern. Write how your team runs it, then pick a target cell. The inject's footprint fills from there and anything running off the grid is wasted.
You learn hit or miss for every cell you fired at, so the board rewards deduction. A hit next to a miss tells you which way the site runs.
You see the inject. You do not see the target. Write your response before the inject resolves. Then it lands, and any of your sites under the footprint go offline for the rest of the exercise.
Your response does not change what the inject does. It is the record the debrief works from, so write it as though you were briefing your county administrator.
Four injects hit the whole jurisdiction. There is no target and no footprint. Two of your cells go offline and you choose which two, which is the only real decision the defending team gets. Play does not resume until you write a continuity statement.
In two-screen mode the file writes the code for you and spells it phonetically. An attack code looks like R3-12-D3. A result code looks like R3-H101, meaning the first and third cells you fired at were hits and the middle one was a miss.
Bottom right, open on every screen. Everything either team writes or attaches lands there and stays, so the two teams are in contact the whole way through rather than only at the handoff. With an automatic relay switched on, text, links and small files cross by themselves.
Large files will not fit. Upload one somewhere the other team can reach and paste the link instead.
Every screen runs its own clock, with its own start, pause, add a minute, take a minute and reset. Nothing about the clock travels between screens.
They stay together anyway, because the phases do. When both attacks have been swapped, every screen enters the response phase at the same moment and starts its own response clock, so the numbers match within a second without anything being sent.
The one thing that does not carry is a pause. If the moderator stops the clock, say so out loud and everyone presses pause. Same for adding time. That is the trade for not hammering the relay all afternoon.
The last box in every attack is the thing you made with AI. Each inject names a specific task and offers a starter prompt if you want one, and the attack will not commit without it.
Push past another paragraph of text. An image, a screenshot, a voice memo, a mailer, a letterhead, a spoofed alert, a fake headline. The exercise is worth far more when the other team has to look at a convincing fake object and decide what to do about it, and that is the honest version of what an office is up against. Attach it in the relay panel so they can actually see it.
Do not generate video. Even a short clip takes longer to render than the round allows, and a team waiting on it loses the part of the round that matters. If the inject calls for something on camera, write the script and make the thumbnail instead.
Start the AI the moment the inject appears. Put one person on the tool while the rest of the table reads the card and argues about the other four boxes. Teams that run it in sequence finish late.
Fictional county, fictional names, no real people or agencies. Label whatever you make as exercise material. None of it gets posted, and it all goes after the debrief.
When the rounds run out, the jurisdiction with fewer sites offline wins. Tie goes to the one with fewer individual cells down, then to a sudden death round.
Every site counts the same. A one-cell polling place is as expensive to lose as the four-cell operations center, which means the operations center is the hardest site to knock out and the polling places are the cheapest.
The exercise saves itself as you go. Close the tab, close the laptop, break for lunch, and reopening this page offers to resume where you stopped. Boards, damage, the log and the relay thread all come back.
Break between rounds, not in the middle of one. Finish the result screen first, so neither side is waiting on a packet from the other.
It comes back on the same browser on the same machine. A different laptop, a different browser or a private window has no memory of it. Do not clear browsing data over the break.
Attached files are the one thing not kept. The name and any link stay in the thread, but the file itself is gone, so paste links rather than attaching if a break is coming.
Export the after-action log at the end anyway. That is the copy you keep.
Put all seven sites on your grid. The other team never sees this. Multi-cell sites must sit in an unbroken line, across or down. Spread them out and single hits cost you a whole site. Cluster them and one footprint can take two.
This inject hits everywhere at once. There is nothing to aim at. Two of their cells go offline and they choose which two, so you will not learn anything about their map.
This one hits everywhere at once, on both sides. There is nothing to aim. You will each choose which two of your own cells go offline.
This clock is yours alone. It runs in step with the teams because the phases start together, but a pause here does not reach them. Call it out loud and have both rooms press pause.
The clock is advisory either way. Nobody is cut off when it hits zero. Whatever is on the page is committed and the round moves.
Everything both teams write and attach arrives in the relay panel, bottom right. That is the debrief record, so there is no need to stop for one during play.
Every inject, every stated method and every stated response, in order. This is the exercise. The board was only the thing that made it happen in a particular order.
A tabletop exercise is designed to find the gaps in your planning. That is the whole product. Finding one is the exercise working, not you failing.
This is a safe space. Nothing written here is an evaluation of anyone, and nothing leaves the room except the action items you decide to take with you.
The goal is to walk out with renewed confidence and a short list of things to go fix.
There are two things going on here and they do not share a scoreboard.
Each jurisdiction hides seven sites on a six-by-six grid. Every round you aim a footprint at the other team's grid without knowing where anything is, and they aim one at yours. Cells get hit, sites go offline, and whoever has knocked out more of the other jurisdiction when the rounds run out wins that part.
This half is luck and deduction. It is a hunt across a hidden map. It is not a measure of how well your office is run, and a site going offline is not a prediction that it would.
The other half is what you write when an inject lands on you. Who owns it, who gets called, what you do in the first hour, what policy it rests on, what goes out and to whom, and what you needed that you did not have. That is the part that gets read back in the retro, and it is the part that turns into action items.
It is not scored during play at all. Nothing you write can save a cell, and nothing you write can lose one.
You can win one and lose the other, and that happens often. A team can get lucky on the board and write thin responses all afternoon. A team can get picked apart on the grid and still walk out with the better plan, which is the outcome that actually matters on Election Day. When the final screen names a winner, it is naming the winner of the board. The retro settles the rest.
You will find things the scenario does not tell you. Make a reasonable assumption, say out loud what you assumed, and keep going. Arguing with the premise burns the clock and teaches you nothing. If an assumption changes your answer, that is worth writing in the gap box.
Answer in your real role, with the authority you actually have and the resources you actually have. If the person who would normally make this call is not in the room, say who that is and what you would do to reach them. That gap is a finding.
Build your scenario around systems, facilities, timing and information. Do not solve a problem by having a colleague, a supervisor or anyone else in the room be harmed, removed or otherwise unavailable. Staffing continuity comes up in its own inject, and it is handled there.
Do not generate a synthetic voice or face of a named real person.
The response boxes ask who owns it and who else has to know because that is the part most plans are thin on. Talk it through as a group before anyone types.
The AI piece produces material that is supposed to look convincing. Label it as exercise material, keep it in the room, and delete it after the retro.
Ten minutes a phase, so about 23 minutes a round with the result screen. Round 1 always runs long while everyone learns the boxes, so expect to add time or treat it as practice.
The last box in every attack is the thing you made with AI. Each inject names a task and offers a starter prompt, and the attack will not commit without it.
Push past another paragraph of text. An image, a screenshot, a voice memo, a mailer, a letterhead, a spoofed alert, a fake headline. The exercise is worth far more when the other team has to look at a convincing fake object and decide what to do about it.
Even a short clip takes longer to render than the round allows, and the team waiting on it loses the part of the round that actually matters. If an inject calls for something on camera, write the script and make the thumbnail instead.
Put one person on the tool the moment the inject appears, while the rest of the table reads the card and works through the other four boxes. Teams that do it in sequence run out of clock.
Attach whatever you make in the relay panel so the other team can see it. Label everything as exercise material. No synthetic voice or face of a named real person.
Nobody is cut off when a clock runs out. Whatever is on the page is committed and the round moves. Ask the moderator for more time if you need it.
Aiming is the first thing you do each round. The target window opens by itself, you click a cell on their hidden grid, and the footprint fills from there. Anything that runs off the edge is wasted, so the shape of the inject matters. Once it is locked, the target sits at the top of the writing screen and you can change it until you commit.
A cell that has been hit turns red and stays down for the rest of the exercise. A site is offline once every one of its cells is down. A miss leaves a grey dot, which tells you something too.
You will hear about it. There is a clip for a hit and three for a sunk site, and they play on the screen whose board took the damage. Switch them off at setup for a quiet room or a recorded session.
Leave the laptop open and on power over lunch, and use the same one after. If it does get closed, the exercise saves itself. Reopening this page offers to pick up where you stopped, with the boards, the damage, the log and the whole relay thread intact. The clock comes back paused, so press start when the room is ready.
Three things to get right:
Export the after-action log before you go anyway. It costs nothing and it is the copy you actually keep.
Open on every screen, bottom right. Everything either team writes or attaches lands there and stays, so you can send a file, ask a question or pass a link at any point. That running record is the retro, which is why there is no separate debrief break during play.
Anything much over a couple of hundred kilobytes is too large to send through it. Upload it somewhere and paste the link.