We've done itbackwards

It starts small. The CEO, the COO and the CFO agree the executive team should hear it, so a meeting goes on the calendar a week out. Two people turn out to be traveling. Staff starts building slides, and nobody has the full picture, so week one isn't a decision. It's a hunt for numbers.

Then the rest of the company gets pulled in, and teams start meeting to get ready for meetings. The team presents. Nobody signs. Someone suggests two more weeks for comments.

At the third meeting they vote while the two who missed the first ones are still reading last week's deck. Legal and HR were never in the room, so they hear about it afterward, and everyone meets again.

Six to ten weeks in, the teams are guessing at their jobs from a decision they only half heard. The fourteen people whose roles change still don't know what happens to them. West wait times haven't moved. The customer who was already waiting too long didn't stay for the third meeting.

Now run it in the other order.

On Monday the COO asks the same question as one sentence someone can sign: do we combine East and West support into one organization? The CEO doesn't book a meeting. The CEO says push it through the engine and see how the executive team, the board and the innovation team respond.

This is the part people get wrong. The engine doesn't hand down a signed answer that everyone nods along to. It does the prep that used to take ten weeks, before anyone walks into the room. A day later every executive and every board member has their own page, and the first meeting is about the real disagreements instead of who read the deck. The weeks that used to go to the vote now go to the work.

Start with a sentence someone can sign.

"Support strategy" is a topic, and topics make decks. "Do we combine East and West support?" is a question someone can put their name under, and it tells you who needs to weigh in and what a finished answer looks like.

The engine then builds the rooms the question needs, in the order it needs them. For East-West that's the executive team, the board and an innovation team. Each room seats up to ten members, and each one brings its own data, its own limits and its own vote.

Three rounds, each with one job.

The rounds aren't more talking. The first gets the facts only one seat has onto the table, the second is a vote where every seat names its conditions, and the third is the final vote on the reworked answer. Research on groups points the same way. How a group closes a decision matters more than how long it argues, and structured voting beats open-ended consensus-seeking (Kaesberg et al., ACL 2025), and AI models drift toward whatever the person asking already believes (Sharma et al., ICLR 2024), so each round is short and has a single job. And when drift detection sees a room agreeing too fast with no new facts, the engine makes at least two seats argue the other side before the final vote.

Real seats, not one voice in ten hats.

A finance seat isn't a chatbot told to act like the CFO. Ask one model the same question ten ways and the answers drift toward agreement, usually with made-up support sitting right next to them, because agreeing is the easy path.

A real seat brings its own data, the limits of its job and a record of what it argued last time, so finance fights for cash while operations fights for what breaks when one office becomes the only office. Each seat pulls from the company's own numbers, fresh market research, the risks and what the team can actually carry. If they all sound like the same reasonable person by lunch, you don't have a debate. You have a press release.

Multi-model, with drift detection.

The engine is multi-model, and every seat gets its input from its own independent agent, so one model's blind spots don't become the whole room's. Drift detection watches every room for the slide toward easy agreement: unanimous votes, confidence bunching together, dissent going quiet. When it sees that, the engine forces the argument back onto the table.

Fourteen facilitators, and none of them votes.

Between rounds, fourteen facilitators check the work: whether the room brought up facts only one seat had or just repeated the shared slides, whether anyone changed a vote without a reason, and whether every claim traces back to where it came from. Then they write it all down: the options, who pushed back on what, the vote by seat and what's still missing.

A facilitator can't carry its own version of the story, or you've just built a second decision-maker. The record is locked, and every step, vote and source is open to every member of the team, so nobody can quietly steer it.

Every executive gets their own page.

A day after the question goes in, every executive and every board member has their own page. Take the CMO. It shows what the rooms decided, the facts the CMO should be looking at, how every other seat voted and why, and every risk on the table, with everyone else's answers one click away. Now the CMO can agree or disagree with all of it.

Nobody says "I don't agree with the AI."

The pushback gets specific instead. "I'm a no for now. Section 5, Exhibit A. The customer data is 8 percent off what we're seeing, and that changes the numbers. My team is checking how much it matters, and I'll have it to you tomorrow." Legal's two requirements were already on the page, with the board room's answer: on this timeline, with these blockers, if you all agree, move.

Behind those pages sits one shared file, and it is the thing that gets signed. This is the East-West file on the morning the COO signs it.

Operations lost the vote and won the gate.

Look at what happened to the losing side. Operations voted against, and its condition, a tested failover before cutover, became the gate the whole plan has to pass. That's the difference between losing an argument and being ignored, because the concern lives in the file with an owner, and if the failover test fails, the decision reopens on its own.

Meeting one. Meeting two at the most.

Everyone walks in already knowing the room, so the argument is about 8 percent, not about who read the deck. Six to ten weeks becomes two, and the staff hours on the decision drop by about 86 percent. In less than a day, for less than one executive meeting costs, any question can run through it. After the signature, any meeting that isn't doing the work or dealing with a reopen fact, one of the named facts that would force a fresh look, is a leak.

Build the rooms you don't have.

If you don't have an innovation team or an investor forum, build one. Copy your executive team, add the roles you don't have yet, or build a board to test the call against. As the CEO you get to hold that board to account, and you walk in knowing three quarters of the room, often with half of it handled before you sit down. The same file moves from room to room without anyone rewriting it for the next audience.

Give someone what they need up front instead of making them hunt for it, and they get faster. That isn't our claim. It's what controlled studies found when they split people into groups with and without AI and measured the actual work.

The people who asked for it haven't used it.

Leaders spent two years asking every team to adopt AI, and three in four knowledge workers now use it at work. Yet only about one executive in six uses it in strategic decisions, according to Capgemini's 2026 survey of 500 of them. The table that demanded it is the one table still running on slides.

That's why the gains don't show up in earnings. Eighty percent of the people McKinsey surveyed say AI made them personally more productive, but only 6 percent work at companies earning 5 percent or more of their profit from it. The speed is real, it just runs into a room that still decides the old way.

The room was slow before AI showed up.

This isn't new. Bain found in 2006 that only about 15 percent of the 365 companies it studied made and carried out decisions well. In 2014, Mankins, Brahm and Caimi traced everything one weekly executive meeting at one large company set in motion over a year. It added up to 300,000 hours of people's time, and the meeting itself was 7,000 of them.

You're already paying for it.

The meeting isn't the hour on the calendar. The heaviest calendars get interrupted every two minutes during core hours, and PowerPoint edits jump 122 percent in the last ten minutes before a meeting (Microsoft, 2025). The managers staging all of it are down to 22 percent engaged worldwide (Gallup, 2026). Meanwhile 81 percent of organizations in McKinsey's 2026 survey of more than 10,000 executives report no real bottom-line gain from AI.

Where the 250 hours come from.

Counted out for East-West, meeting by meeting, the old way looks like this, and it's deliberately on the low side.

The argument survives a worse count.

Even if the old way took only 100 hours and the file took 50, the file would still cut the time in half and turn ten weeks into two. The case does not rest on the exact hours. It rests on the order.

$23,750 the old way. $3,230 in the file.

Put public pay rates on those hours. Bureau of Labor Statistics data puts an operations manager at about $75 an hour with benefits and a financial manager at about $119, so we use $95 for the East-West seats, rounded down from the $97 midpoint to keep every figure a floor. Equity and bonuses aren't in it, which is why real numbers sit above ours.

40 signed decisions a year instead of 12.

One decision is nice, but a year of them is the point. With the executive team signing twice a month and the board four times a quarter, the same calendar holds about 40 signed decisions instead of the 12 the old way manages in a good year. Twelve decisions at 250 hours is about 3,000 hours. Forty at 34 is about 1,360, under half of it.

But the point isn't the hours you save, it's what they open up. The teams that spent weeks building slides now have three times as many signed decisions to work on, so the whole pipe finally runs. That's the premise this paper started with. Take the blocker out of the room at the top, and all the AI speed already sitting at the desk finally has somewhere to go, and it starts to build on itself across the company.

At enterprise size the pile around one weekly executive meeting costs about $22.5 million a year at the same rates, which is 300,000 hours at $75. That's the size of the bill, not the savings, but even a four-week slice of it is $1.7 million.

Nearly half the lost value is gone before the work starts.

Deciding faster matters most at the front. McKinsey's survey of 1,034 people who'd been through a big company change found that nearly half the value that gets lost disappears while the company is still arguing about the target, before any work begins.

Lose in public, not in the meeting.

Today the miss is inside the building: nobody owned the answer, the wait times stayed where they were and a customer left while everyone was still in slides. Decide first and whatever miss is left belongs to the market, where you combined the offices and service still slipped, or the customer just said no. That company lost in public and can learn from it, while the other one never decided at all.

Size changes where you start.

At twenty people the founder is the room, the board and the only memory of why the last call went the way it did, so the file becomes the company's memory. At two hundred to two thousand, start with the executive team, where the pile is heaviest and the hours come back first. Past five thousand, don't try to flip the whole enterprise at once, flip the questions a unit or a region truly owns.

What this paper claims, and what it doesn't.

We don't have a named company that ran this and reported the result, and we're not pretending to. The claim is narrower: the market has already measured that people and teams with the information up front work faster, and the executive table is the one place that hasn't been tried. East-West is a made-up company, our math is a floor built on public pay rates, and Level9 sells software that runs this method, which is why every number here is labeled with where it came from.