Put your leadership team in the same consequential decisions.
See where they disagree about ownership, evidence, escalation and AI authority—then turn that divergence into explicit operating rules the team can use.
- What it isA PAVNESS workshop powered by Pressure Room
- What we needNo systems access. No integration.
- Formats50 minutes · 2 hours · half day
- Where nextAI Rising, Columbus · 19–20 October 2026
What your people leave able to do
See it move
They decide once before any machine participates, then watch their own judgment shift on information they were never shown. Nobody argues with an experience they just had.
Name the owner
Out loud, in under five seconds, for a real category of decision. Most rooms discover the answer is genuinely unclear — and that the unclarity is recent.
Run a routine on Monday
THINK — one page, five questions — plus an agreed answer to what AI may do here, what reasoning stays visible, and when to escalate.
The short version. Your organization already has a written standard for who can sign a contract, approve a hire or commit capital. It probably has none for a decision a machine materially shaped. This is where a team builds that standard — by doing it, not by reading it.
Four questions, answered before you ask them
“Is this another AI keynote?” No. The room decides inside a simulation and watches what moved its own judgment. Nobody presents at you for fifty minutes.
“Do you want access to our systems?” No. No systems integration is required. No access to your production AI, systems or proprietary company data is required for the standard Pressure Room experience.
“Are you anti-AI?” No — and this matters more than the rest. We assume your people should be using AI aggressively. PAVNESS is about what humans still need to own when they do. Slowing adoption down is not the goal and never has been.
“Why not just do Copilot training?” Because this isn’t tool proficiency. Copilot training teaches what the machine can do. This trains what the human still has to do — and those are different problems with different failure modes.
What actually happens
The method is a constructed decision simulation. You make the call before the machine enters. Then the information changes, the recommendation gets more confident, and the clock gets shorter. You decide again. Only afterward is it revealed what changed — and why competent people made different calls from the same underlying problem.
The cases are built, not borrowed. Nobody has to hand over a live decision, open a system, or expose anything real to find out how their judgment behaves under pressure. That is a design choice, not a limitation.
In the two-hour and half-day formats, participants then apply that standard to the kinds of decisions they actually make — in their own words, without handing anything over.
Rather feel it than read about it? The first two minutes are on this site — one decision, ninety seconds, nothing collected.
Who this is for
The people who sign, and the people who brief them.
- CHROs and heads of L&D whose senior population now reviews work a machine produced, without an agreed standard for what “reviewed” means.
- Executive teams and their direct reports in functions where a wrong call is expensive and slow to surface — risk, finance, legal, clinical, engineering, underwriting.
- Leaders running an AI rollout who have solved for adoption and not yet for accountability.
- Boards and partner groups who will be asked, at some point, who approved something.
Who it is not for. Teams looking for AI tool training — we don’t teach prompting, and others do it well. Anyone wanting a motivational keynote; the room does the work here. And anyone hoping for a document certifying their AI use as sound, which is not a thing we produce or believe can be produced.
The three formats
They differ by how deep the change goes, not by how long they run.
All three establish the same thing. In practice a standard means four things: what AI may do here, who owns the call, what reasoning must remain visible, and when escalation is warranted.
USD 5,000
Make the invisible visible.
Participants experience how AI changes judgment, and meet the human expectations that go with it. They leave able to say: I know what my job is when AI participates.
USD 5,000 for the first three sessions booked before 31 October 2026. USD 7,500 after that.
USD 12,000
Make behavior observable.
People decide, get pressured, defend the call, and reconcile where two of them decided the same case differently. By the end the team has agreed what good human review and ownership means for one recurring category of AI-assisted decisions — without giving PAVNESS the underlying work.
USD 17,000
Make expectations shared.
A whole function gets the common experience; subgroups apply it to different categories of work; everyone reconvenes on the same minimum expectations. They leave able to say: we now share a language and a minimum operating expectation across different kinds of decisions.
The founding rate is a date and a count, not a negotiation: the first three 50-minute sessions booked before 31 October 2026. Booking date is what counts — a session booked in October and delivered in November is still at the founding rate. It exists because there is not yet a reel or a public client list, not because the session is worth less. Travel, where a session is delivered in person outside a scheduled trip, is billed at cost and agreed in writing beforehand.
How the half day works
A seventy-person department does not share a decision type, so one generic case means most of the room sits through somebody else’s job.
Everyone begins together for the common briefing and the first simulation. The group then splits into three or four decision lanes chosen in advance by the sponsor — for example strategy and investment, analytics and finance, commercial, product and operations. Each lane runs a case built for the calls that lane actually makes, and translates the standard into one or two of its own recurring decision types. Then everyone reconvenes, and the patterns that showed up across every lane are drawn out together.
That closing is what makes it a shared standard rather than four local ones.
What it does not do. A half day starts this; it does not complete organizational adoption, and we will not tell you otherwise. What a department leaves with is a common language and a minimum operating expectation, plus each lane’s own first application. Embedding it into recurring work is a separate conversation, and one we would rather have after you have seen the room.
What we need from you
No systems integration is required. No access to your production AI, systems or proprietary company data is required for the standard Pressure Room experience.
For the half day, the only input needed is the sponsor’s choice of lanes — which kinds of decisions the department actually makes. Not the decisions themselves.
What this is not
It does not review, approve or validate any AI system, model or vendor, and it produces no finding about whether a given system is fit for a given purpose.
It is not legal, regulatory or compliance advice. Whether a specific organization is covered by any rule, and what that rule requires of it, is a determination for its counsel.
It does not measure any individual’s capability, and nothing said in the room is reported back to the organization as an assessment of a named person. The work describes how decisions are being made, on the day it is run, in the scenarios it tests. Discovering that a decision cannot currently be defended is a useful outcome, not a failure.
Who runs the room
Pav Lertjitbanjong is a decision scientist with 20+ years inside Fortune 500 and global organizations, most recently leading strategic analytics — BBA in Decision Science, Kellogg MBA, Stanford-trained in AI-driven leadership. She designed the method and facilitates every PAVNESS session personally.
This is not a licensed course handed to a junior facilitator. The person you meet on the scope call is the person who runs the room. Full background.
What you leave with
Every session ends with things you can hold, not a feeling. The formats below are samples — the content of each comes from your own session.
The THINK review card — the routine a reviewer runs before signing an AI-assisted decision, on one page.
The four-question team standard — what AI may do here, who owns the call, what reasoning stays visible, when to escalate — as your team agreed it, out loud.
The session summary, in counts — how many reversed, how many could name the owner, how many escalated. Never an assessment of a named individual.
The Monday-morning prompt — the one question to put to the people who review AI-assisted work, and what the answers tell you.
Where you can see it run
AI Rising, Columbus — 19 and 20 October 2026.
A TEDx stage — 1 November 2026, under the title A Machine Will Never Be Brave.
The sessions are run by a decision scientist with more than twenty years inside Fortune 500 and global organizations, most recently leading strategic analytics, who came through more than fifteen rounds of restructuring without being cut.
AI can make the call. It can't take the fall.
Tell us where AI has entered a decision that matters.
Useful if you have them: the function, roughly how many people review AI-assisted work, and the date or window you have in mind. None of it is required.
This goes straight to pav@pavness.com and you'll receive a response within 48 hours. Prefer the calendar? Use the scope-call button above.
Fifteen minutes is enough to tell.
Fifteen minutes, one calendar link, no deck and no follow-up sequence. Bring the function, roughly how many people review AI-assisted work, and any date you have in mind. If it isn't a fit, we'll say so on the call rather than send you a proposal — and either way you'll leave with the question worth asking internally.