The AI facilitator
Every stakeholder gets a 1:1: 25 minutes, conversational, on their own schedule. Questions adapt to the role. Nobody performs for a room, and nobody gets talked over — every answer becomes an attributed requirement.
Stakeholder decisions, made on evidence
“Do we actually all agree on this?”
The problem
Everyone left energized. The whiteboard was full. Someone photographed it. And the facilitator — who met your industry on Monday — left with the only complete picture of the problem in their head, attached to a day rate.
The problem isn't your team. It's the format: hours of synchronous talking, filtered through one person who doesn't do the work, compressed into stickies nobody owns and a summary nobody checks.
Salby replaces the format itself. The facilitator is software — and it does its homework on all of you.
How it works
Your people spend 25 minutes each in a 1:1 with the AI facilitator — on their own time, no meeting. Then two short decision sessions. That's the entire meeting load. Every step produces a typed, attributed artifact, so the final dossier assembles itself instead of costing someone's weekend.
The AI facilitator interviews every stakeholder, 1:1, on their own time. No meeting, no filter, no loudest voice.
It merges every answer into one map — friction, value, volume — with conflicts between perspectives flagged. Your team meets once to resolve them.
It proposes where AI actually applies — every proposal traceable to who said what. Your team votes one target.
Two steps get redesigned around AI — each with an autonomy contract your people set.
Premortem, EU AI Act check, five-axis scorecard. The facilitator plays devil's advocate.
A human expert's verdict — buy, configure, or build — with measurable acceptance tests.
Go or no-go, in writing, with kill criteria. Both outcomes are wins.
What you get
Every stakeholder gets a 1:1: 25 minutes, conversational, on their own schedule. Questions adapt to the role. Nobody performs for a room, and nobody gets talked over — every answer becomes an attributed requirement.
Every interview merges into one model of the work. Where people agree, it consolidates. Where they contradict each other, it flags. Where nobody had an answer, it says so out loud.
Eight families, about eighty cards, written for the people who do the work. Each card says when it shines, when it bites, and what data it needs. Nobody has to know what a transformer is.
The facilitator proposes; your people dispose. Every selection, score, and go/no-go is a human act — and every AI intervention gets a written contract: what it may do, what it may never do, when it hands back.
A human expert delivers the verdict — buy, configure, or build — with measurable acceptance tests. The technical call is our job, not yours.
Auto-generated as the sprint runs, every claim traceable to its source. You end with a decision you can defend, not a photo of a whiteboard.
The offer
Agencies run this arc as a bespoke engagement: discovery interviews, workshops, synthesis, feasibility, roadmap. The synthesis quality depends on one person's understanding — and it leaves when they do.
Salby productizes the arc. The facilitator is software: it captures every perspective, forgets nothing, writes the dossier as it goes — and it's still there for your second sprint, and your tenth.
The Day-14 Guarantee
Every sprint terminates in a decision with kill criteria in writing. If day 14 arrives and you have neither — no charter, no reasoned no — you pay nothing, ever, and you keep every artifact the sprint produced.
Stakeholder decisions, made on evidence.