Protocols for Business · Kickoff

Kickoff · Monday 2 November 2026

Protocols for Business

A year of reading together, learning to see the rules a business really runs on, and putting that to work as agents join in.

Today: introductions · these slides, about 20 minutes · our first reading together · questions

TL;DR

Let's get good at seeing protocols.

  • WhyAgents are making the formal and informal rules of business matter, and very few people can see them well yet.
  • WhatEvery other Monday, 15:30 UTC on Discord, 26 sessions in all. One reading each time, read together. Recorded.
  • You getA skill, a case study from your own work, and a small group of friends who care about this too.
  • We askCome when you can, post what you notice, and carry one case if you have one. That's it.

Where we're coming from

We've spent a year and a half reading wide.

What it taught us

Five lessons shaped this year's plan.

  • Some tensions are handled in the moment; others are built into structure.→Which rules should be hard, and which can stay soft?
  • Good protocols produce non-events, and nobody counts what didn't happen.→What can incident reports show us about everything that usually goes right?
  • Protocols show up when tempo changes: Slack in the office, Boom's R&D against its supply chain.→What shows up when agents speed everything up?
  • Standards spread through incentives and narrow scope, then lock in.→What do our systems emit, and how do we keep rules open to change?
  • Working with agents is management; the scarce work is gathering context.→Who gathers that context where you work, and could it be you?

What's different now

Agents won't ask for manager approval.

A small team can now run thousands of agents. An approval step can't keep up, so work flows around it. What holds is what's built in: into every agent, and into the ground they work on.

Left, an approval queue: agents' work piles up at one approval step and flows around it. Right, hard protocols: agents move freely along guides, and guardrails keep them out of off-limits areas.

This year's question

How do organizations coordinate through what they emit, and what changes when agents read and write those signals too?

  • ThemesSkillDate
  • I. What agents are, in practiceReading what agents can see, do and spendNov–Dec 2026
  • II. Natural coordinationSpotting coordination without a managerJan–Feb 2027
  • III. EmissionsMapping what systems publishFeb–Apr 2027
  • IV. Seeing hardness in incidentsReading incidents for what held or brokeMay–Jun 2027
  • V. Engineering hardnessChoosing what to harden, and howJul–Sep 2027
  • VI. Operational livenessKeeping rules open to changeOct–Nov 2027

What you'll build

By next autumn, you'll see protocols everywhere.

  • SeeSpot the rules a business runs on, written and unwritten.
  • ReadRead incidents and agent traces for where rules held or broke.
  • DesignDecide what to harden, what to keep soft, and try it.
  • WriteDescribe a messy system so the people inside it recognise it.

Your case study

Bring one case: a client, or your day job.

Pick an organization you can see into, and one data operation AI is changing. Carry it for about a year: choose, brief, intervene, gather evidence, publish.

That's how last year's cases started: water rate data, construction bids from PDF drawings, a brand kit agents build from, and how people actually co-work with AI.

Put it to work

If you want, turn it into paid work.

Protocol Vision Advisory is an independent network of consultants and teams, built on this group's open research. Any member can become a network consultant; nobody has to.

  • Work comes inA coordinator takes each inquiry and picks a lead by fit. The lead builds the team.
  • One ruleA field log from day one, shared with the client.

Next steps

Let's get started!