Founding workshop / NYC

Agents for
actual work.

Build the system. Keep the judgment.

Bring one recurring task. In one hands-on day, turn it into a supervised AI workflow your team can test on Monday.

Two collaborators map a real workflow across paper artifacts, tools, checkpoints, and one resolved lime path
S/C–01Possibility → judgment → controlled practice
DateSat / Aug 01
Time10:00–16:30
PlaceManhattan / NYC
Room24 seats
One dayOne real workflowInstructions + limitsFive test casesHuman takeoverSeven-day clinic

01 / THE OUTPUT

Your team leaves with a system, not a folder of prompts.

Most AI classes stop at an impressive demo. We follow the work all the way through permissions, tests, review, recovery, and handoff.

A.01
A charter worksheet organized into mission, input, owner, and stop-condition zones

Agent charter

Mission, inputs, tools, permissions, owner, and stop conditions.

A.02
A modular workflow harness with context, tool, permission, and review components connected by a lime route

Portable harness

Reusable context, instructions, examples, and structured output.

A.03
Five test-case cards arranged for ordinary, edge, failure, and human-review checks

Evaluation pack

Five ordinary, edge, and failure cases with a review rubric.

A.04
Three work positions converging on a single handoff plan with a clear responsibility path

Monday plan

A 14-day implementation runbook with a named human in charge.

02 / WORKBENCHES

Different work. One rigorous operating system.

Selected workbenchREADY

01 / FINANCE + OPS

Move faster. Keep the source.

Build research, spreadsheet review, reporting, monitoring, or knowledge workflows with reconciliation and human sign-off intact.

  • Source-grounded research
  • Workbook QA
  • Recurring reporting
A finance workflow workbench with ledger sheets, reconciliation marks, evidence checks, and a resolved path
01 / FINANCE + OPS / CONTROLLED PRACTICE

03 / THE METHOD

From possibility to controlled practice.

An agent completes a multi-step task. A harness is the operating system around it: instructions, context, tools, permissions, tests, limits, logs, and human checkpoints. You build both.

A supervised workflow test with human review checkpoints, evidence markers, and a deliberate stop condition
S/C–11Human review is part of the system
  1. 01
    INPUT

    Frame the work

    Name the mission, source of truth, failure cost, owner, and stop condition.

  2. 02
    CONNECT

    Assemble the harness

    Give the agent instructions, context, tools, permissions, limits, and a structured output.

  3. 03
    TEST

    Break it on purpose

    Run ordinary, edge, and adversarial cases. Measure quality, cost, speed, and intervention.

  4. 04
    HANDOFF

    Make the handoff

    Decide what can be tested Monday, what needs review, and what should not ship.

04 / THE DAY

A working session, paced like one.

Three participants actively testing and reviewing one shared workflow around a studio table
S/C–12 / THE WORKING DAYNo presentation theater. The work stays on the table.
Working agreementSuccess, safety, responsibility
Agent vs. demoA full path, including failure
Map the workMission, inputs, owners, stop conditions
Build the harnessContext, tools, constraints, review
Working lunchShort sector demonstrations
Build sprint IA working path through the real task
Break it on purposeEdge cases and a peer evaluation
Build sprint IIRecovery, escalation, repeatability
Working demosTask, failure, control, Monday test
REVIEW REQUIRED

SAFE MATERIALS / REAL JUDGMENT

Bring the workflow. Leave confidential data at work.

Use public, synthetic, redacted, or organization-approved material. Do not bring confidential client, employee, financial, healthcare, or unreleased project information. Examples are educational—not financial, legal, architectural, investment, or compliance advice.

05 / FOUNDING COHORT

Bring the people who own the handoffs.

Names for pair and team seats can be supplied after payment. Coffee, a simple working lunch, templates, and the seven-day group clinic are included.

Solo seat

$495

$495 / seat

  • 1 workshop seat
  • 1 complete deliverable pack
  • 7-day implementation clinic
  • Names can be completed later

Pair pass

$900

$450 / seat

  • 2 workshop seats
  • 2 complete deliverable packs
  • 7-day implementation clinic
  • Names can be completed later

Full refund until 72 hours before the workshop. Substitutions accepted until start time. If the organizer cancels or cannot confirm the room, choose an immediate full refund or transfer.

06 / FIELD NOTES

Before you reserve.

Do I need to code?

No. The workshop is designed for operators, analysts, project leads, producers, designers, architects, finance professionals, creative technologists, and managers. Bring a laptop and a task you understand.

Which tools do we use?

Signal & Craft is tool-neutral. We use currently available AI tools to teach durable practice: context, delegation, evaluation, permissions, human review, and recovery. No paid AI subscription is required.

Where exactly is it?

In an accessible Manhattan studio. Exact arrival details are sent to registered participants after the room is confirmed. We do not publish attendee locations in advance.

Can my firm book a private room?

Yes. Private team sessions adapt the same build-test-handoff method to your policies and representative work. Start with a three-seat pass and select “private team” in the intake.

SATURDAY / AUGUST 01 / MANHATTAN

Bring one real workflow.
Leave with a supervised first system.

24 seats / founding cohort