Field Lab · Field Council

Bring field judgment to production AI.

The Field Lab studies how AI systems move from demo to deployment. Each lab starts with a real customer-shaped problem, defines a reliability bar, builds the smallest useful system, measures what breaks, and publishes the field report.

The Field Council is for people who have seen AI systems in the field: forward deployed engineers, applied AI architects, customer engineers, solutions architects, founding engineers, AI infra operators, and enterprise AI builders.

You do not need to represent your company. You do not need to share confidential details. You can contribute anonymously. What matters is field judgment.

Why this exists

Most AI content stops at the demo. The Field Lab focuses on what happens after that:

  • messy customer requirements
  • weak or stale evidence
  • eval gaps
  • enterprise permissions
  • cost and latency tradeoffs
  • workflow ownership
  • escalation paths
  • deployment readiness
  • production failure modes

The goal is to turn field lessons into reusable artifacts: reliability bars, checklists, rubrics, evals, runbooks, and reports that serious AI builders can use.

Ways to participate

Field Reviewer

Review a reliability bar, artifact, or field report.

The lightest way to contribute. You might react to a deployment checklist, a web-agent verification rubric, an eval plan, or a builder report.

Time
10-15 minutes
Output
Anonymous or attributed reviewer note
Best for
FDEs, applied AI architects, AI infra engineers, senior builders

Example asks:

  • Does this reliability bar reflect what you see in real deployments?
  • What failure mode is missing?
  • Would this artifact be useful to a field team?
  • What would make this safer before customer rollout?

I can review a reliability bar

Field Note Contributor

Share one practical lesson from field AI work.

Written or audio, anonymous or attributed. We turn your answers into a short field note and share a draft before publishing.

Time
20-30 minutes, async
Output
Published Field Note
Best for
FDEs, customer engineers, applied AI architects, founding engineers, solutions architects

Field Notes usually cover:

  • what your role actually looks like
  • what breaks between demo and deployment
  • how you scope ambiguous customer problems
  • what ready to ship means in practice
  • what skills matter most in the field
  • one lesson other AI builders should copy

I can contribute a field note

Lab Mentor

Help builders during a Field Lab.

Lab Mentors support a cohort working on a production-shaped AI problem. This can be a short office-hours session, async feedback, final report review, or demo-day judging.

Time
45-60 minutes
Output
Builder feedback, report review, or final judging
Best for
Senior engineers, FDEs, startup operators, AI architects, product-minded technical leaders

Mentors help builders think through:

  • architecture tradeoffs
  • eval design
  • customer-readiness
  • failure modes
  • trust and safety
  • cost and latency constraints
  • what would need to change before production

I can mentor a lab

Problem Scout

Submit a recurring field problem.

If you keep seeing the same AI deployment problem across customers, teams, or startups, submit it to the Problem Bank.

Time
5-10 minutes
Output
An anonymized field problem
Best for
Founders, operators, FDEs, customer engineers, consultants, AI platform teams

Good problems are specific enough to test:

  • Can a web-grounded agent qualify design-partner leads without unsupported claims?
  • Can an internal RAG assistant handle permissions and citations well enough for rollout?
  • Can an agent detect when to stop, verify, or escalate?
  • Can an AI workflow turn messy customer calls into CRM updates without corrupting records?

I have a field problem to submit

Artifact Reviewer

Improve a checklist, rubric, eval, or runbook.

The Field Lab publishes reusable artifacts for production AI teams. Reviewers help make them sharper before they are widely shared.

Time
10-20 minutes
Output
Improved artifact
Best for
People with practical deployment experience

Artifacts may include:

  • AI Deployment Readiness Checklist
  • Agent Escalation Checklist
  • Web-Grounded Claim Verification Rubric
  • RAG Reliability Bar
  • Demo-to-Production Risk Review
  • Field Report Template
  • Founder-Facing Trust Report Template

I can review an artifact

What contributors get

This is a lightweight contribution system, not a formal employment or advisory role. When allowed, contributors receive:

  • public credit as a Field Reviewer, Field Note Contributor, Lab Mentor, or Artifact Reviewer
  • a link to their profile, company, or project
  • early access to lab reports and artifacts
  • invitations to small reviewer sessions and Field Lab events
  • priority access to selected workshops
  • occasional sponsor-provided swag, credits, or prizes
  • a place on the Field Council thank-you page

Anonymous contribution is always available. If your company has policies around public participation, we can keep your name, employer, customer details, and architecture details private.

What we do not ask for

  • confidential customer information
  • private company strategy
  • unreleased product details
  • speaking on behalf of your employer

The best contributions are practical and generalizable:

  • a failure mode you see often
  • a reliability bar you would improve
  • a pattern that works in the field
  • a checklist you wish more teams used
  • a deployment mistake builders should avoid

Why join

Production AI needs more field judgment. The industry has enough demos, launch posts, and abstract takes. What is missing is a public record of how AI systems actually survive contact with customer workflows, enterprise constraints, messy data, and real users.

The Field Council helps turn that knowledge into something reusable. If you have deployed AI systems in the field, your judgment can help builders avoid the same mistakes.

Get involved

Send a short note with one line: "I can review a reliability bar." "I can contribute a field note." "I can mentor a lab." "I have a field problem to submit." "I can review an artifact." That is enough to begin. You can contribute publicly, privately, or anonymously.

New investigations and reviewer asks land in the weekly newsletter. Subscribe.