A public library of sanitized AI workflow problems from startups and operators, turned into buildable briefs with a stated reliability bar.
Submitting or building a problem does not guarantee mentorship, review, publication, referrals, funding, or hiring outcomes.
Featured brief
Build an AI growth agent that helps an early-stage founder discover, reach, convert, and activate the right first users with evidence, not guessing.
A library of real AI workflow problems scoped to be buildable, testable, and useful as portfolio-grade artifacts. Each brief includes the pain, the workflow, the inputs and outputs, evaluation ideas, the boundaries, and a reliability target. The goal is to help builders work on production-shaped problems instead of random demos.
This month's failure brief
Starbucks terminated its AI inventory counting program across all North American stores nine months after a chain-wide launch. The image-recognition system, built by NomadGo, frequently miscounted and mislabeled stock, confusing milk types or failing to detect items entirely, and stores reverted to manual counting. An internal memo reviewed by Reuters read: Effective immediately, Automated Counting will be discontinued.
A vision model that performed under controlled conditions was deployed at scale into thousands of variable retail environments, where lighting, placement, lookalike packaging, and human workflow did not match the training distribution. Instead of catching shortages, the system added a new layer of inaccuracy to the supply chain it was meant to fix.
Eval-production skew at chain scale: the gap between accuracy in a demo environment and consistency under real-world variance was never measured before full deployment. A staged rollout with a shadow-mode comparison against manual ground truth would have surfaced the gap at ten stores instead of all of them.
Sources Source 1
Editorial note: Reuters reporting and the company statement are the primary sources; attach the Reuters link in the editorial verification pass before publish.
This month's build: close the gap that caused it.
Build an evaluation harness that measures how a vision-based inventory counter degrades from controlled to messy real-world conditions, and gates deployment on consistency under variance, not demo accuracy.
Reliability focus eval-production skew, variance testing, deployment gating
View BriefBuild a verifier that checks every citation in an AI-drafted legal document against a real source and flags or strikes anything it cannot ground.
Reliability focus citation grounding, fabrication detection, export gating
View BriefBuild an agent that classifies inbound support tickets, drafts a suggested reply, escalates uncertain cases, and explains why it routed each one.
Reliability focus classification, calibrated escalation, explainability
View BriefSeen this failure in your own stack? Submit your version.
Submit your versionBuild a retrieval copilot that answers technical buyer questions from product, security, and integration documents, with citations and honest gaps.
Reliability focus retrieval, citation, calibrated refusal
View BriefBuild a verifier that checks every citation in an AI-drafted legal document against a real source and flags or strikes anything it cannot ground.
Reliability focus citation grounding, fabrication detection, export gating
View BriefBuild an agent that turns raw meeting notes into owners, action items, risks, and follow-ups, grounded in what was actually said.
Reliability focus extraction, attribution, grounded refusal
View BriefBuild a context layer that reconciles facts scattered across code, chat, docs, CRM, and product tools into one owned, cited, conflict-aware source of truth.
Reliability focus source reconciliation, conflict detection, ownership and freshness
View BriefBuild middleware that routes each request between small and large models by predicted complexity, measuring the tradeoff across cost, latency, and quality.
Reliability focus routing policy, quality floor, measurement
View BriefBuild a merchant copilot that routes messy requests across catalog, sales, operations, and support while preventing unsafe mutations.
Reliability focus routing, safe actions, layered evals
View BriefBuild a guardrail layer that detects when chained AI actions start compounding errors and halts the chain before a cascade causes real damage.
Reliability focus circuit breaking, invariant checks, safe halt
View BriefBuild an agent that classifies inbound support tickets, drafts a suggested reply, escalates uncertain cases, and explains why it routed each one.
Reliability focus classification, calibrated escalation, explainability
View BriefNo briefs match those filters. Clear a filter to see more.
A problem is accepted when it passes all seven questions.
Problems requiring confidential data, production credentials, regulated workflows, vague business goals, or heavy integration work are not accepted.
A good problem is specific, painful, buildable with sample or synthetic data, and evaluable without production access. Selected problems may be rewritten into public briefs, and submitting creates no contractor, delivery, or hiring relationship.
Submit a ProblemBuilders use Field Briefs to create portfolio-grade artifacts. Some problems may later become cohort capstones or teardown candidates.
Join the builder interest list