BAT Challenge

Can your AI-assisted workflow be reconstructed later?

The BAT Challenge invites people working with AI-assisted processes to describe a real or representative workflow and test whether its evidence trail would stand up under later review.

Accountable AI
Challenge

Describe where AI output enters a workflow and what evidence remains around the human decision.

Methodology Context

BAT means Backward Accountability Traceability: starting from an outcome and tracing back through AI contribution, review, evidence, and rationale.

Workflow Examples

Case triage, clinical or research review, compliance checks, grant scoring, HR screening, claims handling, customer support, and policy drafting.

Anonymity

Submit without naming your organisation, clients, patients, customers, or confidential systems. Representative workflows are welcome.

Research Rationale

The aim is to understand where AI use creates accountability gaps between policy, live work, evidence capture, and review.

Introduction

BAT looks backwards from the moment accountability is needed.

Many AI governance conversations focus on the tool. BAT focuses on what happened when someone relied on the output.

Who relied on it, what was checked, what evidence was available, and whether the final decision can be explained by someone who was not present at the time.

BAT is an active research methodology and remains under development.

Use the form below to share one workflow. It can be live, planned, paused, or hypothetical, as long as it reflects a realistic operational pattern.

Looking for a structured review rather than contributing to the research corpus? See the AI Workflow Reliance Review.

Submit a Workflow

Take the BAT Challenge.

This challenge collects workflow examples to test and challenge BAT. Please avoid confidential, personal, or identifying details.

Use of Submissions

How your submission supports BAT research.

Submissions will be reviewed to identify recurring patterns in AI workflow evidence, human review, accountability handoffs, and reconstructability gaps. Insights may inform future writing, talks, research notes, and development of the BAT methodology.

BAT is an active research methodology. Participation helps test whether the method reflects operational reality across different sectors and workflow types.