# Human and AI Responsibility Map

Make task boundaries, approval, and escalation explicit.

Built Into Work | Vic Akosile | Version 1.0 | October 5, 2026

For: Process owners, domain experts, and delivery teams
Use when: When redesigning a workflow or preparing a handoff
Suggested time: 25 minutes

Use synthetic or authorized information. Keep sensitive work in approved tools. A named person owns each decision. These tools support discussion and planning; they are not validated assessments.

## 1. Assign responsibility by step

Record the work at the level where a person can tell whether the system acted within its boundary. Avoid assigning accountability to AI.

### AI may

Which bounded action is permitted?

Your response:

________________________________________________________________

### AI may not

Which action is outside its authority?

Your response:

________________________________________________________________

### Human review

Who checks what, using which evidence?

Your response:

________________________________________________________________

### Approval owner

Who authorizes the consequential action?

Your response:

________________________________________________________________

## 2. Specify exceptions

Write a trigger and a response rather than a general promise of human oversight.

### Escalation trigger

What observed condition requires a person?

Your response:

________________________________________________________________

### Escalation destination

Who receives the case, and what context do they need?

Your response:

________________________________________________________________

### Stop and fallback

Who can pause the workflow, and how does work continue?

Your response:

________________________________________________________________

### Review cadence

Who checks whether boundaries still fit the work?

Your response:

________________________________________________________________

## Completed companion example

All organizations, workflows, numbers, and results in these examples are fictional and provided for learning. Targets are proposed decision criteria, not validated standards or promised outcomes.

**AI may**
Find policy passages and draft a cited explanation.

**AI may not**
Decide eligibility, access personal files, change a record, or send an answer.

**Human review**
Policy specialist checks the current policy, citation support, and case category.

**Approval owner**
Service-desk analyst sends only after specialist approval during the pilot.

**Escalation trigger**
Missing policy, conflicting sources, individual eligibility, or a request for restricted information.

**Escalation destination**
Policy specialist receives the question, cited excerpts, and the reason the draft was withheld.

**Stop and fallback**
Service-desk lead pauses drafting; analysts return to approved manual policy lookup.

**Review cadence**
Owner reviews exceptions weekly during the pilot and after any source change.

## Working record

Owner:
Review date:
Decision:
Evidence still needed:

Original guide: https://builtinto.work/resources/human-ai-responsibility-map