# TTC-115 — Workplace task design

Redesign a workplace task as explicit inputs, transformations, checks, decisions, and handoffs before deciding which bounded steps AI should assist.

Level: practitioner · Version: 1.0.0 · Last reviewed: 2026-10-03
Review status: reviewed

## Learning outcome
Map and redesign a work process so AI assistance has a bounded purpose, owner, evidence trail, and failure path.

## Explanation
Automating a vague process usually preserves its confusion. Observe the current work, identify its customer or user, trigger, inputs, exceptions, decisions, outputs, and accountable owner. Separate deterministic steps from judgment and distinguish assistance from execution. Remove unnecessary data and duplicate handoffs before adding AI. For each assisted step, define source, expected artifact, acceptance check, failure route, and measure that reflects the real outcome—not merely speed or volume. Pilot with synthetic or low-risk cases and include the people affected by the change.

## Worked fictional example
Fictional case: a cooperative's meeting workflow becomes: approved notes in, draft decisions and questions out, coordinator checks names and dates, chair approves, then the existing system distributes. The assistant cannot infer absent commitments or send the minutes; turnaround and correction rate are measured together.

## Reusable exercise
Map one fictional recurring process with swim lanes for person, AI assistance, deterministic system, and external actor. Mark data, decisions, exceptions, approvals, and failure recovery. Remove one needless step, then design a five-case pilot including ambiguity and refusal.

## Observable success criteria
- Every step has an owner, input, output, acceptance condition, and exception route.
- AI assistance is used only where its error mode and human check are explicit.
- Pilot measures include quality, correction burden, affected-user impact, and failures—not speed alone.

## Limitations
- A process map can miss informal work, power differences, accessibility needs, and impacts visible only to affected people.
- Efficiency evidence from a pilot does not prove that broader deployment is fair, safe, or worthwhile.

## Next review
Workplace evidence shows a material unmeasured impact or the accountability model changes.; A cited primary source is materially revised, replaced, or becomes unavailable.; Repeated learner results show that the exercise or success criteria are ambiguous.

## Copyable material
```text
# TTC-115 — Workplace task design
Objective: Improve one end-to-end work outcome with bounded, reviewable AI assistance.
Procedure: Map trigger, actors, inputs, data, transformations, decisions, exceptions, outputs, checks, and accountable handoffs before tool selection.
Required evidence: Retain baseline, redesigned map, pilot cases, quality and correction measures, affected-user feedback, and decision log.
Boundaries: Do not automate unclear authority, conceal labor, expand data access, or let throughput stand in for outcome quality.
Completion test: The pilot meets its quality and boundary criteria and failed cases return safely to an accountable owner.
Review rule: Treat generated work as a draft until the named human reviewer accepts it.
```

## Primary sources
- National Institute of Standards and Technology: Artificial Intelligence Risk Management Framework (AI RMF 1.0) — https://www.nist.gov/itl/ai-risk-management-framework
- European Union: Regulation (EU) 2024/1689 (Artificial Intelligence Act) — https://eur-lex.europa.eu/eli/reg/2024/1689/oj

Canonical URL: https://teachthecompany.com/school/workplace-task-design/